<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ben Resnic on Medium]]></title>
        <description><![CDATA[Stories by Ben Resnic on Medium]]></description>
        <link>https://medium.com/@bsresnic?source=rss-7db84f61f81f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*XRrf3p2vqRd4hqL2</url>
            <title>Stories by Ben Resnic on Medium</title>
            <link>https://medium.com/@bsresnic?source=rss-7db84f61f81f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 20:11:40 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bsresnic/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Adjusted Run Efficiency Score (ARES): College Baseball’s Most Predictive Rating System]]></title>
            <link>https://medium.com/@bsresnic/adjusted-run-efficiency-score-ares-college-baseballs-most-predictive-rating-system-251b72557a36?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/251b72557a36</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Thu, 30 Apr 2026 21:00:49 GMT</pubDate>
            <atom:updated>2026-04-30T21:00:49.271Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WasBuhbyO6xitfeKZxAzqA.png" /></figure><p><a href="https://huggingface.co/spaces/benresnic/ARES_methodology">Adjusted Run Efficiency Score * Projected runs better than an average team at a neutral site per 9 innings - a Hugging Face Space by benresnic</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=251b72557a36" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Balls are Juiced in MLB the Show]]></title>
            <link>https://medium.com/@bsresnic/the-balls-are-juiced-in-mlb-the-show-42cea3cfa5e8?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/42cea3cfa5e8</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Tue, 14 Apr 2026 18:18:27 GMT</pubDate>
            <atom:updated>2026-04-14T18:25:47.855Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9jOsebmJ-yoN_zlM" /></figure><p><a href="https://drive.google.com/file/d/1WvCo8UtRZnrlyz-Wmiattp6-gk5WRm0h/view?usp=sharing">The_Balls_Are_Juiced_in_MLB_The_Show.pdf</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=42cea3cfa5e8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Red Sox Select Kyson Witherspoon with the 15th Overall Pick]]></title>
            <link>https://medium.com/@bsresnic/red-sox-select-kyson-witherspoon-with-the-15th-overall-pick-b88a5387a586?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/b88a5387a586</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Mon, 14 Jul 2025 00:17:32 GMT</pubDate>
            <atom:updated>2025-07-14T01:00:42.142Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*qu5hZ6oQ9oNdyugq" /></figure><p>A RHP out of Oklahoma, Witherspoon is an electric arm with a fastball that can touch 99. It’s a funky release, where he hitches a little after leg lift and fires a cannon over the top. Pre draft I ranked <a href="https://medium.com/@bsresnic/top-30-college-prospects-ahead-of-the-mlb-draft-20e1f0b4270a">Witherspoon</a> as my 7th overall college prospect, and 6th best college pitcher.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mYVS6eVoU4xw3c3GUvjsqA.png" /></figure><p>The fastball averages 19 inches of induced vertical break and plays to a -4.4 vertical approach angle. I project it to lose some carry with the transition to an MLB ball, but still a 60 fastball nonetheless. He saw some trouble with the fastball getting hit hard in the zone, and didn’t generate a ton of whiff compared to his other pitches.</p><p>Witherspoon also throws a cutter and slider at 89 and 85 mph respectively. They have a tendency to overlap shapes, and this is something I’d like to see the Red Sox develop in the coming years. The cutter averages 6 inches of rise and 4 inches of glove side break, while the slider has a similar movement except stays around zero vertical break. Kyson should be able to keep the cutter up and throw it harder in the low 90’s, maybe with around 6–7 more inches of rise but less cut. This would get better separation from the slider but also help blend with the fastball. Both pitches receive a 55 grade.</p><p>His curveball averages 79 mph with -17 inches of induced vertical break and 10 inches of glove side. It’s his least used breaking ball at 7% and could see an uptick in pro ball. He generates an insane 53% whiff but doesn’t zone it well which leads to few chases. It grades as a 55.</p><p>Lastly, Witherspoon threw a changeup sparingly to lefties, and is the pitch in his arsenal that will need the most work. It’s 89 mph but the shape has no deviation from his arm slot. Witherspoon has a slight supination bias so I imagine the Red Sox can get him to cut the changeup for more depth, which will help create a discrepancy from the fastball that it plays off of. As currently constructed it’s a 45 offering, but look for the Red Sox to do an overhaul here.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/908/1*4fK2w9_zyssPZB_ZOs92Bw.png" /></figure><p>The command is a 55. Witherspoon only walked 6.5% of hitters this spring.</p><p>I’m content with this pick at 15. While I would have preferred Gage Wood out of Arkansas, who I ranked as my second overall college prospect, Witherspoon has the potential to be the #2 behind Garrett Crochet for a long time. I love Witherspoon’s archetype of pitcher, whose arsenal plays up because he can throw the fastball high, and then tunnel the breaking balls off of it. Good selection by Craig Breslow in this spot, getting 55 FV from the middle of the first round.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b88a5387a586" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ranking the Top 30 College Prospects Ahead of the MLB Draft]]></title>
            <link>https://medium.com/@bsresnic/top-30-college-prospects-ahead-of-the-mlb-draft-20e1f0b4270a?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/20e1f0b4270a</guid>
            <category><![CDATA[mlb]]></category>
            <category><![CDATA[prospects]]></category>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[draft]]></category>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Sun, 13 Jul 2025 16:27:10 GMT</pubDate>
            <atom:updated>2025-07-14T00:24:37.564Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*VsBMO_RSp8a31j1J" /></figure><p>It’s draft day! In what is probably a weaker draft than most years, there are future all stars in this class waiting to be picked. I’ve written up a short excerpt on my top 30 college prospects that will be selected in Atlanta, including grades and predictions on where they’ll be selected.</p><p><strong>30. Kane Kepley (CF, <em>North Carolina</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*iaD1VEcvO_A3QQYK" /></figure><p>Kepley clearly has the best outfield glove out of everyone at the top of this draft with platinum glove potential. Combine that with elite plate discipline and contact skills, and he becomes an intriguing prospect. He has a small frame and is unlikely to develop any power. Kane doesn’t generate good launch angles and doesn’t pull the ball in the air which can make low raw power numbers play up. Jumps are below average in center, but posts great numbers due to his closing speed. With some work defense could jump from 60 grade to even higher.</p><p><strong>Hit: 45, Game Power: 30, Raw Power: 35, Fielding: 60+, FV: 45</strong></p><p>Prediction: Kansas City Royals (2.61)</p><p><strong>29. Zach Root (LHP, <em>Arkansas</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ywEZxYCGCxyoYYKr" /></figure><p>Root’s breaking balls induced an insane amount of whiffs leading to his 92nd percentile K%. He has the ability to add velocity on all his pitches, most importantly the fastball. Probably needs to zone his pitches more at higher levels, was able to get away with it in college posting only an 8% BB rate. Root’s bad fastball shape will get worse with an MLB ball. He couldn’t generate whiffs, and could benefit from making his current “fastball” a sinker and adding a true 4 seamer with some cut and ride, if possible. I see a team like Cincinnati snagging him in the second round and giving him a makeover.</p><p><strong>Fastball: 35, Changeup: 50, Slider: 50, Curveball: 55, Command: 50, FV: 45</strong></p><p>Prediction: Cincinnati Reds (2.51)</p><p><strong>28. Charles Davalan (LF, <em>Arkansas</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*FLe8KRPc3e7iFjBn" /></figure><p>Davalan hit leadoff for Arkansas and boasts a 60 grade hit tool derived from his ability to avoid strikeouts and hit for average. He makes contact deep behind the plate and hits the ball to the opposite field often on a line. Plays above average defense in LF despite a really bad error in Omaha against LSU. With a small frame Davalan is unlikely to develop real home run power. He hits the ball on the ground more often than you’d like and isn’t a blazing runner. Just a very solid all around player that knows how to win.</p><p><strong>Hit: 60, Game Power: 45, Raw Power: 40, Fielding: 55, FV: 45</strong></p><p>Prediction: Pittsburgh Pirates (2.50)</p><p><strong>27. Mason Neville (CF, <em>Oregon</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*MDF5ZLzLk8iiDqR3" /></figure><p>Neville crushes righties with an OPS over 1.400 and posts good contact rates as well. Pulls the ball in the air well and his raw power plays even higher in game when he connects. He walks often and doesn’t chase. However he struggles mightily vs lefties and non fastballs. Whiffs at a high rate all over the zone, makes bad swing decisions, and doesn’t make quality contact. Strikes out 42% of the time vs lefties and only walks in 7% of PA. Will have a lot to develop in the minors against everything but righty fastballs. That being said the potential is potent and could be an overslot guy early, but I could also see him slipping down the board a fair bit. He will probably get fed spin in the majors and struggle, but he could be a righty killer.</p><p><strong>Hit: 35, Game Power: 65, Raw Power: 60, Fielding: 40, FV: 45</strong></p><p>Prediction: New York Yankees (Competitive Balance Round A, 39)</p><p><strong>26. Henry Ford (1B/OF, <em>Virginia</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*EKdU5HwmwBlBHHMH" /></figure><p>Ford hits lefties well, doesn’t strike out a ton, and pulls the ball well. Average power now could play up as he develops. Is also a very solid defender in left field. Hits velocity well with a .430 wOBAcon vs 93+. Below average plate discipline formed by a 33% chase rate leads to few walks. Whiffs often against non-fastballs and contact quality against them is a big concern.</p><p><strong>Hit: 45, Game Power: 50, Raw Power: 50, Fielding: 55, FV: 45</strong></p><p>Prediction: Los Angeles Angels (2.47)</p><p><strong>25. Anthony Eyanson (RHP, <em>Louisiana State</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*F3Sc415L22M6xNTx" /></figure><p>Eyanson came onto the scene this year after being a great number 2 guy for the national champion Tigers. An average fastball in college will likely not play well in pro ball. Eyanson has an over the top slot but doesn’t generate elite carry on the fastball, and what projects as 15–16 IVB with an MLB ball won’t generate results especially with that -6 VAA. His best pitches are his short gyro slider at 84 and big breaking curveball at 78. Both graded out well compared to the average college pitch but they’ll need to add velocity for Eyanson to stick in the bigs.</p><p><strong>Fastball: 45, Slider: 55, Curveball: 50, Command 50, FV: 45</strong></p><p>Prediction: Tampa Bay Rays (Competitive Balance Round A, 42)</p><p><strong>24. Chase Shores (RHP, <em>Louisiana State</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bMOSdl4q6DnOB4Nb" /></figure><p>The biggest question around Shores is if he can make it as a starter in MLB. After pitching in 23 games for LSU this spring and only making 9 starts, Shores could probably make an impact out of the pen this fall for a playoff team if they absolutely needed it. For his development however, trying him as a starter would squeeze out the most value. Sitting 100 as a closer in college would have to turn into 96–97 starting games, the command would have to improve; he can’t just throw high 90’s down the middle anymore. Shores would also benefit from more distinct shapes on his arm side pitches. His fastball, sinker, and changeup all blend together, and I think he needs just a little more carry on the four seamer. Shores also offers a deadly 87 mph slider which made batters whiff at an ungodly 42% clip. A bit of a wild card, a team like the Dodgers could snag him in between round 1 and 2 for some value.</p><p><strong>Fastball: 50, Sinker: 55, Slider: 60, Changeup: 55, Command: 40, FV: 45</strong></p><p>Prediction: Los Angeles Dodgers (Competitive Balance Round A, 40)</p><p><strong>23. Andrew Fischer (1B, <em>Tennessee</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bML8_hCXQcRb7ZDP" /></figure><p>Looking at Fischer’s numbers it’s clear why he’ll be the first 1B of the board. The only major reason he’s not higher is the position he plays, and some whiff and quality of contact issues vs better quality pitching. Fischer’s swing is an extreme uppercut, which helped him hit 56 homers in college, but also had him whiffing up in the zone vs fastballs. 69% zone-contact vs 93+ up in the zone, and 38% whiff overall will have to improve at higher levels. It’s definitely a little nit picky but pitchers in MLB will attack him up and have success. Besides that, the only other knock is his defense at first base. 100th percentile xwOBA and xSLG are absolutely first round numbers.</p><p><strong>Hit: 40, Game Power: 60, Raw Power: 60, Fielding: 35, FV: 50</strong></p><p>Prediction: Houston Astros (1.21)</p><p><strong>22. Luke Stevenson (C, <em>North Carolina</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KRLrNrzZbSc27u3_" /></figure><p>Stevenson is a really hard player to project. On one hand his 70 grade defense at catcher paired with 60 grade raw power and a 22% walk rate make him primed to be top 10 on this list, and on the other hand he can’t stop whiffing and has some slightly concerning quality of contact issues. Stevenson does a great job not chasing, so all of his whiffs come in the zone, and he strikes out a ton. Stevenson would probably go near the top of the first round with a 40 hit tool, but in MLB pitchers will abuse him with breakers and have success.</p><p><strong>Hit: 30, Game Power: 55, Raw Power: 60, Fielding: 70, FV: 50</strong></p><p>Prediction: Kansas City Royals (1.23)</p><p><strong>21. Ike Irish (OF/C, <em>Auburn</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bJPv2wPA08kImPPk" /></figure><p>I’m polarised on Ike Irish. While he showed he can crush college pitching, he has a similar problem to Stevenson where he whiffs a lot in the zone, but the strikeout problem isn’t there and overall makes average swing decisions. 60 grade raw power will play down to 55 game power if he can’t start pulling the ball in the air. This spring Irish hit far too many ground balls and saw some launch angle issues. My biggest knock on Irish is his defense, which for me would slide him out of a first round selection. Irish split time between catcher and the outfield at Auburn, and posted a negative framing run value behind the fish, and -5 outs above average between left and right field. He let 6 balls drop with a catch probability greater than 80%, went 4/8 on 2 star catches (70–90%), and 0/10 on 4 star or harder catches (&lt; 50%). It’s a small sample, but the jumps weren’t good on video and he doesn’t have the closing speed to overcome that.