<?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 RG Pitching -- Art of W.A.R.  on Medium]]></title>
        <description><![CDATA[Stories by RG Pitching -- Art of W.A.R.  on Medium]]></description>
        <link>https://medium.com/@reinha3?source=rss-8dcb15e5d8e2------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*3czaZ210JP7Z4X-Q</url>
            <title>Stories by RG Pitching -- Art of W.A.R.  on Medium</title>
            <link>https://medium.com/@reinha3?source=rss-8dcb15e5d8e2------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 16:19:53 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@reinha3/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[Enhancement of Collegiate Pitching Plus Model With Arsenal Coherence Effects]]></title>
            <link>https://medium.com/@reinha3/enhancement-of-collegiate-pitching-plus-model-with-arsenal-coherence-effects-c633e3dfe5cf?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/c633e3dfe5cf</guid>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Thu, 13 Mar 2025 21:56:03 GMT</pubDate>
            <atom:updated>2025-03-13T21:56:03.339Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Abstract</strong></p><p>Pitching+ models have taken over baseball analytics as an advanced way to predict and evaluate pitching performance. Last year, we developed a set of Division III Stuff+, Location+, and Pitching+ models. Our Stuff+ quantifies the value of pitches in a vacuum, a simple way to evaluate pitch value. However, pitches do not exist in a vacuum. Pitchers rely on multiple pitches in conjunction to create an arsenal that can deceive hitters. Stuff+ evaluates the physical tools of a pitcher, but arsenal design is their blueprint for success. This project proposes a new way to quantify and evaluate arsenal coherence at the collegiate level.</p><p>Research on arsenal coherence has become a topic of growing interest. Marek Ramilo and Jack Lambert of Driveline released their Mix+ and Match+ statistics. Match+ is described as “how long pitches in a player’s arsenal remain on the same trajectory.” (Link). They define Mix+ as “the ability to make things move in different ways.” (Link) Their work helped advance Pitching+ to evaluate a pitcher more comprehensively. Another similar approach was Dylan Drummey of Prospects Live, titled Arsenal+. Arsenal+ is described as a “Pitching+ metric on steroids” an evaluation dependent on changes from one pitch to another. (Link). Arsenal+ considers prior pitches thrown, another interesting technique to quantify arsenal.</p><p>We propose the quantification of arsenal coherence through the metric of Arsenal +. The purpose of our research into Arsenal + is to understand arsenal coherence at the college level in two ways. How do physical pitch interactions and pitch sequencing affect success in collegiate pitching? We found that the gap between pitchers’ Stuff+ and their actual results can be largely accounted for by their ability to effectively command an arsenal.</p><p>Arsenal Points make Pitching+. a more complete statistic for evaluation, continuing to solve the problem of predicting future success with the smallest possible sample. Arsenal Points also increases detail for development with college pitching, and provides a quantitative understanding of what arsenals are most effective at the collegiate level. Using Arsenal Points as a foundation, pitchers can receive actionable insights for development beyond Stuff + development. This model is another piece to our final goal of creating publicly accessible and easy to use baseball analytics software to data drive baseball development around the country.</p><p><strong>Redefining Run Values</strong></p><p>The first part to creating quality models for player performance is to find a detailed measure of previous performance. Traditionally the best measure for pitching performance has been ERA, but we have now long passed by that tradition. Previously, we have created a Pitching + model for Division III college baseball, using FIP as our measurement for performance. This was done due to the size of our data. However, Art of W.A.R. ‘s data size has grown substantially in the past year. Collegiate run values which previously seemed impossible are now available to us. These run values were used in the Fall of 2024 to recreate our Stuff + Model. Run values simply take into account the value of each pitch a player throws. When breaking down value into each individual pitch, we are able to analyze past performance as accurately as possible.