</p><p><strong>Hit: 35, Game Power: 55 Raw Power: 60, Fielding: 30, FV: 50</strong></p><p>Prediction: Minnesota Twins (1.16)</p><p><strong>20. Patrick Forbes (RHP, <em>Louisville</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*a0hBZ_PfOjiZPtb5" /></figure><p>Forbes has one of the fresher arms in this year’s draft, having pitched 71.1 innings this spring, but only 69.1 IP in college/summer ball before that. Forbes was a top prospect of high school for the 2022 draft, but as a hitter, so there isn’t a lot of mileage on him. Forbes was a 2 pitch guy this spring at Louisville, but has shown he can develop more as he progresses. He offers a mid-90’s fastball that gets a flat VAA from a low slot, and a short sweeper at 82 mph. He got hit hard in the air in college, and a true sinker shape should be in his future to help limit that, and it’ll also help the fastball. He needs to zone his pitches better and lower his walk rate, but the right organization could get a stud in the right situation.</p><p><strong>Fastball: 50, Slider: 55, Cutter: 50, Changeup: 50, Command: 45, FV: 50</strong></p><p>Prediction: Baltimore Orioles (Competitive Balance Round A, 30)</p><p><strong>19. Marek Houston (SS, <em>Wake Forest</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PHeDSnAWmIAULztn" /></figure><p>I’m not as high on Marek Houston as some other draft boards. The big discrepancy in the public view comes from his defense. Most major publications have him with at least a 55 fielding, such as MLB Pipeline with a 60 grade, but I’m sticking with him at 45. He fared fine on balls with higher out probabilities, but lacked an ability to make the difficult plays. Houston converted only 2 out of 39 5 star plays (&lt; 25%), and 8/26 on 4 stars (26–50%). I agree with the masses that the glove has the potential as his motions look fluid, but just isn’t where it needs to be at the moment. As for the bat Houston posted great numbers in Winston-Salem with an OPS north of 1.000, but drastically overperformed his expected numbers, beating his xBA and xwOBA by 50 points and his xwOBAcon by 80.</p><p><strong>Hit: 55, Game Power: 50, Raw Power: 45, Fielding: 45, FV: 50</strong></p><p>Prediction: Texas Rangers (1.12)</p><p><strong>18. Gavin Kilen (MIF, <em>Tennessee</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ntbdvQmjjhMcdlTq" /></figure><p>Kilen is the prototypical great bad bad glove in this draft. He split time between second base (-1.2 OAA) and shortstop (-3.3 OAA) this spring in Knoxville. Despite a below average chase rate Kilen made good swing decisions at the plate, and backed that up by walking and striking out at above average clips. Kilen hits righties for power, but posted a low ISO vs lefties. He also follows a common trend of chasing vs non fastballs, inducing bad contact quality. Vs 93+ he has a .540 wOBAcon in 28 batted balls, though the contact rates drop. Look for Kilen to go to a team looking for an already polished college bat, like the Red Sox, who drafted Kilen out of high school in 2022. Kilen didn’t sign and went to Louisville.</p><p><strong>Hit: 55, Game Power: 50, Raw Power: 50, Fielding: 40, FV: 50</strong></p><p>Prediction: Boston Red Sox (Competitive Balance Round A, 33)</p><p><strong>17. Caden Bodine (C, <em>Coastal Carolina</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*tBkbUSurmRDYGRCG" /></figure><p>Bodine might be one of the most polished hitters in this entire class. Great hit tool but power is limited, but he could grow into it. He shoots the ball the other way vs fastballs and pulls slower stuff in the air, but it doesn’t lead to slug. Bodine won’t have a problem with the transition into the majors with a 97% zone-contact rate vs 93+ and a 52% hard hit rate. Bodine won the Buster Posey award as the best college catcher this season, and backs it up with the highest framing run value in this class, and 4th best in the country. That being said, some of the framing numbers you’ll see in college can be a little overblown. While framing run value adjusts for college umpires being terrible, there isn’t a reliable way to separate the catcher being good at framing from an umpire making a bad call (yet, maybe coming soon). Bodine has an average-ish arm but throws out runners at a good clip due to his accuracy. However, it’s hard to say no to a switch hitting catcher with elite defense and hit tool.</p><p><strong>Hit: 60, Game Power: 45, Raw Power: 40, Fielding: 70, FV: 50</strong></p><p>Prediction: New York Mets (Competitive Balance Round A, 38)</p><p><strong>16. Ethan Conrad (OF, <em>Wake Forest</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Pj1c656IH5BLv9AK" /></figure><p>A shoulder injury limited Ethan Conrad to only 21 games and under 100 plate appearances this spring for the Demon Deacons, but in those games it’s hard to find a college hitter who produced better numbers. Conrad produced 100th percentile xwOBA and xSLG, an 18% barrel rate, and 1/3 of his fly balls were home runs. Vs high velocity Conrad hit .500 and saw no decrease in plate discipline metrics. He played a slightly below average CF, but with such a small sample size (18 plays), I won’t knock it because the only reason he’s below average is one play with a 90% catch probability that landed. The only reason Conrad isn’t ranked higher is because his season was cut short, don’t be surprised if a team reaches for him early on Sunday.</p><p><strong>Hit: 55, Game Power: 50, Raw Power: 60, Fielding: 50, FV: 50</strong></p><p>Prediction: Baltimore Orioles (1.19)</p><p><strong>15. Kyle Lodise (SS, <em>Georgia Tech</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6IOcZcpv0m2OFzux" /></figure><p>The lesser known Lodise brother, but just as good. Kyle has reverse splits hitting from the right side, which can actually be a good thing as the hit tool vs lefties should be easier to develop than the other way around. Whiff problems eat up Lodise vs spin and offspeed. While he doesn’t chase vs those pitches he doesn’t make contact, and when he does the quality drops. Lodise displays better power than expected considering his size, partly because he does a great job of pulling the ball in the air. In the field Lodise is smooth at SS, with 6.7 OAA and makes the hard plays look easy while making the routine play every time as well.</p><p><strong>Hit: 45, Game Power: 55, Raw Power: 50, Fielding: 60, FV: 50</strong></p><p>Prediction: Tampa Bay Rays (2.53)</p><p><strong>14. Cam Cannarella (CF, <em>Clemson</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*w1h1rR1SxAXcequ3" /></figure><p>If Cannarella goes in the first round it will be because a team sees him as a gold glove center fielder. He had 5 OAA in center field this spring for Clemson, excelling on balls in the 3–5 star range. With the bat Canneralla walks almost as much as he strikes out, and posts decent exit velocity numbers, but is unlikely to provide above average power at the big league level. He whiffs in the zone too much and hits the ball on the ground or a low line when he makes contact, which fits his playing style as a plus runner.</p><p><strong>Hit: 45, Game Power: 45, Raw Power: 45, Fielding: 65, FV: 50</strong></p><p>Prediction: Arizona Diamondbacks (Compensation, 29)</p><p><strong>13: Jace Laviolette (CF, <em>Texas A&amp;M</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*naQjFNOgSV76_ZL4" /></figure><p>This time last year Jace LaViolette was projected to go top 5 in this year’s draft. After a disappointing season in Aggieland where they missed the postseason, I don’t think Jace has slipped <em>that</em> much. Here’s a guy who has atrocious contact rates and doesn’t hit for average, but that’s about it. He swings and misses because of his swing, not due to chase. LaViolette walks as much as anyone in the country, and smacks the ball all over the park. The most underrated part of Jace’s game is his glove. It’s 99th percentile OAA in center, catching every single ball hit to him with a probability of over 40%. At his peak, LaViolette is Kyle Schwarber but with a glove.</p><p><strong>Hit: 30, Game Power: 65, Raw Power: 70, Fielding: 60, FV: 50</strong></p><p>Prediction: Baltimore Orioles (Competitive Balance Round A, 31)</p><p><strong>12. Alex Lodise (SS, <em>Florida State</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZleZIYvfODmNbV2o" /></figure><p>Alex Lodise will stick at shortstop and be a plus defender there for a long time. He has a well above average range on balls that no other shortstops get to, and makes the easier plays every single time. With the bat Lodise definitely has holes in his swing. He chases and whiffs way too much, and doesn’t walk to balance that out. Lodise particularly struggles with velocity up in the zone and spin away from him. He provides plus power from a premium position, and when he makes contact the quality is top tier. Lodise is this high primarily because of his defense at SS, and if he can develop the hit tool a bit more he’ll be well worth the pick.</p><p><strong>Hit: 40, Game Power: 60, Raw Power: 55, Fielding: 60, FV: 55</strong></p><p>Prediction: Cleveland Guardians (1.27)</p><p><strong>11. Wehiwa Aloy (SS, <em>Arkansas</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*5ap9xpYDivp0U1eW" /></figure><p>Aloy is a very similar player to Alex Lodise. Plus power at shortstop, great defender, but whiff and chase problems limit his upside. The hit tool isn’t great, but there is so much value in playing shortstop well, and Aloy could easily run into 20–25 homers with his raw power. Just like Lodise, Aloy whiffs in fastballs up, and his in zone whiff map on sliders is under 70% in the bottom and middle third. He shoots the ball to the opposite field a lot for extra bases, and with a wood bat the ball won’t carry and those will turn into outs. The Golden Spikes winner is a 20–80 outcome player. I could see him flaming out in AAA getting spun to death, but I could also see him win a gold glove at shortstop and lead the position in homers. I lead towards the latter.</p><p><strong>Hit: 40, Game Power: 55, Raw Power: 65, Fielding: 60, FV: 55</strong></p><p>Prediction: Milwaukee Brewers (1.20)</p><p><strong>10. Devin Taylor (OF, <em>Indiana</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZWcx6F_DFjm3zYU7" /></figure><p>Devin Taylor is one of my 2 draft crushes this cycle. He does everything well, has a high floor, and I’d bet a lot of money he’s a solid, every day MLB outfielder at the minimum. Taylor was by far the best hitter in the Big 10 this year. He walks more than he K’s, and is 100th percentile in xwOBA, xBA, xSLG and Average Exit Velocity. He plays a good left field too. Won’t make the flashy play (0/13 on 5 star plays), but was above average on 4 stars and made every single catch on balls easier than that. He has a large frame and I think he’ll end up being an average left fielder. Taylor also posted a casual 2.300 OPS vs 93+ while seeing no dip on contact rates. Someone at the end of the first round is going to get a steal later today.</p><p><strong>Hit: 55, Game Power: 55, Raw Power: 55, Fielding: 50, FV: 55</strong></p><p>Prediction: Detroit Tigers (1.24)</p><p><strong>9. Brendan Summerhill (RF, <em>Arizona</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*JQFdAH30U2kRhQMv" /></figure><p>Summerhill saw an injury limit some playing time in Arizona this spring, but that won’t stop him from being a high pick later today. Just a polished hit tool with average power, and the potential to be an above average defender. Summerhill makes great swing decisions and excels at hitting the ball on a line. The OAA numbers from the spring aren’t pretty, but he was dealing with a bad hamstring all year and combine that with a few bad jumps, there’s nothing else to expect. Summerhill has the speed to be at least an average defender, and will continue to develop getting better jumps in the minors.</p><p><strong>Hit: 60, Game Power: 50, Raw Power: 50, Fielding: 50, FV: 55</strong></p><p>Prediction: San Francisco Giants (1.13)</p><p><strong>8. Riley Quick (RHP, <em>Alabama</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*bOSoT4efSUp1K-sk" /></figure><p>Riley Quick has flown way under the radar this draft. Quick is a high floor guy, offering five distinct pitches that produced great results at Alabama. At 6&#39;6&quot; and 250 lbs, he offers a strong heater at 97 mph, a sinker at the same velocity, a depthy changeup, and a hard cutter and slider. The slider is his best pitch that he threw to righties more than any other pitch. It’s like a mini slurve at 87, and got hitters to whiff at a 35% rate. Both fastball shapes could use some improvement. The 4 seam is deadzone at 13 inches of horizontal break and14 vertical, and that’s actually the pitch that performed the worst for him this spring by run value. The sinker needs to be separated from the fastball more as well, and he should not be throwing it to lefties as much as he does. I’d bet a lot of money that Quick sticks in the middle of a rotation at the least.</p><p><strong>Fastball: 50, Sinker: 50, Slider: 60, Cutter: 45, Changeup: 55, Command: 50, FV: 55</strong></p><p>Prediction: Detroit Tigers (Competitive Balance Round A, 34)</p><p><strong>7. Kyson Witherspoon (RHP, <em>Oklahoma</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*7kvKq6LgJwoGsmnb" /></figure><p>Kyson Witherspoon has the second best fastball in this draft, generating good carry at 96 from a unique, over the top slot. He also offers a gyro cutter and slider that can blend together at times, a steep curveball at 79, and sometimes offers a changeup just to mix things up vs lefties. I’d describe Witherspoon as electric. I love this archetype of pitcher, whose arsenal plays up because he can throw the fastball up, and then tunnel the breaking balls off of it. The changeup will need some work to throw it in MLB, whether that’s getting more depth or just improving the command on it. Witherspoon has ace potential, and don’t be surprised when he adds even more velocity.</p><p><strong>Fastball: 60, Slider: 55, Cutter: 55, Changeup: 45, Curveball: 55, Command: 55, FV: 55</strong></p><p>Prediction: Toronto Blue Jays (1.8)</p><p><strong>6. Tyler Bremner (RHP, <em>California Santa Barbara</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*eXKdudZwTdhUoaus" /></figure><p>Bremner had a slow start to the year but picked it up towards the end. He only throws three pitches: a mid-90’s fastball with outlier shape, a changeup with good run and decent depth, and a short gyro slider at 86. At the next level Bremner will need at least one more pitch. I think he can keep the current slider and has potential to throw that harder. He’s a big pronator so he might not be able to get to sweeper shapes effectively, but even if it doesn’t grade out well I don’t necessarily believe he can start in MLB with his current arsenal. That being said, the potential is there, he’s just one pitch away.</p><p><strong>Fastball: 55, Slider: 55, Changeup: 55, Command: 50, FV: 55</strong></p><p>Prediction: Boston Red Sox (1.15)</p><p><strong>5. Liam Doyle (LHP, <em>Tennessee</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pU7BBOVi2BeIKGnx" /></figure><p>With the best fastball in the draft, Doyle depended on it constantly to get outs. Doyle topped out at 100 from the left side while generating great carry. In pro ball however he can’t rely on it as much as he did in college, throwing it 60% of the time. The secondaries will need work, though I really love the splitter. It kills spin well and provides a massive change in depth for hitters. Doyle would benefit from throwing the cutter and slider harder, but I’m unsure if he physically can do that. A big concern with Doyle is how hard he will be throwing the fastball considering he’ll need to be prioritizing the longevity of making 35 starts and throwing over 200 innings in MLB. The fastball has a stark velocity drop as the game goes on, starting in the upper 90’s at times and dropping 3–4 mph by the end. The delivery looks stiff, and I can see him profiling as more of a reliever as he ages. I think the Angels take him at 2 and rush him to the majors by this time next year at the latest, because that’s what the Angels do.</p><p><strong>Fastball: 65, Slider: 55, Splitter: 55, Cutter: 50, Command: 45, FV: 55</strong></p><p>Prediction: Los Angeles Angels (1.2)</p><p><strong>4. Jamie Arnold (LHP, <em>Florida State</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NfkGVOdIw5436ucE" /></figure><p>Jamie Arnold is a bit of a wildcard. He could not hit the side of a barn door with his changeup this spring, which is why he only threw it 8% of the time. However the shape on it is so outlier, and I believe he’ll be able to control it as time goes on, as it’s a relatively new pitch for him. From a low slot on the left side, Arnold’s fastball has the appearance of rise at 94, but got hit the most out of all his pitches. To begin the season he was only throwing one breaker, the sweepy slider in the mid 80’s that served as his primary. Recently Arnold developed a cutter that helps balance out his arsenal to mix things up. He’s such a unique arm that will confuse hitters at the plate, and his success will depend on it staying that way.</p><p><strong>Fastball: 55, Sinker: 50, Slider: 60, Changeup: 65, Cutter: 50, Command: 55, FV: 60</strong></p><p>Prediction: St. Louis Cardinals (1.5)</p><p><strong>3. Aiva Arquette (SS, <em>Oregon State</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KUjUIKJ1zngsUtUl" /></figure><p>Arquette is a big guy for a shortstop but still moves well and plays the position well — reminds me of Marcelo Mayer. He has just as much power as shortstops lower on this list like Lodise and Aloy, but makes much better swing decisions and the hit tool is far more serviceable, although still a 50. As long as the whiff and chase don’t get worse as time goes on, everything about Arquette screams 2 WAR at the minimum. He runs into some weak flyouts to right field, but when he pulls the ball it goes forever. His contact quality is either really or really bad, popping up and hitting grounders more than he should, but most of the time it’s elite. There’s been a lot of buzz about the Mariners going with an arm at 3, but I think Arquette goes here in a bit of a surprise.</p><p><strong>Hit: 50, Game Power: 60, Raw Power: 60, Fielding: 60, FV: 60</strong></p><p>Prediction: Seattle Mariners (1.3)</p><p><strong>2. Gage Wood (RHP, <em>Arkansas</em>)</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wWEdE7NFqObqTBIb" /></figure><p>My second draft crush, I don’t understand why Gage Wood isn’t valued higher. I don’t care that he got hit hard in the SEC, the shapes are too good to pass on. Wood offers four pitches from the right side, though only relied on three at Arkansas. From a low release height, his mid-upper 90’s fastball gets a flat VAA and otherworldly whiff and chase rates. The curveball is the best pitch out of anyone in this class. A 70 grade might not be cutting it, for an 83 mph pitch it breaks more than just about anyone, and deviates slightly from his arm angle. The gyro slider at 89 is nothing to sneeze at either and provides a good bridge between the fastball and curve. He threw a splitter sparingly, and I see that pitch only getting better as he develops. He’s a supinator and should be able to cut the ball to generate more depth. I haven’t even mentioned that he doesn’t walk batters, and strikes out HALF of the hitters he faces. Whoever snags Gage Wood will get the steal of the draft.</p><p><strong>Fastball: 60, Slider: 60, Splitter: 45, Curveball: 70, Command: 60, FV: 60</strong></p><p>Prediction: Chicago Cubs (1.17)</p><ol><li><strong>Kade Anderson (LHP, <em>Louisiana State</em>)</strong></li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*abTit1VhKtU_kDTs" /></figure><p>Kade Anderson could give a team 6 innings of 3 run ball tomorrow if need be. The most pro ready arm in the draft this year, Anderson rose to the top of the draft rankings after being lights out in Omaha for the national title winning Tigers. Anderson throws four pitches that all grade out above average at least. He has a wiry frame at 6&#39;2&quot; and 180 lbs. A 93 mph fastball with big induced vertical break has the potential to add velocity. Anderson threw his slider at 85 in the zone constantly and batters could not stop whiffing. He also has depthy curveball at 79 and a changeup that will only get better. He might throw his pitches in the zone a bit too much, as he feels like a guy that could generate more chase with his kind of stuff. The most polished pitcher in the draft will hear his name called first by the Washington Nationals.</p><p><strong>Fastball: 60, Slider: 60, Curveball: 60, Changeup: 55, Command: 55, FV: 60</strong></p><p>Prediction: Washington Nationals (1.1)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=20e1f0b4270a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introducing Deception Runs and Deception+, a Measurement of a Pitcher’s Ability to Force a Batter…]]></title>
            <link>https://medium.com/@bsresnic/introducing-deception-runs-and-deception-a-measurement-of-a-pitchers-ability-to-force-a-batter-d161ddd7603a?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/d161ddd7603a</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Thu, 20 Feb 2025 17:55:08 GMT</pubDate>
            <atom:updated>2025-02-20T17:55:08.051Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ROyJC5xsDK0axLZA.jpg" /></figure><h3>Introducing Deception Runs and Deception+, a Measurement of a Pitcher’s Ability to Force a Batter into Bad Swing Decisions</h3><p>For batters, having good plate discipline is a skill. Choosing whether or not to swing at any given pitch is almost solely dependent on the batter’s ability to know what pitches are “good” balls to hit and which ones are “bad”. But what’s the best way to quantify it? And an even bigger question is how much of a pitcher’s ability has an affect on these decisions? This article will be an introduction to a new metric called Deception Runs. Deception Runs simply measure how good a pitcher is at making batters make the wrong swing decision, while also accounting for how bad of a decision it was.</p><p><strong>What is a “correct swing”?</strong></p><p>In order to evaluate swing decisions in the first place, we have to create a new metric. Correct Swing Percentage is what it’s called, and the formula is defined as:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/726/1*bBrJsuWUTFLHDdR58cwYMw.png" /></figure><p>The formula is fairly simple, but defining what a correct swing looks like gets a bit tricky. I’ll explain the math and the models behind coming to such conclusions later, but for now, think of a correct swing as the batter making a decision (swing or no swing), and the expected run value from doing so being greater than if they had done the opposite. In other words, if the batter had made the other decision, would that pitch have been worth more or less runs? The leaderboard for Correct Swing Percentage in 2024 looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GXuYuvoauBssBf4w5SaMnA.png" /><figcaption>Min. 500 pitches seen</figcaption></figure><p>There are many ways to validate this metric, of which I did (for example the stickiness, also known as year to year correlation, is .78), but for the sake of length I’m going to make one point: Javier Báez is last. And as a Red Sox fan, Ceddanne Rafaela being the 4th lowest makes me trust the metric just as much. Everyone on this list has two things in common: they don’t chase, and they take walks. You may be questioning DJ Stewart at #1, but he was 100th percentile in chase and walk rate last year, although not qualified.</p><p>To get a better sense of the distribution, the quantiles are as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/486/1*JfSdDCPO_TEFRvUyFDmp8g.png" /></figure><p>And it follows a relatively normal distribution:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NqVLG2FpX_VzPFElB97knA.png" /></figure><p>So now that we have a metric, how did we get here? To model the expected run values that the Correct Swing Percentage formula is derived from, we split the data into two subsets: swings and takes. For each one, a LightGBM regressor was fit to the data, with the following features for <em>both</em> models:</p><p>Balls</p><p>Strikes</p><p>Outs</p><p>Horizontal Plate Location</p><p>Vertical Plate Location</p><p>Base State (factor)</p><p>Pitch Category (As a factor, 0 = Fastballs, 1 = Breaking Balls, 2 = Offspeed)</p><p>Stuff Plus</p><p>As for feature importance, for the take model the location and the balls were pretty much all that factored in. For the swing model however, stuff plus and the base state became pivotal features, and the pitch category became a non-zero.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JjEGMW6qYfoSudDZUG5fMw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lasuKjvvoXUcDJMD7Gvtdg.png" /></figure><p>I took both these models, which are trained on only swings and takes respectively, and predicted both on every row in the dataset regardless of swing decision. Now for every pitch regardless of the decision, we have the expected run value if they swing, and the expected run value if they take.</p><p>Then, a correct swing is defined as:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*U2zyzEchkNhQbU_k43R5Jg.png" /></figure><p>As I explained earlier, the batter gets a correct swing if they swing and the expected run value from swinging is greater than if they had taken the pitch, or if the batter takes a pitch and the expected run value from not swinging is greater than if they had swung.</p><p>Then taking the sum of correct swings and dividing by pitches seen gets a Correct Swing Percentage.</p><p>The models both make one very strong assumption: a “good” pitch to swing at is the <strong><em>same</em></strong> across all hitters, meaning it is assumed they are league average. It is a false assumption, but the reason I can make it is because I’m using the metric to evaluate pitchers, not hitters. And over a large sample, we can assume that the average hitter a pitcher faces is well, league average. Originally my plan was to adjust for batter ability, but then it became abundantly clear that for each pitcher, measuring the skill of the hitter they are facing will converge to league average over the course of a season. For pitchers, lower Correct Swing Percentage is better.</p><p>By pitcher, the 2024 leaderboard looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2KceTUAsT6L7-ErRP10ycQ.png" /><figcaption>Min. 500 pitches</figcaption></figure><p>By pitch type, the distribution is wider (2024):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EtUngkxGJQKiswzJom-K5w.png" /><figcaption>Min. 200 pitches</figcaption></figure><p>On average, breaking balls fooled hitters more than any other pitch group, followed by offspeed and then fastballs. Slurves were the individual pitch type with the lowest correct swing percentage at 65%, with 4-seam fastballs tied with cutters for the highest with 71%. This follows the Stuff+ trend, where breaking balls (particularly sweepers and sliders) have the typically highest values, followed by the same order of offspeed then fastballs.</p><p>Deception Runs and Deception+ are very similar to Correct Swing Percentage, but I make one key change: instead of categorizing each pitch as a correct decision or not, I <em>quantify </em>how good the decision was, by taking the <em>difference</em> in expected run values for each pitch. This way, each individual pitch has a continuous, numeric value instead of a binary 1 or 0. Taking the average of this difference gets the amount of runs gained or lost due to batter swing decisions. This is called Deception Runs. This way, inducing a swing on a pitch 2 feet outside the zone is more valuable than getting one 2 inches away. It’s important to note that this metric is not just chase rate. While it is heavily correlated for obvious reasons, there’s a difference between labeling a pitch as just a “chase” or a “correct swing” and <strong><em>measuring</em></strong> how much that chase mattered, or what was the value of getting that swing decision. Because like I’ve already mentioned, not all chases have the same impact. Deception Runs also takes into account pitches in the zone too (which Correct Swing Percentage also does), which are equally as important for accumulating the stat. This is where the models come into play. What is considered a good pitch to swing at is not black and white on each edge of a strike zone. Using a simple formula like (In Zone Swing Swings + O-Zone Takes) / Total Pitches Seen is similar, but not good enough because it’s not always a good decision to swing at every pitch in the zone, and sometimes with two strikes it’s beneficial to chase a 50/50 ball on the black. The stuff plus also plays a considerable role. The range of positive swing xRV values shrinks considerably with better stuff.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JDHFWol9uwRgJtFliw5pMQ.png" /><figcaption>Stuff+ on traditional 20–80 scale</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rPKrhoBdnbRs67GSqagpSQ.png" /><figcaption>Stuff+ on traditional 20–80 scale</figcaption></figure><p>Consider this example with Chris Sale:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ud8tq6BW0EPA19F5KUXYEw.png" /></figure><p>On the left, he throws a bad slider down the middle to Daz Cameron in a 2–2 count. The expected run value if he had taken that pitch is -.398 runs. If he swings, which is obviously the correct decision given the situation, the expected run value is .098 runs. Because Cameron made the correct swing decision, or in other words Chris Sale failed to make him make the wrong one, the Deception Runs from this pitch is -.398 — .098 = -.5 Deception Runs. Had he gotten a take, we would flip it, and the Deception Runs would be equal to positive .5.</p><p>In the second scenario on the right, Chris Sale throws a nasty slider on the black and gets Tyler O’Neill to go down looking. Similar to before, the Deception Runs form this pitch is equal to the swing xRV — take xRV, because in this situation O’neill should have swung but didn’t. For this particular pitch, the xRV of a take was -.51, and only .01 for a swing because the location of the pitch was in such a good spot. This results in a positive Deception Runs of .52.</p><p>If the batter makes the correct decision, Deception Runs will always be negative, and vice versa.