</p><p>Traditional run values simply take into account the outcome of each ball put into play (Flyball, Groundball, Linedrive, Popup), but that leaves nuance in the form of exit velocity.. While the launch angle of a batted ball is something to take into consideration, quality of contact also plays a major role in batted ball results. A 95 EV double in the gap is not more valuable than a slow ground ball simply because it is in the air, but also because it is hit hard. Hard hit balls typically mean a mistake from the pitcher, and should have a different run value than a ball hit softly.</p><p>In our research, we found 87 mph to be the optimal cutoff when delineating between soft and hard hit balls. This may feel like an arbitrary boundary, but there is numerical evidence to support this decision making. It is all in relation to the average wOBACON value for each exit velocity. As figure 1 below shows, when EV is below 87 mph there is a fairly flat slope in wOBACON. Then at 86.77 mph, the slope begins to increase. Thus the barrier is placed there.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/552/1*d2yQEM4rOz5kd_r8wNIcpQ@2x.jpeg" /></figure><p>Figure 1: Graph displaying wOBACON against Exit Velocity of a hit. Showing around 87 mph is where wOBACON begins to increase.</p><p>Other than that distinction our method for calculating run values remains fairly standard. Averages for each outcome were calculated using a set of data from both Division III and Division I games. Our run values are shown in table 1 below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zQDJip0UBFg0dH8-ocyFYg@2x.jpeg" /></figure><p>Table 1: Table displaying Run Values for each pitch outcome with adjusted approach</p><p>The Role of Physical Interaction.</p><p>(Supplemental charts may be found at end of paper)</p><p>One way that a coherent arsenal can be quantified is through the physical interactions between a player’s primary pitch and each of their secondary pitches. This is commonly referred to as “tunneling”. However, to properly model the effectiveness of pitch tunneling, we first need to quantify the effects of different tunnels.</p><p>Our decisions for modeling were all based around using a similar system to the Stuff + models we have created in the past. However, differently to that model we measured each metric in relation to that player’s Fastball. Based on our data we felt it made sense to separate secondary pitches into three groups, Sliders, Curveballs, and Offspeeds. The following table displays how we divided up pitch categories. It is also important to note that secondary fastballs such as sinkers, two seamers etc were tested and removed due to sample size.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s5eIRthLM7sxx9VrCunLRg@2x.jpeg" /></figure><p>Table 2: Table displaying the groupings of each secondary pitch into their larger categories</p><p>From there we determined a set of variables to use in our model. We started with traditional thrown ball metrics and transformed them into variables of percent change from the fastball metrics. We took metrics for each secondary pitch and subtracted them from that player’s Fastball metrics. dividing that difference by the Fastball metric, calculates a percent difference in the pitches physical movement, release, velocity, and spin. This creates variables like horizontal break difference which shows the extent at which the secondary pitch moves differently from that player’s fastball. All of the metrics in the physical interaction work this way except the angle metric.</p><p>The angle metric requires a little geometry, but remains conceptually straightforward. To calculate this metric take the straight line from a pitchers fastball to the offspeed on a typical break chart. Then create another line from that offspeed pitch to the origin of the graph. Find the interior angle between those two lines. That angle’s degree value is this metric. We have previously hypothesized that the angle metric quantifies some efficiency in how two pitches move away from one another, or tunnel. In previous research we have found correlations to this metric and secondary pitch success. Hence its inclusion in the physical interaction model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/548/1*BaqWwV1BiSzjDA58T-cGag@2x.jpeg" /></figure><p>Figure 2: Figure displaying the calculation of the angle metric. The angles shown below each offspeed represent the angle between the red and black lines on the pitch.