</p><p>Because batters are generally “good” at making decisions, averaging the right decision 69% of the time, Deception Runs will always be negative for qualified pitchers. In other words, every pitcher will pretty much always “lose” runs based on the quality of batter swing decisions that they induce, but some negatives are still better than others. In 2024, Mason Miller led all pitchers (min. 500 pitches) with only 3.35 runs lost per 100 pitches, with Kyle Nicolas in dead last at 471st losing 7.59 runs per 100 pitches.</p><p>The distribution of Deception Runs per 100 Pitches looks like:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N3asSBTegPOpSEU5nBm46g.png" /><figcaption>Min. 500 pitches</figcaption></figure><p>Here’s the full Deception Runs leaderboard from 2024:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B2pxUl5WFtu81lISrXxqgA.png" /><figcaption>Min. 500 pitches</figcaption></figure><p>The way to interpret these results is “Blake Treinen only lost 3.77 runs per 100 pitches in 2024 due to batter swing decisions”, and such for each player. Is a difference of only 4.36 runs from first to last a lot? The answer is it’s more than you’d think. Remember that this is a rate statistic, and while 4 runs separating first from last doesn’t seem like a lot, it’s per 100 pitches, and that 4 runs adds up over time. For what it’s worth, the difference between the best pitcher in actual run value / 100 in 2024 and the worst (Emmanuel Clase vs Tanner Gordon) by the same qualifications (500 pitches) is 6.5 runs.</p><p>These days plus stats are all the rave. All a plus stat means is that the average is 100 and follows a normal distribution, with higher being better, making the stat easier to compare players. With a known standard deviation (in this case I use 10), it is very easy to say “player X is Y% better than player Z in this stat”. Turning Deception Runs into a plus stat, the leaders for 2024 are the same:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vOifZM_xPiTEMetIYKqkoQ.png" /><figcaption>Min. 500 pitches</figcaption></figure><p>By pitch type (2024):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EbiihYrhwB-QXPD8yXX1aA.png" /><figcaption>Min. 200 pitches</figcaption></figure><p>Deception Runs are more useful than Deception+, because it provides an actual run value that has a real context. Deception+ is used to easily compare these Deception Run values against each other, because sometimes a stat isn’t that easy to interpret on its own. Both stats have good correlation with pitcher success, and I think that although it can take a larger sample to stabilize, the usefulness of Deception Runs is palpable. If you have any questions about anything in this article, as some of the machine learning and math can be overwhelming, feel free to reach out to me on X/Twitter: <a href="https://x.com/BenResnic_">https://x.com/BenResnic_</a>, and give me a follow! Thanks for reading!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d161ddd7603a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Connor Wong is Not a Good Enough Primary Catcher]]></title>
            <link>https://medium.com/@bsresnic/connor-wong-is-not-a-good-enough-primary-catcher-982cee160d46?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/982cee160d46</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Sun, 19 Jan 2025 21:09:36 GMT</pubDate>
            <atom:updated>2025-01-19T21:09:36.762Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*L0raWpzi75OUb6PG.jpg" /></figure><p>Kyle Teel is no longer the catcher of the future. Garrett Crochet is here, and I’m unbelievably ecstatic about the acquisition and what we gave up. That being said, it left the Red Sox with one catcher on the roster. So they went out and traded international bonus pool money for Carlos Narvaez, a catcher formerly in the Yankees system who spent the season in AAA. Then just a few days ago Boston traded for Blake Sabol from the Giants to complete their catching depth. These moves came after the Red Sox completely whiffed on the entire catching free agent market, which had some decent options like Carson Kelly and Travis d’Arnaud. The problem? The catching room is not nearly good enough, and it stems from the top. I’m not going to be angry that the backup catchers have glaring flaws, but I will be that the starter does.</p><p>Wong produced good numbers in 2024. You will gladly take a .758 OPS from your catcher in nearly 500 plate appearances every single time. The problem is that he got lucky. Like really lucky. Actually, he was THE luckiest hitter in 2024, by wOBAcon — xwOBAcon (weighted on base average on balls in play, x stands for expected). Put simply, on batted ball events, Connor Wong had the most success above what he was expected to out of any batter in baseball with at least 250 balls put into play.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/794/1*MSaSSyPTDYhDHdt0sRgZSA.png" /></figure><p>What does this mean? Wong ranked 64th out of 234 hitters I studied in wOBAcon, but 159th in xwOBAcon. You might be thinking “so what, who cares that he was expected to do it only matters what the actual results were”. And you would be correct. The issue comes from what I’m trying to do, which is to predict his 2025 season. If he puts up similar expected numbers this year, he WILL regress to the mean, or at least near it, and won’t be nearly as valuable. This is because expected stats are accurate. The average difference between a stat and it’s expected version over a large enough sample is zero, or approaches zero as more and more samples come in. Aside from just the batted ball data, he whiffs on 29% of swings, rarely walks, chases at a below league average rate, and all that corresponds to being in the 38th percentile in K rate. The blue bubbles on Connor Wong’s baseball savant page aren’t lying. Under the hood, he wasn’t very good last year, and the expected numbers are a much better indicator of how he will hit in 2025 than what’s traditionally found in a box score.</p><p>Then there’s another problem. Especially for a catcher, hitting is only half the battle of evaluating a player. Wong was the single worst defensive catcher last year too. That’s right, we had the luckiest hitter and the worst defensive catcher behind the plate for most games. Wong is average at throwing out runners, but is the second worst framer, and is 3rd percentile in blocking. Statcast has their own “framing runs” metric to evaluate framing, but I created my own stat just to see the extent of Wong’s horrid framing. I made a simple model that takes the location coordinate of the pitch, and the heights of the strike zone, to predict if the umpire would call any given pitch a ball or strike on a pitch that wasn’t swung at. Every strike gets assigned either a 1 for a strike, or 0 for a ball. If the pitch was a strike, and the model predicted that there is a 90% probability that it would be called a strike, then the catcher of record would get .1 <em>strikes added. </em>Coming from 1 — .9 (The code for a strike minus the probability of a strike). Catchers who excel at framing would rack up strikes added. Below is a table showing the top and bottom 5 this past year in strikes added (min 1000 pitches caught).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/776/1*mVq3NWxazUIClh_mbx8J0A.png" /></figure><p>Rounding to -74, the way to interpret this is over Wong’s season, 74 pitches that should have been called strikes were called balls compared to <em>average</em>. That doesn’t sound like a lot, but think of it like this. One pitch has .01 strikes added, the next has -.09, then .03, then .02, -.02, .08, -.13, .01, -.06, -.02. Every pitch fluctuates, some have small positive values and some small negative. But over time they add up. Just in this made up 10 pitch example there are -.17 strikes added. Again, this number is small, but they add up, and strikes are valuable, and Connor Wong was the second worst in MLB at gaining strikes. So why does this matter? The better a catcher is at framing, the better pitchers will perform when throwing to them. It’s that simple. Below are two simple scatter plots. On the Y axis of both of them is pitching run value when catching. Basically, it is a way to measure the pitcher’s performance for each catcher he has. In general, the better defense by the catcher, the better the pitcher performs, which is what actually matters here. But also note that the pitcher has to still be productive no matter the catcher to be successful, as the catcher can help, but he’s not the one throwing the pitches.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fJcEO9fkGJwbCWdITfV2Ww.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dKWC-JiN1iYji2NandYTbw.png" /></figure><p>Of the top 20 battery duos on the Red Sox last year by run value, Wong was the catcher in just 5 of them, despite catching 61% of all pitches. Reese Mcguire isn’t labeled on either graph above to save space, but if he were he’d be above average with 22 strikes added and a pitcher run value when catching of 23 runs. The difference between McGuire and Wong is over 50 runs. Now you don’t need me to tell you that that’s a lot, and considering that that number is with the same pitching staff, it’s hard not to wonder what the results would have been if Wong were even just a league average defender last year. Even when they put him in the field after trading for Danny Jansen, he was terrible, accumulating -4 OAA in very limited innings.</p><p>The Red Sox hired a new catching coach, Parker Quinn, from the Yankees. Who knows if that will make a difference, but it’s worth noting that the Yankees did have the 4th and 6th best defensive catchers last year by fielding run value, minimum 500 innings. But for right now, going into 2025 with the worst defensive catcher in MLB, who also way over performed on his expected stats, isn’t good enough.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=982cee160d46" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Create a Pitching Summary Using R]]></title>
            <link>https://medium.com/@bsresnic/how-to-create-a-pitching-summary-using-r-046fb4f4ec25?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/046fb4f4ec25</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Thu, 19 Dec 2024 03:14:15 GMT</pubDate>
            <atom:updated>2024-12-19T03:14:15.036Z</atom:updated>
            <content:encoded><![CDATA[<p>MLB provides so much data for free to the public, but most people don’t have access to it or don’t know what to do with it if they did. This article will walk you through my process to turn these endless rows of play by play data into a summary that you can take something away from. The basis of my pitching cards is to take raw data and make it simple and understandable for people to learn everything about a pitcher’s arsenal. I do this by making graphics that are as visually appealing as possible, easy to read and with colors that pop. The goal is to provide information about only the metrics on each pitch, not the results, so it could be used to summarize just a bullpen session for example.</p><p>I’ll start with the packages. I have the latest version of R downloaded, and you will need to install and read in these 14 packages.</p><pre>library(tidyverse)<br>library(gt)<br>library(gtExtras)<br>library(xgboost)<br>library(lubridate)<br>library(webshot2)<br>library(magick)<br>library(grid)<br>library(gridExtra)<br>library(ggforce)<br>library(mlbplotR)<br>library(stringr)<br>library(geomtextpath)<br>library(rvest)</pre><p>Then I load in the data. All of my data is downloaded using a scraper built by Scott Powers, an assistant professor at Rice. It only downloads regular season games, so we can set the start and end dates to span the whole year. Also, for some reason, you can only download one season at a time. I also read in my stuff plus model that I have already trained on this exact data. This isn’t necessary to recreate, but it is something fun to try out and evaluate pitcher’s raw stuff. I also calculate vertical approach angle using physics from <a href="https://blogs.fangraphs.com/a-visualized-primer-on-vertical-approach-angle-vaa/"><strong>this</strong></a> Fangraphs article written by Alex Chamberlain.</p><pre>devtools::install_github(repo = &quot;saberpowers/sabRmetrics&quot;)<br><br>d21 &lt;- sabRmetrics::download_baseballsavant(<br>  start_date = &quot;2021-01-01&quot;,<br>  end_date = &quot;2021-12-31&quot;<br>)<br><br>d22 &lt;- sabRmetrics::download_baseballsavant(<br>  start_date = &quot;2022-01-01&quot;,<br>  end_date = &quot;2022-12-31&quot;<br>)<br><br>d23 &lt;- sabRmetrics::download_baseballsavant(<br>  start_date = &quot;2023-01-01&quot;,<br>  end_date = &quot;2023-12-31&quot;<br>)<br><br>d24 &lt;- sabRmetrics::download_baseballsavant(<br> start_date = &quot;2024-01-01&quot;,<br>  end_date = &quot;2024-12-31&quot;<br>)<br><br>load(&quot;stuffplus.RData&quot;)<br><br><br>calculate_vaa &lt;- function(data) {<br>  yf &lt;- 17 / 12      <br>  <br>  data &lt;- data %&gt;%<br>    mutate(<br>      vy_f = -sqrt(vy0 ** 2 - (ay * 2 * (50 - yf))), <br>      time = (vy_f - vy0) / ay,<br>      vz_f = vz0 + az * time,  <br>      vaa = -atan(vz_f / vy_f) * (180 / pi)  <br>    )<br>  <br>  return(data)<br>}<br><br>data &lt;- calculate_vaa(data)</pre><p>Now that we have all the play by play data, it’s time to clean the data in order to prepare it for each function later that creates a graph. Unfortunately, this data does not have a column for the pitcher’s name. I don’t know why, but we will have to create it. I use a separate dataset to do this with biographical information. It has every player in MLB’s name, id, height, weight, and age. We will use the rest of this information later.</p><pre>data &lt;- rbind(d21, d22, d23, d24)<br><br>ids &lt;- read_csv(&quot;player_ids.csv&quot;)<br><br>pitcher_ids &lt;- ids %&gt;% <br>    select(player_id, name) %&gt;% <br>    rename(pitcher_id = player_id)<br><br>data &lt;- merge(data, pitcher_ids, by = &quot;pitcher_id&quot;)<br><br>data$pitcher_team &lt;- ifelse(data$inning_topbot == &quot;Top&quot;, data$home_team, data$away_team)</pre><p>The next step is transforming the data so that I can predict the stuff plus model on it. In order to do this I need to alter the data and create a few new columns. Again, if you’re following along and recreating this, the code that follows isn’t necessary besides converting certain columns from feet to inches (multiplying by 12).</p><pre><br>data &lt;- data %&gt;% <br>   mutate(stand = ifelse(bat_side == &quot;R&quot;, 1, 0),<br>         plate_x = plate_x * -12, p_throws = ifelse(pitch_hand == &quot;R&quot;, 1, 0), pfx_x = pfx_x * -12) %&gt;% <br>  mutate(plate_x = round(plate_x, 1),<br>           plate_z = round(plate_z, 1),<br>         on_3b = ifelse(!is.na(pre_runner_3b_id), 1, 0),<br>         on_2b = ifelse(!is.na(pre_runner_2b_id), 1, 0),<br>         on_1b = ifelse(!is.na(pre_runner_1b_id), 1, 0))<br><br>max &lt;- data %&gt;% <br>  filter(pitch_type %in% c(&quot;FF&quot;, &quot;FA&quot;, &quot;SI&quot;)) %&gt;% <br>  select(pitcher_id, pitch_type, release_speed, az, year) %&gt;% <br>  na.omit() %&gt;% <br>  group_by(pitcher_id, year) %&gt;% <br>  summarise(mean_release_speed = mean(release_speed), mean_az = mean(az))<br><br>data &lt;- left_join(data, max, by = c(&quot;pitcher_id&quot;, &quot;year&quot;))<br><br>data &lt;- data %&gt;% <br>  mutate(delta_release_speed = release_speed - mean_release_speed,<br>         delta_az = az - mean_az) %&gt;% <br>  select(-mean_az, -mean_release_speed)<br><br>data &lt;- data %&gt;% <br>  group_by(balls, strikes, description, on_3b, on_2b, on_1b) %&gt;%<br>  mutate(mean_delta_run_exp = mean(delta_pitcher_run_exp, na.rm = TRUE)) %&gt;%<br>  ungroup() %&gt;%<br>  mutate(delta_run_exp = ifelse(description == &quot;hit_into_play&quot;, delta_pitcher_run_exp, mean_delta_run_exp)) %&gt;%<br>  select(-mean_delta_run_exp)<br><br>data &lt;- data %&gt;% <br>  mutate(ax = ax * 12,<br>         az = az * 12, <br>         plate_z = plate_z * 12, <br>         pfx_z = pfx_z * 12, <br>         ay = ay * 12)<br><br>data &lt;- data %&gt;% <br>  mutate(ax = abs(ax))<br><br>data &lt;- data %&gt;% <br> drop_na(ax, az, release_speed, pitcher_id, name, pitch_name, p_throws, arm_angle, release_pos_z, release_spin_rate, <br>          delta_release_speed, delta_az, delta_run_exp, extension, year)<br><br>matrix &lt;- xgb.DMatrix(as.matrix(select(data, release_speed, ax, az, release_pos_z,<br>                                       extension, release_spin_rate,<br>                                       delta_release_speed, delta_az, arm_angle)),<br>                      label = data$delta_run_exp)<br><br>data$stuff_plus &lt;- predict(bst, matrix)<br><br>data$stuff_plus &lt;- round(scale(data$stuff_plus) * 20 + 100)</pre><p>Now that we’re all set on that, I will read in one more csv file that I have previously downloaded from baseball savant directly. It gives the spin direction (measured and observed) for every pitch as well as the spin efficiency.</p><pre>spin &lt;- read_csv(&quot;spin-direction-pitches.csv&quot;) %&gt;% <br>  rename(name = `last_name, first_name`)<br><br>spin &lt;- spin %&gt;%<br>  select(name, api_pitch_name, active_spin_formatted,<br>         hawkeye_measured_clock_label, movement_inferred_clock_label,<br>         hawkeye_measured, movement_inferred) %&gt;%<br>  separate(name, into = c(&quot;LastName_Pitcher&quot;, &quot;FirstName_Pitcher&quot;), sep = &quot;, &quot;) %&gt;%<br>  unite(player_name, FirstName_Pitcher, LastName_Pitcher, sep = &quot; &quot;) %&gt;%<br>  rename(`Pitch Type` = api_pitch_name,<br>         active_spin = active_spin_formatted,<br>         measured_spin = hawkeye_measured_clock_label,<br>         observed_spin = movement_inferred_clock_label)</pre><p>Then, I will take the data and make a separate but similar data frame used for the pitch trajectories. It takes a function on the original data to add new columns for the coefficients used in the parametric equations.</p><pre>trajectory_data &lt;- data |&gt;<br>  sabRmetrics::get_quadratic_coef(source = &quot;baseballsavant&quot;) |&gt;<br>  sabRmetrics::get_trackman_metrics()<br><br><br>trajectory_data &lt;- trajectory_data %&gt;% <br>  mutate(t_0 = 0, <br>         bx = bx * 12, <br>         bz = bz * 12, <br>         release_x = release_x * 12, <br>         release_z = release_z * 12, <br>         by = by * 12, <br>         release_y = release_y * 12)</pre><p>Lastly I establish a color scheme for the graphics. This is arguably the most important part. The colors flow from one pitch type to another, pop out, and provide consistency throughout the whole thing. It also takes away the need to have a legend on every single plot, and makes the graphic much cleaner. For simplicity, I deem a knuckle curveball and a curveball the same pitch.</p><pre><br>data &lt;- data %&gt;% <br>  mutate(pitch_type = case_when(<br>    pitch_type == &quot;KC&quot; ~ &quot;CU&quot;,<br>    T ~ pitch_type<br>  )) %&gt;% <br>   mutate(pitch_name = case_when(<br>    pitch_name == &quot;Knuckle Curve&quot; ~ &quot;Curveball&quot;,<br>    T ~ pitch_name<br>  ))<br><br>pitch_colors &lt;- c(<br>  &quot;4-Seam Fastball&quot; = &quot;#0259d4&quot;,<br>  &quot;Changeup&quot; = &quot;#0a6e28&quot;,<br>  &quot;Curveball&quot; = &quot;#f73620&quot;,<br>  &quot;Cutter&quot; = &quot;#5604c9&quot;,<br>  &quot;Eephus&quot; = &quot;#0cf7e4&quot;,<br>  &quot;Forkball&quot; = &quot;#027525&quot;,<br>  &quot;Knuckleball&quot; = &quot;#b35f0b&quot;,<br>  &quot;Other&quot; = &quot;black&quot;,<br>  &quot;Screwball&quot; = &quot;#8a810b&quot;,<br>  &quot;Sinker&quot; = &quot;#02e1f5&quot;,<br>  &quot;Slider&quot; = &quot;#be3af2&quot;,<br>  &quot;Slurve&quot; = &quot;#f05411&quot;,<br>  &quot;Split-Finger&quot; = &quot;#11b880&quot;,<br>  &quot;Sweeper&quot; = &quot;#fc03b1&quot;<br>)</pre><p>Finally! The boring part is done, and now it’s time to create the fun plots that will make up the graphic as a whole. For all of the following plots there is a function that creates it. Every function (besides this first one) takes in three arguments: player, game_year, and data. The first plot we will make is the header, which displays the pitcher’s official MLB headshot, name and information, and team logo. The nice thing about the headshot and logo is that both are a PNG, meaning there is no background. The player info function only takes in the player as an argument. For the example in this article, I will create the summary for Zack Wheeler in 2024.</p><pre>player_info &lt;- function(player) {<br>  <br>#calculate age<br>  ids &lt;- ids %&gt;% <br>    mutate(player_age = floor(interval(birthDate, today()) / years(1))) %&gt;% <br>    filter(name == player) <br><br> #get handedness<br>  hand &lt;- data %&gt;% <br>    filter(name == player) %&gt;% <br>    mutate(pitcher_hand = ifelse(p_throws == 0, &quot;LHP&quot;, &quot;RHP&quot;)) %&gt;% <br>    slice_head(n = 1) %&gt;% <br>    pull(pitcher_hand)<br><br>#pull player&#39;s name, age, height, weight, and id<br>  name_parts &lt;- strsplit(player, &quot; &quot;)[[1]]<br>  first_name &lt;- name_parts[1]<br>  last_name &lt;- name_parts[2]<br>  age &lt;- ids %&gt;%<br>    pull(player_age)<br>  height &lt;- ids %&gt;%<br>    pull(height)<br>  weight &lt;- ids %&gt;%<br>    pull(weight)<br>  <br>  id &lt;- ids %&gt;%<br>    pull(player_id)<br>  <br>#this scrapes the MLB api to download a headshot directly<br>  url &lt;- paste0(<br>    &quot;https://img.mlbstatic.com/mlb-photos/image/&quot;,<br>    &quot;upload/d_people:generic:headshot:67:current.png&quot;,<br>    &quot;/w_640,q_auto:best/v1/people/&quot;, id, &quot;/headshot/silo/current.png&quot;<br>  )<br>  <br>  img &lt;- image_read(url)<br>  raster_img &lt;- rasterGrob(as.raster(img)) <br>  <br>#plot the headshot<br>  p1 &lt;- ggplot() +<br>    annotation_custom(raster_img, xmin = -Inf, xmax = Inf, ymin = -Inf, ymax = Inf) +<br>    theme_void() +<br>    theme(plot.margin = margin(0, 0, 0, 0))<br>  <br>#plot information<br>  p2 &lt;- ggplot() +<br>  annotate(<br>    &quot;text&quot;, x = 0, y = 0.00, <br>    label = paste0(first_name, &quot; &quot;, last_name),<br>    color = &quot;#0c2340&quot;, size = 24, hjust = 0.5, vjust = .25, fontface = &quot;bold&quot;<br>  ) +<br>  annotate(<br>    &quot;text&quot;, x = 0, y = 0, <br>    label = paste0(hand, &quot; &quot;, height, &quot; &quot;, weight, &quot; lbs Age &quot;, age),<br>    color = &quot;#0c2340&quot;, size = 15, hjust = 0.5, vjust = 2.4, fontface = &quot;italic&quot;, lineheight = 0.8<br>  ) +<br>  theme_void() +<br>  theme(plot.margin = margin(0, 0, 0, 0))<br>  <br>#plot the team logo<br>  abv &lt;- data %&gt;%<br>    filter(name == player) %&gt;%<br>    arrange(desc(game_date)) %&gt;%<br>    slice_head(n = 1) %&gt;%<br>    pull(pitcher_team)<br>  <br>  p3 &lt;- ggplot() +<br>    mlbplotR::geom_mlb_logos(aes(x = 0, y = 0, team_abbr = abv), width = 0.5, vjust = .5) +<br>    theme_void() +<br>    theme(plot.margin = margin(0, 0, 0, 0))<br>  <br>#combine all three aspects onto one graph<br>  grid.arrange(p1, p2, p3, ncol = 3, widths = c(0.25, 0.5, 0.25))<br>}</pre><p>And here it is, the header for Zack Wheeler’s pitching summary!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7d3WlZVgENTfgLAZNIT-ug.png" /></figure><p>Next is the table with all the pitch metrics. Using the package gt, we are able to create a clean table with all the data you could possibly want about each of our player’s pitches.</p><pre>pitch_summary &lt;- function(pitcher, game_year, data){   <br>  selected_pitcher &lt;- pitcher<br>  <br>  data &lt;- data %&gt;% <br>    filter(year == game_year)<br><br>#create the domains for the color scales<br>  ext_domain &lt;- data %&gt;%<br>    group_by(name) %&gt;%<br>    summarise(extension = mean(extension)) %&gt;%<br>    arrange(desc(extension)) %&gt;%<br>    summarise(min = floor(min(extension)), max = ceiling(max(extension)))<br><br>  domains &lt;- data %&gt;%<br>    group_by(name, pitch_name) %&gt;%<br>    summarise(n = n(), mean = mean(release_speed), spin = mean(release_spin_rate)) %&gt;%<br>    filter(n &gt; 50) %&gt;%<br>    ungroup() %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    summarise(min_velo = floor(min(mean)), max_velo = ceiling(max(mean)),<br>              min_spin = floor(min(spin)), max_spin = ceiling(max(spin))) %&gt;%<br>    mutate(min_velo = as.numeric(min_velo),<br>           max_velo = as.numeric(max_velo),<br>           min_spin = as.numeric(min_spin),<br>           max_spin = as.numeric(max_spin))<br><br>#filter for pitches thrown more than 1% of the time<br>  types &lt;- data %&gt;%<br>    filter(name == selected_pitcher) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    summarise(n = 100 * n() / nrow(data %&gt;% filter(name == selected_pitcher))) %&gt;%<br>    filter(n &gt;= 1)<br><br>  types &lt;- unique(types$pitch_name)<br><br>  spin &lt;- spin %&gt;%<br>    filter(player_name == selected_pitcher) %&gt;%<br>    select(-player_name)<br><br>#group by pitch and summarize all metrics<br>  pitcher_summary &lt;- data %&gt;%<br>    filter(name == selected_pitcher) %&gt;%<br>    filter(pitch_name %in% types) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    summarise(<br>      Usage = round((n() / nrow(data %&gt;% filter(name == selected_pitcher))) * 100, 1),<br>      Velo = round(mean(release_speed), 1),<br>      spin = round(mean(release_spin_rate)),<br>      hb = round(mean(pfx_x), 1),<br>      ivb = round(mean(pfx_z), 1),<br>      vaa = round(mean(vaa), 1),<br>      ext = round(mean(extension), 1),<br>      hrel = round(mean(-release_pos_x), 1),<br>      vrel = round(mean(release_pos_z), 1),<br>      `Stuff+` = round(mean(stuff_plus))<br>    ) %&gt;%<br>    rename(`Pitch Type` = pitch_name) %&gt;%<br>    arrange(desc(Usage))<br><br>  pitcher_summary &lt;- pitcher_summary %&gt;%<br>    left_join(spin, by = c(&quot;Pitch Type&quot;)) %&gt;%<br>    rename(eff = active_spin,<br>           measured = measured_spin,<br>           observed = observed_spin) %&gt;%<br>    mutate(<br>      measured = substr(measured, 1, 5),<br>      observed = substr(observed, 1, 5)<br>    ) %&gt;%<br>    filter(Usage &gt;= 1)<br><br>  pitcher_summary &lt;- pitcher_summary %&gt;%<br>    select(`Pitch Type`, Usage, Velo, spin, eff, measured, observed, vaa,<br>           hb, ivb, ext, hrel, vrel, `Stuff+`)<br><br>#create table, using special gt theme<br>  pitcher_summary &lt;- pitcher_summary %&gt;%<br>    gt() %&gt;%<br>    gt_theme_538() %&gt;%<br>    cols_align(align = &quot;center&quot;, columns = everything())<br><br>#assign color to columns that I specify<br>  pitcher_summary &lt;- pitcher_summary %&gt;%<br>    data_color(<br>      columns = `Pitch Type`,<br>      colors = scales::col_factor(<br>        palette = pitch_colors,<br>        domain = names(pitch_colors),<br>        na.color = &quot;white&quot;<br>      )<br>    )<br><br>  for (pitch in types) {<br>    min_velo &lt;- domains %&gt;% filter(pitch_name == pitch) %&gt;% pull(min_velo)<br>    max_velo &lt;- domains %&gt;% filter(pitch_name == pitch) %&gt;% pull(max_velo)<br><br>    pitcher_summary &lt;- pitcher_summary %&gt;%<br>      data_color(<br>        columns = Velo,<br>        rows = `Pitch Type` == pitch,<br>        palette = c(&quot;#056af7&quot;, &quot;white&quot;, &quot;#ff5100&quot;),<br>        domain = c(min_velo, max_velo)<br>      )<br><br>    min_spin &lt;- domains %&gt;% filter(pitch_name == pitch) %&gt;% pull(min_spin)<br>    max_spin &lt;- domains %&gt;% filter(pitch_name == pitch) %&gt;% pull(max_spin)<br><br>    pitcher_summary &lt;- pitcher_summary %&gt;%<br>      data_color(<br>        columns = spin,<br>        rows = `Pitch Type` == pitch,<br>        palette = c(&quot;#056af7&quot;, &quot;white&quot;, &quot;#ff5100&quot;),<br>        domain = c(min_spin, max_spin)<br>      )<br>  }<br><br>  pitcher_summary &lt;- pitcher_summary %&gt;%<br>    data_color(<br>      columns = `Stuff+`,<br>      colors = scales::col_numeric(<br>        palette = c(&quot;#056af7&quot;, &quot;white&quot;, &quot;#ff5100&quot;),<br>        domain = c(50, 150),<br>        na.color = &quot;white&quot;<br>      )<br>    ) %&gt;%<br>    data_color(<br>      columns = c(ext),<br>      colors = scales::col_numeric(<br>        palette = c(&quot;#056af7&quot;, &quot;white&quot;, &quot;#ff5100&quot;),<br>        domain = c(ext_domain$min, ext_domain$max),<br>        na.color = &quot;white&quot;<br>      )<br>    ) %&gt;%<br>    data_color(<br>      columns = c(eff),<br>      colors = scales::col_numeric(<br>        palette = c(&quot;#056af7&quot;, &quot;white&quot;, &quot;#ff5100&quot;),<br>        domain = c(0, 100),<br>        na.color = &quot;white&quot;<br>      )<br>    )<br><br>  pitcher_summary<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YWqt7a-48bgfSQZHvnuKLw.png" /></figure><p>Next for a better understanding of the pitcher’s velocity distribution, we can plot a kde for each pitch.</p><pre>velo &lt;- function(player, game_year, data){    <br>  <br>  data &lt;- data %&gt;% <br>    filter(year == game_year)<br>  <br>#filter for pitches thrown more than 1% of the time<br>  types &lt;- data %&gt;%<br>    filter(name == player) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    summarise(n = 100 * n() / nrow(data %&gt;% filter(name == player))) %&gt;%<br>    filter(n &gt;= 1) %&gt;%<br>    arrange(desc(n))<br>  <br>  types &lt;- unique(types$pitch_name)<br>  <br>  data_filtered &lt;- data %&gt;%<br>    filter(name == player &amp; pitch_name %in% types) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    mutate(total_pitches = n()) %&gt;%<br>    ungroup() %&gt;%<br>    rename(`Pitch Type` = pitch_name)<br>  <br>#create x and y limits<br>  min_velo &lt;- data_filtered %&gt;%<br>    filter(!is.na(release_speed)) %&gt;%<br>    summarise(min = min(release_speed)) %&gt;%<br>    pull(min)<br>  <br>  min_velo &lt;- min_velo - 3<br>  <br>  max_velo &lt;- data_filtered %&gt;%<br>    filter(!is.na(release_speed)) %&gt;%<br>    summarise(max = max(release_speed)) %&gt;%<br>    pull(max)<br>  <br>  max_velo &lt;- max_velo + 3<br>  <br>#find average velo for each pitch<br>  avg_speed &lt;- data_filtered %&gt;%<br>    group_by(`Pitch Type`) %&gt;%<br>    summarize(<br>      avg_speed = mean(release_speed),<br>      usage_pct = n() / nrow(data_filtered) * 100<br>    )<br>  <br>#create the plot<br>  data_filtered %&gt;%<br>    arrange(desc(total_pitches)) %&gt;%<br>    ggplot(aes(x = release_speed, fill = `Pitch Type`)) + <br>    geom_density(alpha = 0.6) + <br>    geom_point(data = avg_speed, aes(x = avg_speed, y = 0.01, color = `Pitch Type`), size = 3, shape = 21, stroke = 0.3) + <br>    geom_segment(data = avg_speed, aes(x = avg_speed, xend = avg_speed, y = 0, yend = Inf, color = `Pitch Type`), linetype = &quot;dashed&quot;) + <br>    facet_grid(`Pitch Type` ~ ., scales = &quot;free_y&quot;, switch = &quot;y&quot;) + <br>    theme_minimal() + <br>    xlim(min_velo, max_velo) + <br>    xlab(&quot;Velocity (MPH)&quot;) + <br>    theme(<br>      legend.position = &quot;none&quot;,<br>      strip.text.y = element_blank(),<br>      plot.title = element_blank(),<br>      axis.text.y = element_blank(),<br>      axis.ticks.y = element_blank(),<br>      strip.background = element_blank(),<br>      panel.grid.major.y = element_blank(),<br>      panel.grid.minor.y = element_blank(),<br>      legend.title = element_blank(),<br>      axis.text.x = element_text(size = 25),<br>      axis.title.y = element_blank(),<br>      axis.title.x = element_text(size = 25),<br>      panel.background = element_rect(fill = &quot;#ffffff&quot;)<br>    ) + <br>    scale_fill_manual(values = pitch_colors) + <br>    scale_color_manual(values = pitch_colors)<br>}</pre><p>Boom!