</p><p>Next we create a detailed set of models which explain the effects of pitch interaction on preventing runs. A machine learning program, XGBoost, was utilized to construct these models. A model unique to each pitch was run. Overall we found each of those models to be very accurate given our data. Here is a series of visuals displaying our individual pitch results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/856/1*DlWDifeaApd7JlDJMg8-Yg@2x.jpeg" /></figure><p>Figure 3: Figure displaying the model results of the Slider specific Physical Interaction model</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/862/1*zp8nJBdUdP-ItYLWfu4FAw@2x.jpeg" /></figure><p>Figure 4: Figure displaying the model results of the Offspeed specific Physical Interaction model</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/855/1*UcAM5R80pKcm8SCcPkePfg@2x.jpeg" /></figure><p>Figure 5: Figure displaying the model results of the Curveball specific Physical Interaction model</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K7qgauctA6T9xiPS_VSyjA@2x.jpeg" /></figure><p>Table 3: Table displaying the numeric results of each pitch specific model. RMSE around 3 typically represents a level of accuracy given data size.</p><p>Results, both visually and numerically indicate a large amount of predictive success for the physical interaction models. The visuals may speak for themselves, but RMSE indicates that each pitch by pitch model does a very good job of holistically predicting the tunneling effects on a player’s success. RMSE here is in units of Run Values indicating that our mean error for all models is around 3 RV.</p><p>Putting Individual Pitch Models Together</p><p>The pitch specific models look good. However, this wouldn’t be a real arsenal model if these models themselves didn’t cohere. A full model allows us to not only understand the tunneling value of each pitch, but see how the entire arsenal of a pitcher works together to deceive hitters. To do this, a second level of model is introduced to quantify the value of adding each pitch into an arsenal. But where is the cutoff for a pitch being part of a player’s arsenal? We have two criteria in order for a pitch to be “qualified”. First, it needs to be thrown at least 5% of the time, and second the pitches Stuff + must be at least an 80 in our collegiate model. If a player doesn’t trust the pitch enough to throw it 5% of the time then hitters probably aren’t worried about that pitch all that much, and the same is true if a pitch has stuff below two standard deviations of average. Getting an arbitrary boost in Arsenal Coherence for throwing a bad 4th pitch 2% of the time didn’t seem right, so these criteria were added.</p><p>From there, another XGBoost was run to find proper values for the addition of any secondary pitches. Table 4 below shows the gain that adding any of these pitches to an arsenal has, determining each of the secondary pitches respective values in comparison to each other.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/744/1*k4RDb-4umY13WdpG_bvpGg@2x.jpeg" /></figure><p>Figure 6: The displayed feature importance of having each qualified pitch in one’s arsenal.</p><p>The above table highlighted specific insights of the model, in conjunction with our Stuff + model, throwing a hard bullet slider remains to be the best offspeed pitch in college baseball regardless of level of play. However, this does not mean that this pitch belongs in every pitcher’s repertoire. Every pitcher is different and requires an individualized approach to optimizing their pitch mix. This research serves as a tool to guide this process.</p><p>Before moving on it is important to analyze the results of the full model after putting everything together. The two leveled models were applied in conjunction using a weighted average, forming a final ‘Physical Interaction’ or ‘Tunnel +’ metric. Results indicate that this model remains a high level of predictive accuracy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/859/1*ObPiaVn7x3zUgJ8X_5yrpg@2x.jpeg" /></figure><p>Figure 7: Figure displaying predicted run values based on full physical interaction model against actual run value. Regression visually indicates predictive success.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9E-Z6_9yIEoX2RS3fCKubw@2x.jpeg" /></figure><p>Table 4: Table displaying numeric results of full physical interaction model</p><p>The Role of Sequencing</p><p>(Full Sequence + results may be found at end of paper)</p><p>While possessing an elite arsenal does a great job of setting a pitcher up for success, they need to be able to effectively utilize that arsenal to succeed. We can quantify trends in usage with Sequence +. Sequence + analyzes pitch selection using in-game context to determine how effective a certain pitch is in each given state.</p><p>The main point of context that pitchers are beholden to is the count. Each of the 12 unique counts carries their own impact on the game state. An 0–2 slider will likely bring about a different result than a 3–0 slider, even if the pitches are thrown in the same spot with the same physical characteristics. To account for this, we created 3 count-states from the 12 available counts. A pitch may be thrown from the ahead, even, or behind in the count. This grouping was done to ensure that each count state had a large enough sample size to be statistically viable.