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/606/1*BAG2EX8ITFBdOVGYWPsToA.png" /></figure><p>Maybe the most important plot to visualize an arsenal is the movement plot. On the x axis the the horizontal movement, and on the y is the induced vertical movement. Read more about how to contextualize these numbers <a href="https://www.mlb.com/glossary/statcast/pitch-movement"><strong>here</strong></a><strong>.</strong></p><pre>movement &lt;- function(player, game_year, data) {<br>  <br>  data &lt;- data %&gt;% <br>    filter(year == game_year)<br>  <br>  data &lt;- data %&gt;%<br>    filter(name == player) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    mutate(<br>      pct = n() / nrow(data %&gt;% filter(name == player))<br>    ) %&gt;% <br>    filter(pct &gt;= .01) %&gt;% <br>    ungroup()<br>  <br>  data %&gt;% <br>  ggplot(aes(x = pfx_x, y = pfx_z)) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = 25), size = .25, color = &quot;black&quot;, fill = NA, linetype = &quot;dashed&quot;, alpha = .5) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = 20), size = .25, color = &quot;black&quot;, fill = NA) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = 15), size = .25, color = &quot;black&quot;, fill = NA, linetype = &quot;dashed&quot;, alpha = .5) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = 10), size = .25, color = &quot;black&quot;, fill = NA) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = 5), size = .25, color = &quot;black&quot;, fill = NA, linetype = &quot;dashed&quot;, alpha = .5) +<br>    geom_segment(aes(x = 0, y = -25, xend = 0, yend = 25), color = &quot;black&quot;, linetype = &quot;dashed&quot;) +<br>    geom_segment(aes(x = -25, y = 0, xend = 25, yend = 0), color = &quot;black&quot;, linetype = &quot;dashed&quot;) +<br>    geom_point(aes(fill = pitch_name), alpha = .8, shape = 21, color = &quot;black&quot;, stroke = .5, size = 5) +<br>    annotate(&quot;text&quot;, x = 20, y = 0, label = &quot;20&quot;, vjust = -.5, hjust = 1.2) +<br>    annotate(&quot;text&quot;, x = 10, y = 0, label = &quot;10&quot;, vjust = -.5, hjust = 1.2) +<br>    annotate(&quot;text&quot;, x = -20, y = 0, label = &quot;20&quot;, vjust = -.5, hjust = -.2) +<br>    annotate(&quot;text&quot;, x = -10, y = 0, label = &quot;10&quot;, vjust = -.5, hjust = -.2) +<br>    annotate(&quot;text&quot;, x = 0, y = 20, label = &quot;20&quot;, hjust = -.2, vjust = -.5) +<br>    annotate(&quot;text&quot;, x = 0, y = -20, label = &quot;20&quot;, hjust = -.2, vjust = -.5) +<br>    annotate(&quot;text&quot;, x = 0, y = 10, label = &quot;10&quot;, hjust = -.2, vjust = -.5) +<br>    annotate(&quot;text&quot;, x = 0, y = -10, label = &quot;10&quot;, hjust = -.2, vjust = -.5) +<br>    annotate(&quot;text&quot;, x = -18.5, y = 25,  <br>            label = &quot;← Towards 1B&quot;, color = &quot;black&quot;, size = 5, hjust = 0.5) +<br>   annotate(&quot;text&quot;, x = 18.5, y = 25, <br>            label = &quot;Towards 3B →&quot;, color = &quot;black&quot;, size = 5, hjust = 0.5) +<br>    scale_x_continuous(&quot;Horizontal Movement (in)&quot;, limits = c(-25, 25), breaks = seq(-20, 20, by = 10)) +<br>    scale_y_continuous(&quot;Induced Vertical Movement (in)&quot;, limits = c(-25, 25), breaks = seq(-20, 20, by = 10)) +<br>    scale_fill_manual(values = pitch_colors) +<br>    scale_color_manual(values = pitch_colors) +<br>    coord_equal() +<br>    theme_minimal() +<br>    theme(<br>      plot.title = element_blank(),<br>      legend.position = &quot;none&quot;,<br>      panel.background = element_rect(fill = &quot;#ffffff&quot;),<br>      legend.title = element_blank(),<br>      axis.text.x = element_text(size = 20),<br>                  axis.text.y = element_text(size = 20),<br>                  axis.title.x = element_text(size = 20),<br>                  axis.title.y = element_text(size = 20)<br>    )<br>}</pre><p>Below is the pitch movement plot for Zack Wheeler. I add the circles for added context, as sometimes it is hard to notice how much total movement a pitch has if it isn’t near the x or y axis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/538/1*IMGw07RCEC7KwysG0r7Rdw.png" /></figure><p>Plotting the pitch trajectories is tricky. This obviously uses the trajectory_data dataset that we created earlier. For each pitch, we plot it’s trajectory, where the initial trajectory of each one is right down the middle. The plotting is done with a paramtric equation for each dimension. Because velocities and accelerations differ slightly depending on the location of the pitch, instead of taking the average velocity and acceleration of each pitch, we plot a specific pitch that is as close as possible to our target location.</p><pre>tunnels &lt;- function(player, game_year, data){<br>  <br>#create a target of right down the middle<br>  target_x &lt;- 0<br>  target_z &lt;- 30<br>  <br>  data &lt;- data %&gt;%<br>   filter(year == game_year)<br>    <br>   data &lt;- data %&gt;%<br>   filter(name == player) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    mutate(<br>      pct = n() / nrow(data %&gt;% filter(name == player))<br>    ) %&gt;% <br>    filter(pct &gt;= .01) %&gt;% <br>    ungroup()<br>  <br>#filter for rows where pitch release is close to the mean<br>  data &lt;- data %&gt;%<br>    filter(name == player) %&gt;% <br>    group_by(pitch_name) %&gt;%<br>    mutate(<br>      release_x_mean = mean(release_pos_x, na.rm = TRUE),<br>      release_z_mean = mean(release_pos_z, na.rm = TRUE)<br>    ) %&gt;%<br>    ungroup() %&gt;%<br>    filter(<br>      abs(release_pos_x - release_x_mean) &lt;= 0.2 &amp; <br>      abs(release_pos_z - release_z_mean) &lt;= 0.2<br>    )<br>  <br>#find distance from initial coordinates to our target<br>  data &lt;- data %&gt;%<br>    filter(pitch_name %in% names(pitch_colors)) %&gt;% <br>    mutate(<br>      initial_x = plate_x - pfx_x,<br>      initial_z = plate_z - pfx_z,<br>      distance_to_0_30 = sqrt((initial_x - 0)^2 + (initial_z - 30)^2)  <br>    )<br> <br>  #find the pitch for each type closest to right down the middle<br>  data &lt;- data %&gt;%<br>    filter(name == player) %&gt;% <br>    mutate(distance = sqrt((initial_x - target_x)^2 + (initial_z - target_z)^2)) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    slice_min(distance, n = 1) %&gt;% <br>    ungroup() <br>  <br>#plot the location of the pitch in the air at 50 points during it&#39;s journey<br>#locations that end in 0 are release and 50 is the final location<br>  for (i in 0:50) {<br>    t &lt;- (i / 50) * data$plate_time<br>    <br>    data &lt;- data %&gt;%<br>      mutate(<br>        !!paste0(&quot;x_&quot;, i) := release_x + bx * t + 0.5 * ax * t^2,<br>        !!paste0(&quot;z_&quot;, i) := release_z + bz * t + 0.5 * az * t^2, <br>        !!paste0(&quot;y_&quot;, i) := release_y + by * t + 0.5 * ay * t^2<br>      )<br>  }<br> <br>  <br>  x_columns &lt;- paste0(&quot;x_&quot;, 0:50)<br>  z_columns &lt;- paste0(&quot;z_&quot;, 0:50)<br>  <br>#define plot limits<br>  x_max &lt;- round(max(abs(data[x_columns]), na.rm = TRUE) + 3)<br>  z_max &lt;- round(max(abs(data[z_columns]), na.rm = TRUE) + 3)<br><br>#plot the plate and strike zone<br>  p &lt;- ggplot(data) + <br>    geom_rect(<br>      xmin = -8.5, xmax = 8.5, ymin = 18, ymax = 42,<br>      fill = NA, color = &quot;black&quot;<br>    ) +<br>    geom_segment(aes(x = -7.5, y = 6, xend = 7.5, yend = 6), color = &quot;black&quot;) +<br>    geom_segment(aes(x = -7.5, y = 6, xend = -10, yend = 4), color = &quot;black&quot;) +<br>              geom_segment(aes(x = 7.5, y = 6, xend = 10, yend = 4), color = &quot;black&quot;) +<br>              geom_segment(aes(x = 0, y = 2, xend = 10, yend = 4), color = &quot;black&quot;) +<br>              geom_segment(aes(x = 0, y = 2, xend = -10, yend = 4), color = &quot;black&quot;) +<br>    scale_x_continuous(&quot;Horizontal Location (in)&quot;, limits = c(-x_max, x_max), breaks = seq(-x_max, x_max, by = 12)) +<br>    scale_y_continuous(&quot;Vertical Location (in)&quot;, limits = c(0, z_max), breaks = seq(0, z_max, by = 12)) +<br>    theme_void() + <br>    coord_equal() +<br>    ggtitle(paste0(&quot;Pitch Trajectories&quot;)) +<br>    theme(legend.position = &quot;none&quot;, <br>          plot.title = element_text(hjust = .5, size = 20))<br>    <br>  #add all of the points <br>  for (t in 0:50) {<br>    p &lt;- p + <br>      geom_point(aes(x = !!sym(paste0(&quot;x_&quot;, t)), y = !!sym(paste0(&quot;z_&quot;, t)), fill = pitch_name), shape = 21, <br>                 color = &quot;black&quot;, stroke = .15, size = 3) + <br>      scale_fill_manual(values = pitch_colors) <br>  }<br>  <br>  print(p)<br>  <br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/462/1*YHth9T8PYBnnzedzUmiYtQ.png" /></figure><p>Next is the tilts. These can be kind of complicated, so I’ll explain how it works. Spin axis, or tilt, explains how the baseball spins in the air, using hours and minutes on a clock. The tilt measures where the ball is spinning <em>from, </em>looking at the ball from the front. So 12:00 represents pure backspin, and 6:00 is pure topspin. Measured spin is the actual, real way the ball is spinning. Observed spin is different, and takes the vector of the ball’s movement and plots how the baseball <em>should</em> be spinning. These often deviate slightly, as there are more factors that go into movement besides just the way it spins. To graph a clock that visualizes both of these, we use the following code.</p><pre>tilt &lt;- function(player, game_year, data) {<br><br> #transform the spin data to make sure that data on both sides of 12 are in<br> #fact different, then pivot<br>  pitches &lt;- spin %&gt;%<br>  filter(player_name == player) %&gt;%<br>  mutate(hawkeye_measured = -hawkeye_measured + 540,<br>         movement_inferred = -movement_inferred + 540) %&gt;%<br>  mutate(hawkeye_measured = ifelse(hawkeye_measured &gt; 360, hawkeye_measured - 360, hawkeye_measured),<br>         movement_inferred = ifelse(movement_inferred &gt; 360, movement_inferred - 360, movement_inferred)) %&gt;%<br>  mutate(hawkeye_measured = as.numeric(hawkeye_measured),<br>         movement_inferred = as.numeric(movement_inferred)) %&gt;%<br>  rename(Measured = hawkeye_measured,<br>         Observed = movement_inferred) %&gt;%<br>  select(`Pitch Type`, Measured, Observed) %&gt;%<br>  pivot_longer(cols = c(Measured, Observed), <br>               names_to = &quot;measurement_type&quot;, <br>               values_to = &quot;angle&quot;)<br>  <br>#create the plot of a clock<br>  hour_positions &lt;- data.frame(<br>    hour = 1:12,<br>    angle = seq(30, 360, by = 30)<br>  )<br>  <br>  types &lt;- data %&gt;% <br>     filter(name == player&amp; year == game_year) %&gt;% <br>    group_by(pitch_name) %&gt;% <br>    summarise(n = 100 * n() / nrow(data %&gt;% <br>     filter(name == player &amp; year == game_year))) %&gt;% <br>    filter(n &gt;= 1) <br>  <br>  types &lt;- unique(types$pitch_name)<br>  <br>#plot an arrow with the tilt on the clock plot<br>  pitches %&gt;% <br>    filter(`Pitch Type` %in% types) %&gt;% <br>  ggplot() +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = 1), size = 1, color = &quot;black&quot;, fill = &quot;#ffffff&quot;) +<br>    geom_text(data = hour_positions,<br>              aes(x = 0.85 * cos(pi/180 * (90 - angle)),<br>                  y = 0.85 * sin(pi/180 * (90 - angle)),<br>                  label = hour), size = 5) +<br>    geom_segment(aes(x = 0, y = 0,<br>                     xend = cos(pi/180 * (90 - angle)),<br>                     yend = sin(pi/180 * (90 - angle)),<br>                     color = `Pitch Type`),<br>                 size = 1.5, arrow = arrow(length = unit(0.5, &quot;cm&quot;))) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = .25), size = 1, color = &quot;black&quot;, fill = &quot;white&quot;) +<br>    geom_circle(aes(x0 = 0, y0 = 0, r = .06), size = 1, color = &quot;black&quot;, fill = &quot;black&quot;) +<br>    scale_color_manual(values = pitch_colors) +<br>    ggtitle(&quot;&quot;) +<br>    coord_fixed() + <br>    theme_void() +<br>      facet_wrap(~measurement_type, ncol = 2) + <br>    theme(legend.position = &quot;none&quot;, <br>          strip.text = element_text(size = 20))<br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/872/1*W2waCAQzK40rEhZXVLB1vw.png" /></figure><p>Finally there is only one more plot to create, but it is the most extensive. It uses the trajectory data to plot a one standard deviation kde of the release and approach angles for each pitch type. These are important in understanding how similar they appear to a hitter.</p><pre>approach_angles &lt;- function(player, game_year, data){<br>  <br>  dat &lt;- data %&gt;% <br>   filter(name == player &amp; year == game_year)<br>  <br>  dat &lt;- dat %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    mutate(<br>      pct = n() / nrow(dat %&gt;% filter(name == player))<br>    ) %&gt;% <br>    filter(pct &gt;= .01 &amp; pitch_name %in% names(pitch_colors)) %&gt;% <br>    ungroup()<br><br>  #only for release and approach, we plot the coordinates of where a tangent<br>  #line from the coordinate through the previous coordinate would cross the <br>  #plate. This coordinate is then used later to calculate the angle<br>  for (i in c(1, 50)) {<br>  dat &lt;- dat %&gt;%<br>    mutate(<br>      !!paste0(&quot;z_pos_&quot;, i - 1) := release_z + bz * (((i - 1) / 50) * plate_time) + 0.5 * az * (((i - 1) / 50) * plate_time)^2,<br>      !!paste0(&quot;z_pos_&quot;, i) := release_z + bz * ((i / 50) * plate_time) + 0.5 * az * ((i / 50) * plate_time)^2,<br>      !!paste0(&quot;y_pos_&quot;, i - 1) := release_y + by * (((i - 1) / 50) * plate_time) + 0.5 * ay * (((i - 1) / 50) * plate_time)^2,<br>      !!paste0(&quot;y_pos_&quot;, i) := release_y + by * ((i / 50) * plate_time) + 0.5 * ay * ((i / 50) * plate_time)^2, <br>      !!paste0(&quot;x_pos_&quot;, i - 1) := release_x + bx * (((i - 1) / 50) * plate_time) + 0.5 * ax * (((i - 1) / 50) * plate_time)^2,<br>      !!paste0(&quot;x_pos_&quot;, i) := release_x + bx * ((i / 50) * plate_time) + 0.5 * ax * ((i / 50) * plate_time)^2<br>    ) %&gt;%<br>    mutate(<br>      !!paste0(&quot;slope_&quot;, i) := (!!sym(paste0(&quot;z_pos_&quot;, i)) - !!sym(paste0(&quot;z_pos_&quot;, i - 1))) / (!!sym(paste0(&quot;y_pos_&quot;, i)) - !!sym(paste0(&quot;y_pos_&quot;, i - 1))),<br>      !!paste0(&quot;vertical_angle_&quot;, i) := -atan(!!sym(paste0(&quot;slope_&quot;, i))) * (180 / pi), <br>      !!paste0(&quot;slope_x_&quot;, i) := (!!sym(paste0(&quot;x_pos_&quot;, i)) - !!sym(paste0(&quot;x_pos_&quot;, i - 1))) / (!!sym(paste0(&quot;y_pos_&quot;, i)) - !!sym(paste0(&quot;y_pos_&quot;, i - 1))),<br>      !!paste0(&quot;horizontal_angle_&quot;, i) := -atan(!!sym(paste0(&quot;slope_x_&quot;, i))) * (180 / pi)<br>    )<br>  }<br> #plot the mean angle for each pitch, as well as a one sd kde around it<br>p1 &lt;- dat %&gt;%<br>  group_by(pitch_name) %&gt;% mutate(mean_hra = -mean(horizontal_angle_1), mean_vra = mean(vertical_angle_1)) %&gt;% <br>  ggplot(aes(x = -horizontal_angle_1, y = vertical_angle_1, color = pitch_name)) +<br>  stat_ellipse(type = &quot;norm&quot;, level = 0.68, linetype = &quot;dashed&quot;, linewidth = 1.25) +<br>  geom_point(aes(mean_hra, mean_vra, fill = pitch_name, color = &quot;black&quot;), size = 5, shape = 21) +<br>  labs(<br>    x = &quot;Horizontal Release Angle&quot;,<br>    y = &quot;Vertical Release Angle&quot;,<br>    color = &quot;Pitch Type&quot;<br>  ) +<br>  theme_minimal() + <br>  scale_colour_manual(values = pitch_colors) + <br>  scale_fill_manual(values = pitch_colors) + <br>  theme(legend.position = &quot;none&quot;, <br>        axis.text.x = element_text(size = 20),<br>                  axis.text.y = element_text(size = 20),<br>                  axis.title.x = element_text(size = 20),<br>                  axis.title.y = element_text(size = 20)) <br><br>p2 &lt;- dat %&gt;%<br>  group_by(pitch_name) %&gt;% mutate(mean_hra = -mean(horizontal_angle_50), mean_vra = mean(vertical_angle_50)) %&gt;% <br>  ggplot(aes(x = -horizontal_angle_50, y = vertical_angle_50, color = pitch_name)) +<br>  stat_ellipse(type = &quot;norm&quot;, level = 0.68, linetype = &quot;dashed&quot;, linewidth = 1.25) +<br>  geom_point(aes(mean_hra, mean_vra, fill = pitch_name, color = &quot;black&quot;), size = 5, shape = 21) +<br>  labs(<br>    x = &quot;Horizontal Approach Angle&quot;,<br>    y = &quot;Vertical Approach Angle&quot;,<br>    color = &quot;Pitch Type&quot;<br>  ) +<br>  theme_minimal() + <br>  scale_colour_manual(values = pitch_colors) + <br>  scale_fill_manual(values = pitch_colors) + <br>  theme(legend.position = &quot;none&quot;, <br>        axis.text.x = element_text(size = 20),<br>                  axis.text.y = element_text(size = 20),<br>                  axis.title.x = element_text(size = 20),<br>                  axis.title.y = element_text(size = 20))<br><br><br>grid.arrange(p1, p2, ncol = 2)<br><br>}</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/916/1*yaDvN-WlQ76RzgAoirlO0A.png" /></figure><p>Time to arrange all the plots. We have two problems though.</p><ol><li>We will be using the gridExtra library to put everything together, but this format only takes in grobs, or ggplot objects. The summary table we created earlier is a gt object, not a grob. There is no good way to convert it, so we create one. We will create a function that takes an screenshot of the table, saves it, then reads it back in as a grob.</li></ol><pre>gt_to_grob &lt;- function(gt_table) {<br>  temp_file &lt;- tempfile(fileext = &quot;.png&quot;)<br>  gtsave(gt_table, filename = temp_file)<br>  img &lt;- png::readPNG(temp_file)<br>  grob &lt;- rasterGrob(img, interpolate = TRUE)<br>  return(grob)<br>  }</pre><p>2. Not every plot can be the same size, because the table is different sizes depending on how many different pitches our selected pitcher throws. To solve this we have different dimensions depending on this number.</p><p>With that, here is the function that creates the final summary.</p><pre>pitching_dashboard &lt;- function(player, game_year, output_file) {<br>  <br>  <br>  data &lt;- data %&gt;% <br>    filter(year == game_year)<br>  <br>  trajectory_data &lt;- trajectory_data %&gt;% <br>    filter(year == game_year)<br><br>  stats_gt &lt;- pitch_summary(player, game_year, data)        <br>  stats_plot &lt;- gt_to_grob(stats_gt)<br>  velo_plot &lt;- velo(player, game_year, data)               <br>  movements_plot &lt;- movement(player, game_year, data)    <br>  tilts &lt;- tilt(player, game_year, data)<br>  info &lt;- player_info(player)<br>  angles_plot &lt;- approach_angles(player, game_year, trajectory_data)<br>  tunnels_plot &lt;- tunnels(player, game_year, trajectory_data)<br>  <br> <br>  <br>   pitches &lt;- data %&gt;%<br>    filter(name == player) %&gt;%<br>    group_by(pitch_name) %&gt;%<br>    mutate(<br>      pct = n() / nrow(data %&gt;% filter(name == player))<br>    ) %&gt;% <br>    filter(pct &gt;= .01) %&gt;% <br>    ungroup() %&gt;% <br>     summarise(unique_count = n_distinct(pitch_name)) %&gt;%<br>        pull(unique_count)<br>  <br>  if (pitches &lt; 3){<br>    png(output_file, width = 20, height = 15, units = &quot;in&quot;, res = 300)<br>    <br>     grid.arrange(<br>       arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3),<br>      c(4)<br>    ),<br>    heights = c(.25, 0.3, 0.5, 0.45)<br>     )<br>  <br>  } else if (pitches == 3){ <br>     png(output_file, width = 20, height = 16, units = &quot;in&quot;, res = 300)<br>    <br>    grid.arrange(<br>       arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3),<br>      c(4)<br>    ),<br>    heights = c(.25, 0.4, 0.5, 0.45)<br>    )<br>  <br>  } else if (pitches == 4){<br>    png(output_file, width = 20, height = 18, units = &quot;in&quot;, res = 300)<br>    <br>    grid.arrange(<br>      arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3),<br>      c(4)<br>    ),<br>    heights = c(.25, 0.5, 0.5, 0.5)<br>  )<br>  } else if(pitches == 5){<br>      png(output_file, width = 20, height = 20, units = &quot;in&quot;, res = 300)<br>    <br>     grid.arrange(<br>       arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3),<br>      c(4)<br>    ),<br>    heights = c(.25, 0.5, 0.5, 0.45)<br>  )<br>  } else if(pitches == 6){<br>      png(output_file, width = 20, height = 20, units = &quot;in&quot;, res = 300)<br>    grid.arrange(<br>      arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3),<br>      c(4)<br>    ),<br>   heights = c(.25, 0.6, 0.5, 0.45)<br>  )<br>  } else if (pitches == 7){<br>      png(output_file, width = 20, height = 22, units = &quot;in&quot;, res = 300)<br>    <br>    grid.arrange(<br>    arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3), <br>      c(4)<br>    ),<br>    heights = c(.25, 0.65, 0.5, 0.45)<br>  )<br>  } else{<br>       png(output_file, width = 20, height = 22, units = &quot;in&quot;, res = 300)<br>    <br>     grid.arrange(<br>       arrangeGrob(info, ncol = 1, <br>                heights = unit(1, &quot;null&quot;), widths = unit(20, &quot;in&quot;)),<br>    arrangeGrob(stats_plot, ncol = 1, heights = unit(3, &quot;null&quot;), widths = unit(20, &quot;in&quot;)), <br>    <br>    arrangeGrob(<br>      velo_plot, movements_plot, tunnels_plot, <br>      ncol = 3, widths = c(1/3, 1/3, 1/3), heights = unit(1, &quot;null&quot;)<br>    ),<br>    <br>    arrangeGrob(tilts, angles_plot, ncol = 2, heights = unit(2, &quot;null&quot;)),<br>    <br>    layout_matrix = rbind(<br>      c(1),  <br>      c(2),  <br>      c(3),<br>      c(4)<br>    ),<br>    heights = c(.25, 0.7, 0.5, 0.45)<br>  )<br>    }<br>  <br>  <br>  dev.off()<br>}<br><br>pitching_dashboard(&quot;Zack Wheeler&quot;, 2024, &quot;summary.png&quot;)</pre><p>And here it is, Zack Wheeler’s pitching summary for 2024!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*raveA2q_rgeB7WDZdiLvpg.png" /></figure><p>If you’ve made it this far, go ahead and give it a try! Baseball analytics are becoming a bigger part of the sport, and making cool things like this fascinates me. As always this can be easily integrated into another coding language of your choice. Make sure to give it a go, and I hope that this article was useful and interesting. For another example, because why not, here is the newest member of the Boston Red Sox, Garrett Crochet. Unfortunately until he pitches a game, he will still have the White Sox logo, but his headshot is updated.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*MD5IgL9hTfc4mcmYroljKw.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=046fb4f4ec25" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Mariners and Red Sox are a Perfect Match, but Finding a Middle Ground Might be More Difficult…]]></title>
            <link>https://medium.com/@bsresnic/the-mariners-and-red-sox-are-a-perfect-match-but-finding-a-middle-ground-might-be-more-difficult-67c27f432de7?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/67c27f432de7</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Fri, 13 Dec 2024 21:56:49 GMT</pubDate>
            <atom:updated>2024-12-13T21:56:49.175Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Mariners and Red Sox are a Perfect Match, but Finding a Middle Ground Might be More Difficult than it Seems</h3><p>For months now, the Red Sox and Mariners have been linked together as potential trade partners. The Mariners have tons of quality pitching, but can’t hit, and the Red Sox have an excess of hitters, but the pitching isn’t where it could be. As a team last year, the Mariners led MLB in ERA and ranked third in FIP. On the other side, they had a .687 team OPS and scored the 10th fewest runs. A Boston bat for a Seattle arm has always made sense, but a trade can be complicated.</p><p>After finally making a big splash for Garrett Crochet, it has been reported that Boston is still looking for more arms.</p><iframe src="https://cdn.embedly.com/widgets/media.html?type=text%2Fhtml&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;schema=twitter&amp;url=https%3A//x.com/jonmorosi/status/1866970941223473275&amp;image=" width="500" height="281" frameborder="0" scrolling="no"><a href="https://medium.com/media/10505c5d5c854e33e58669f3538b2104/href">https://medium.com/media/10505c5d5c854e33e58669f3538b2104/href</a></iframe><p>Shortly after the swap became official rumors spread like wildfire about a Triston Casas for Luis Castillo swap, which is the general framework of a trade that has been talked about for a long time now. On Wednesday afternoon, Adam Jude, a writer with the Seattle Times, reported that “The Mariners and Red Sox have engaged in trade talks. Boston has been persistent in its pursuit of one of Seattle’s starters, and a Luis Castillo-for-Triston Casas deal does seem to make sense for each side.” Then approximately thirty minutes later, Alex Speier shut the rumor down completely.</p><iframe src="https://cdn.embedly.com/widgets/media.html?type=text%2Fhtml&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;schema=twitter&amp;url=https%3A//x.com/alexspeier/status/1866983449330168222&amp;image=" width="500" height="281" frameborder="0" scrolling="no"><a href="https://medium.com/media/c49eb67495e6268faf18c2c65308f431/href">https://medium.com/media/c49eb67495e6268faf18c2c65308f431/href</a></iframe><p>Where there’s smoke there’s fire. I would not be surprised in the slightest if these two teams do end up making a similar trade. In fact I hope it happens, because at the end of the day it does make all the sense in the world, but finding middle ground for a fair trade isn’t as simple as it looks.</p><p>The Mariners have the best rotation in baseball, with Logan Gilbert, George Kirby, Bryce Miller, Bryan Woo, and Luis Castillo. The problem is that it doesn’t seem like the Mariners want to move off from any of them. And if they do, it would be Castillo. The more I think about it, the less Luis Castillo as a Red Sox would make sense. Not only due to his age, contract situation, and arsenal, but also what it would take to get him. 2024 was Castillo’s worst season since breaking out as an All Star in 2019. That doesn’t mean he was bad: Castillo still finished in the 86th percentile of pitching run value, but the underlying metrics are on a decline. Since that season, this past year he posted his worst numbers in ERA+, FIP, K/9, and HR/9. His stuff has also gone downhill. 2024 was the first time in his career his slider dipped below average, and his arsenal as a whole did not grade out well.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LxnYsx6QzBB1_Syw5STdRg.png" /></figure><p>Although stuff plus isn’t everything, a trend like this one can be dangerous, due to the stickiness of stuff plus. Stickiness refers to how stuff plus generally doesn’t change year over year. Trends like these are rare, so when they do occur they’re noteworthy, and Castillo having two pitches heading in the wrong direction could be troublesome.</p><p>I would trust Breslow and Bailey to fix it, but at a certain point it has to be thought that he might just simply be regressing. Castillo threw his four seamer 45% of the time last year at an 84 stuff+. I can assure anyone with 100% confidence that it wouldn’t occur on the 2025 Red Sox, but even his sinker still had a negative run value. The fastball is right in the dead zone, where his horizontal and vertical movement numbers are the same.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bEspTlPl7N3z32PVn2aAgA.png" /></figure><p>Castillo also has 3 years at $24.15 million left on his contract, so the deal involves the Red Sox taking on his money. My issue with this isn’t them adding the payroll, it’s the way they’re doing it. If the front office is willing to take on $24 million a year for a pitcher’s age 32–34 seasons, great. But why not just sign Corbin Burns at that point? Yes he would cost more dollars wise, but with it being a signing the only thing the Red Sox would lose would be the pick attached to the qualifying offer.</p><p>The Mariners don’t want prospects from the Red Sox; they’re trying to win baseball games as well. So whatever Boston gives up in a potential deal would, and should be big league talent. Triston Casas or Wilyer Abreu by themselves for Luis Castillo is a massive overpay, and the Mariners would likely ask for more than just that one piece. The Red Sox counter to that proposal would likely be a massive underpay, because besides those two that are constantly being floated in trade rumors, the Red Sox don’t have any other hitters on the 40 man roster that it would make sense for them to part with that would also be enough to headline this potential trade.</p><p>On the other hand, if Boston were to pivot to asking for Gilbert, Kirby, Miller, or Woo, Casas or Abreu might not be enough for the Mariners. At that point the Red Sox would walk away from the talks. It’s as if a package that the Red Sox are willing to part with is too much for Castillo but not enough for any of the other starters, and Castillo isn’t enough for the Red Sox to get back for Casas or Abreu, but the rest of the Mariner’s starters would be an overpay. There’s no middle ground!</p><p>All in all I don’t think a trade between these two teams will end up happening. Especially if the Mariners keep being stingy. I don’t think the Mariners actually realize how badly they need hitting in that ballpark. In my opinion, something like Wilyer Abreu and Vaughn Grissom for Bryce Miller makes all the sense in the world, but I’m not in charge, so who knows.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=67c27f432de7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Who Should Be the Red Sox Closer in 2025?]]></title>
            <link>https://medium.com/@bsresnic/who-should-be-the-red-sox-closer-in-2025-38270beaf811?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/38270beaf811</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Tue, 03 Dec 2024 15:51:50 GMT</pubDate>