</p><p>The other point of context we used in this research was the prior pitch. Hitters often make guesses on what they believe will be thrown next, due to previous information. One of the main clues is the last pitch that the hitter saw. While this game of cat and mouse goes well beyond just the previous pitch, we found observing sequencing through this lens provided significant results.</p><p>By combining both the count state and previous pitch, we were able to develop a model to quantify Sequencing at the collegiate level. We utilized an xGBoost model, comparing each unique pitch-and-count state against baseline Run Values in a count vacuum. We were able to determine significant benefits that are useful in helping instruct pitchers in how to utilize their arsenal. Figure 8 below shows a visual representation of the decision making tree of our model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/918/1*H2fE6eOzUUGvfmvX49RYCw@2x.jpeg" /></figure><p>Figure 8: Decision making tree of Sequence +</p><p>One of the most interesting results was the high value of throwing a slider-slider sequence when behind in the count. Throwing a slider when behind in the count seems to fly in the face of traditional baseball wisdom, but our model found that this combo can prove to be incredibly effective. The model found the slider-slider combo to be effective in all situations, aligning with our research into importance quality. When utilized effectively, a good slider can be an incredibly important tool in a pitcher’s arsenal.</p><p>The sequencing model we ran was entirely characteristic agnostic. We did not take into account the physical qualities of a pitch, looking solely at its performance in the game state. This was done to avoid any additional statistical noise stemming from these variables. Sequence + is meant as a tool to be used in conjunction with Arsenal + to help guide pitchers towards their best selves. Where Arsenal + may be the ingredients, Sequence + can be seen as the cookbook.</p><p>Supplemental Charts</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/922/1*YCtqv4iAq_UXCMw03D9YfA@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/923/1*LPqCzt0InTZ8wU4juftFdg@2x.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/909/1*eZkQDWWfKZ2wFfWiuAET-A@2x.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c633e3dfe5cf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Examining the Collegiate Pitchers Drafted in 2024
by Expected Run Value]]></title>
            <link>https://medium.com/@reinha3/examining-the-collegiate-pitchers-drafted-in-2024-by-expected-run-value-dc5f3e3d477f?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/dc5f3e3d477f</guid>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Sun, 08 Dec 2024 22:38:04 GMT</pubDate>
            <atom:updated>2024-12-08T22:38:04.922Z</atom:updated>
            <content:encoded><![CDATA[<h2>Examining the Collegiate Pitchers Drafted in 2024 by Expected Run Value</h2><p>By Liam Kennedy</p><p>What qualities should a team search for when selecting a collegiate pitcher? You may covet a blazing fastball, nasty slider, or freakish control. However, these are all just means to an end. The only thing that matters, as a pitcher, is preventing runs.</p><p>There are many ways we attempt to quantify a pitcher’s prowess at run prevention. The most common statistic is ERA, a useful but flawed way to measure pitcher performance as it factors in both defense and batted ball luck. FIP gets us closer to measuring just the pitcher’s performance, but it measures output from a macro level.</p><p>To look at a pitcher’s pitch-by-pitch contribution to run prevention, I have created a run expectancy model that analyzes individual pitches results and their effectiveness at run prevention in the D1 environment. This model looks at both the pitch call and batted ball type. A called strike is good for preventing runs, a ball is bad for preventing runs, and a hard hit line drive is very bad at preventing runs. This way, we can determine how effective a pitcher is at utilizing their arsenal to put themselves in positions that lower their own run expectancy.</p><p>With expected run value in mind, I wanted to take a closer look at collegiate pitchers taken in the 2024 MLB Draft. I looked at the first 20 college pitchers taken off the board and ranked them with regards to their expected run value on a per pitch basis. I then compared this leaderboard to the order these pitchers were selected to see who was drafted below their ERV, above their ERV, or right where ERV projects them to go.</p><p>What I found was quite surprising. 