            <atom:updated>2024-12-03T15:56:31.975Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8G2BEj2R7uuNJ4GDjFmeDw.jpeg" /></figure><p>It’s no secret that the Red Sox need bullpen help. In 2024, they had the third worst save percentage, and the seventh worst pen ERA. With Kenley Jansen likely gone next year, the Sox will need to hand off the closer role to someone else. The tough question though is who? Kenley provided valuable innings the last two years, saving 56 games in Boston with a solid 128 ERA+. They have three options for 2025: someone internal, a trade, or a free agent signing. I’ll rank them in order of least to most likely options, and explain some candidates for the role in each category.</p><p><strong><em>Signing a Free Agent Closer</em></strong></p><p>Remember, pitchers are expensive. Taking one look at the ones that have already come off the board proves that. The Red Sox however have been telling everyone in the media that they’re planning on spending. But in order to prove it, the Red Sox would have to go full throttle and drop some cash instead of just talking a big game about it.</p><p><strong><em>Tanner Scott </em></strong>— 4 years $68 million ($17 million AAV)</p><p>The best closer on the market, Scott had a tremendous season in 2024, and forced the Padres to give up a ton to get him. Scott’s two pitches are both deadly, and he gives the Sox what they need more than almost anything: a left handed pitcher. Now on the free agent market, if the Red Sox are serious about spending, one way to show it is to spend big money for a reliever. Unfortunately he’s probably going to cost more than the Red Sox are willing to spend.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2prVX6YMkDEQfMW46Xp9TA.png" /></figure><p><strong><em>Jeff Hoffman </em></strong><em>— </em>3 years $39 million ($13 million AAV)</p><p>Hoffman had a career resurgence when he signed with the Phillies prior to the 2023 season. Since then, all he’s done is pitch to a sub-3 FIP and ERA in 118.2 innings, most of which in high leverage situations. Some view Hoffman as a reliever to starter transformation candidate, but I think that he’s best served as a closer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fgr_9O2ASgePB38zA_Sh7w.png" /></figure><p><strong><em>Paul Sewald</em></strong><em> — </em>1 year $10 million</p><p>Sewald made a name for himself in Seattle before being traded to Arizona at the 2023 trade deadline. Sewald’s two pitch arsenal is very unique, as his fastball is at the bottom of the league in velocity, but it flourishes due to the flat vertical approach angle, which is caused by his arm angle. His time with the Diamondbacks didn’t go as exactly as planned, and maybe it’s time for him to switch coasts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wtJb3ooWFlc80d3zU4JGCA.png" /></figure><p><strong><em>Making a Deal</em></strong></p><p>Primarily built by Chaim Bloom, the farm system is as good as it’s ever been. Now it’s time to cash in on it. Craig Breslow traded for two relievers at the trade deadline last year and missed on both. With one of these trades, he can prove to ownership, the fanbase, and current free agents that the Red Sox are back and ready to win.</p><p>Red Sox receive: <strong><em>Devin Williams</em></strong></p><p>Brewers receive: Kutter Crawford, Chase Meidroth</p><p>Devin Williams has undoubtedly the best changeup I’ve ever seen. The Brewers have already shown how they value Williams by declining his option and saving a few bucks through arbitration. The Brewers gain control with both Crawford and Meidroth, while the Red Sox get Williams and look to the checkbook for an extension. All we can hope for is that he doesn’t punch any more walls. Giving up Kutter is a hard pill to swallow, but would be based on them signing at least one starter in free agency, which is likely. I like Meidroth, but similar to a higher ranked prospect I’ll mention later, the only reason he’s yet to debut is because there’s nowhere for him to play. If Williams pitches a whole year, he’ll end up as a top three closer in MLB easily. The Red Sox need a lockdown guy like that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*euODvLVCynaucFTUBdQ_CA.png" /><figcaption>Note: Cutter’s sample size is too small to register spin directions</figcaption></figure><p>Red Sox receive:<em> </em><strong><em>Alexis Díaz</em></strong></p><p>Reds receive: Luis Guerrero, Bryan Mata</p><p>This is a classic example of trades costing less in the offseason, and buying low-ish on Díaz after a “down” year. I think this is a fair trade. Díaz won’t be a free agent until 2028, and has proven experience as a closer. In 2022 he saved 37 games, and the year before he pitched to a 1.84 ERA, although the xERA was almost a run higher and the FIP started with 3. Guerrero was excellent in his short stint to end last year, but his potential flatlines as a 7th, maybe 8th inning guy. Nothing can make me comprehend how Bryan Mata is still a prospect, and I think it’s time to give him a fresh start in Cincinnati. For context, when Mata signed as an IFA, David Ortiz was the Red Sox DH. Throw Díaz in the Andrew Bailey lab of pitching to re-find himself, and the Red Sox might just find themselves a closer for the next three years.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oUbIs-ilyWnEomvoqbBwag.png" /></figure><p>Red Sox receive:<strong> <em>Ryan Helsley</em></strong><em>, </em>Nolan Arenado</p><p>Cardinals receive: Braden Montgomery, Blaze Jordan</p><p>Giving up Montgomery without him making his pro debut is a tough ask, but when we reach his 2027 ETA, where would he play in the outfield? Especially with Juan Soto in left (A man can hope). You can find my thoughts on Nolan Arenado at Fenway<a href="https://medium.com/@bsresnic/the-red-sox-trading-for-nolan-arenado-would-be-a-net-neutral-at-best-7b62efeafa4a"> here</a>, but with the Helsley trade, it makes the most sense to add him in and commit to the position changes. Helsley is coming off winning NL Reliever of the Year, with 49 saves and a stunning 206 ERA+. Blaze Jordan is a much bigger name than the value he would add, but let’s be honest, there’s no way Chaim Bloom would make a trade with the Red Sox without getting someone he drafted. Like Williams, Helsley is a free agent after 2025, so you don’t make this trade without an extension. That being said, if the Red Sox want to make a win now statement to the rest of baseball, this is how they do it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*whhP86e44auJMB-speTslg.png" /></figure><p><strong><em>Promoting from Within</em></strong></p><p>This is the most likely option, because that it obviously wouldn’t cost anything. And given the recent lack in spending, I’m not totally counting on them making a big splash that would cost money or prospects.</p><p><strong><em>Zach Penrod</em></strong></p><p>Penrod is definitely the least likely outcome of the three options I’ve come up with, but it wouldn’t be the end of the world. Penrod has great stuff, and has shown that his biggest flaw is being lengthened out. He could be a serviceable closer in 2025 if called upon.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CZOdZzfRQAhNrr0KRiEOuw.png" /></figure><p><strong><em>Garrett Whitlock</em></strong></p><p>Remember him? Whitlock posted a 2.8 bWAR his rookie year out of the bullpen. Then the Red Sox decided to start him and the injuries came and the results worsened. The experiment failed, and it’s okay because Whitlock, when healthy, is a force out of the pen. Although not a save situation, Whitlock did record the final three outs in the 2021 AL Wild Card Game, and many more big outs throughout those playoffs. His experience and Cora knowing exactly what he’s getting makes Whitlock a good option, but in my eyes there’s one better candidate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N9SmFGwhtnRo-pWm2ekE_Q.png" /></figure><p><strong><em>Justin Slaten</em></strong></p><p>Slaten is the most likely closer out of everyone I’ve listed. The Red Sox threw Slaten into the fire in his first career appearance against the Mariners, giving up a walk-off to Julio Rodriguez. Slaten pitched the majority of his innings last year in the 7th and 8th innings. In the 9th, he threw 6 innings of no run ball. His combo of elite vertical break on the fastball with 3 plus secondaries makes him the Red Sos most likely closer in my opinion. Entering his second season, Slaten will emerge as one of the best high leverage relievers in baseball. The question is what role that’ll be in.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*InsE8g6nkotMdNKWwXf9Cg.png" /></figure><p>You may notice I didn’t mention Liam Hendriks. I’m not sure that coming off his lengthy injury that he should be thrust into such a crucial role. As I’m writing this, it’s being reported that the Red Sox are working towards a deal with Aroldis Chapman. At 36, he’s still got it. I’m not sure what their plan is with him, but last year with the pirates he pitched most of his innings in the 8th as a setup guy. Chapman to Slaten in the last two innings to finish out a close game could a deadly combo if they go down that route, and Chapman is the lefty flamethrower the Red Sox need.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j2wRlPypaIMccO73uGRB-g.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=38270beaf811" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Red Sox trading for Nolan Arenado would be a net-neutral at best]]></title>
            <link>https://medium.com/@bsresnic/the-red-sox-trading-for-nolan-arenado-would-be-a-net-neutral-at-best-7b62efeafa4a?source=rss-7db84f61f81f------2</link>
            <guid isPermaLink="false">https://medium.com/p/7b62efeafa4a</guid>
            <dc:creator><![CDATA[Ben Resnic]]></dc:creator>
            <pubDate>Wed, 27 Nov 2024 19:35:34 GMT</pubDate>
            <atom:updated>2024-11-27T19:38:21.605Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*u343lNxf7uLynObd" /></figure><p>Believe it or not, the Red Sox already have a third baseman. His name is Rafael Devers, he’s one of the best all around hitters in MLB, and he isn’t going anywhere. That being said, his defense at third base is a problem that the Red Sox would be wise to address. Since 2019, Devers has put up seasons of -4, -13, -2, -8, and -6 Outs Above Average. So what should the Red Sox do to fix this issue? I’m not sure Nolan Arenado is the number one priority. Arenado would be a stellar replacement there, and the idea of upgrading defense at third makes even more sense if the Red Sox signing Max Fried is a real conversation, given his nature as a heavy ground ball pitcher. However, his bat continues to get worse, and even if his defense stays at premium quality, how much would the Red Sox actually improve by switching up the diamond like this?</p><p>I’ll start with the upside. Nolan Arenado and Trevor Story reuniting on the left side of the infield would be the best defensive tandem in baseball. For a team that has struggled so mightily on that side of the game, the Red Sox truly could go from the bottom to the top in defense with just this move. I don’t need to remind everyone how good Nolan Arenado is at playing the hot corner, so I won’t write much about it.</p><p>Unfortunately, that’s all the upside there is. It’s extremely important, and a +16 OOA swing would significantly raise the ceiling of the team, but the defense would be the only upgrade, albeit a huge one.</p><p>I still have lots of questions though. This talk of the him being open to a position change seems real fishy. Why in the hell would he be willing to change positions, when he’s one of the best defensive third baseman of all time and is still near the top of his game on that side? On the wrong side of thirty, Arenado’s game is built on his defense, and moving him from third to first would take away all of his value. Imagine if the Red Sox stuck with Ceddanne Rafaela at short this year just for the “fit”. It’s all a bunch of nonsense. If the Red Sox do trade for Arenado, there is a zero percent chance that they would entertain this idea, so I won’t even write about how terrible that would be, but the fact that it’s even being floated around makes me a bit scared about how his overall game will age. This feels like a “we made a move just to say we’ve made a move” type of deal. Especially if they miss out on all the top free agents, which I’m not ruling out. Remember, the Red Sox have <strong>far</strong> bigger issues than this. It’s also important to remember that he’s not a free agent, and I don’t think that this trade would necessarily be on the cheap side. Most fans think that because Chaim Bloom signed Yoshida, he must still like him and would be willing to take on his contract in such a trade. I’m having a hard time believing that the Cardinals, who would begin a rebuild by trading away a player of Arenado’s caliber, would want to take on that money instead of just prospects, even if Chaim still likes Yoshida.</p><p>Arenado’s bat is in decline in almost every way, which isn’t uncommon for a 33 year old corner infielder. I don’t want to call his third place MVP finish 2022 season an outlier, but when you compare it to the rest of his time in St. Louis, it kinda is. Since being traded, the batting component of his fWAR has gone 9.6, <strong>35.5</strong>, 4.9, 1.7. Even going back one more year to the shortened COVID season in Colorado, he posted a -6.1. Overall, his offensive game is just in a slow decline.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SBlriBtG4--8uylTFEEeiA.png" /><figcaption>WPA stands for win probability added</figcaption></figure><p>Just looking at swing path, he has 28th percentile bat speed, and a long swing length in the 15th percentile at 8.2 feet. Not a good combo, and the bat speed will likely only decrease unless Arenado lives at Driveline over the winter. His average exit velocity has also plummeted over the years as he’s aged, along with every correlated metric.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QfWYJHsGLynCkXqAdbpEvQ.png" /></figure><p>Along with all this, Arenado’s batted ball profiles aren’t as Fenway friendly as people think. If he played all of his games at Fenway last year, he would have hit 12 homers, 4 less than he actually did. Arenado doesn’t pull fly balls at an extraordinarily high clip, unlike what you’d want out of a righty at Fenway.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fz3q5owYFk1IekRMbXTE8w.png" /><figcaption>NOTE: Percentages are based strictly on spray and launch angles, not the official classifications by scorekeepers. Due to this, percentages differ from batted ball profiles on Baseball Savant and other sites</figcaption></figure><p>The Red Sox should entertain moving Devers to DH, and open up a spot for better defense at third. People try to make the argument that having Devers be that full time DH would be a terrible look because they shouldn’t be paying north of $30 million for that position. He’s already been paid, and there is nothing that can change that. If the team will win more games with Devers at DH and a positive defender at third, the Red Sox and Devers need to swallow their pride and do it. Also, it is likely that Devers would not only stay healthier not playing the field, but would be a better hitter as well. I like the idea of this, and if the Sox prioritize upgrading third base defense, fine, but that player just shouldn’t be Nolan Arenado. With his bat in a decline, the Red Sox should look elsewhere to fill this roster hole. They can knock two birds with one stone with other players that may not bring platinum glove quality defense such as Arenado, but are still a major upgrade over Devers and will provide offensive output.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7b62efeafa4a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>