12 of the 20 pitchers were drafted within 2 spots of where ERV would place them, including 7 of the top 10 college arms off the board. If we expand the scope a bit, we find that 16 fell within 5 spots of their ERV projection. This shows that there is some value in using ERV as a guide evaluating pitchers in the collegiate ranks.</p><p>However, we can learn just as much from the 4 “outlier” pitchers. Jurrangelo Cijntje, Khal Stephen, Ryan Johnson, and Ryan Prager all either over or underperformed their ERV ranking by a substantial margin. Diving into what led these pitchers to end up in these positions can help shed some light on the imperfect art of drafting pitchers from the college level.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/585/1*N3r4AmFjW2wKZZnazWMNGw@2x.jpeg" /></figure><p>Jurrangelo Cijntje is by far the biggest outlier on this list. He had the 17th best expected run value in this sample, yet he was the third collegiate pitcher taken in the 2024 draft. However, Cijntje possesses an incredibly unique trait that goes beyond the scope of this model.</p><p>He has demonstrated the ability to be an effective right and left handed pitcher. Both of Cijntje’s arms have garnered pro comps, and his extreme versatility is incredibly enticing at the next level. While he is an unfinished product, his unique traits and potential from both sides has allowed Cijtnje to get drafted well above what his ERV may suggest.</p><p>Khal Stephen was a bit of a late bloomer when it came to draft position. The 6’4 righty was left off of D1 Baseball’s list of top 20 pitchers heading into the 2024 season, yet he ended the year as the 11th pitcher taken off the board in the 2024 draft. He was able to achieve this growth thanks to his consistent arsenal. He has incredible feel for all four pitches he throws, with his high-rise fastball being the standout pitch in his consistent arsenal.</p><p>His ERV disparity may be caused by the size of the study itself. I looked at the expected run value of a pitcher from the 2023 and 2024 season. For much of this sample, Khal Stephen did not look quite like the pitcher that the Blue Jays drafted. His in-season growth as a pitcher is something the model does not account for, but definitely played a role in how teams viewed him in the draft process.</p><p>Ryan Johnson had a spectacular collegiate campaign. The Dallas Baptist product holds the team record for strikeouts in a season and maintains a stellar 10.8:1 K to BB ratio. These numbers had him listed as the 9th best pitcher by ERV, yet he was the 16th collegiate arm off the board. Some of this may be his competition, as he was of three pitchers that did not come from a P4 school.</p><p>However, I believe the main culprit in Johnson’s slide was his delivery. Described as “scary” by scouts, he has a wonky delivery that features a slide step from the wind up that seems to take a lot of effort. This delivery has raised concerns over what it means for Johnson’s durability and consistency.</p><p>Ryan Prager was instrumental in the 2024 Texas A&amp;M season. His College World Series campaign was highlighted by a gem against Kentucky to send his team to the CWS semifinals. These 6.2 innings of shutout ball included 6 no-hit innings, and helped Prager achieve the 10th best ERV in this sample. However, he was the 19th collegiate pitcher drafted.</p><p>Some of this hesitancy may arise from his 2022 Tommy John, which forced him to miss the entire 2023 campaign. Scouts may also be wary of his fastball velocity. In a league where 100 mph fastballs are becoming more and more common each day, Prager provides a bit of nostalgia. His fastball averages 90 mph but is located exceptionally well. There is some concern as to how this pitch will play at the next level, as the quality of hitters only continues to rise.</p><p>While examining expected run value can be a useful tool in evaluating collegiate pitchers, there are other factors that must be considered. By looking at these players with high variance from ERV and draft position, we are able to peer into these other factors that matter to major league teams. We are also able to discern how heavily these teams weigh collegiate performance, giving us a better idea of what to look for when evaluating college pitchers and where they may end up in the draft.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dc5f3e3d477f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Stuff+ Model Revamp]]></title>
            <link>https://medium.com/@reinha3/stuff-model-revamp-d54f239cbbea?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/d54f239cbbea</guid>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[college-baseball]]></category>
            <category><![CDATA[baseball]]></category>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Fri, 15 Nov 2024 01:12:39 GMT</pubDate>
            <atom:updated>2024-11-15T01:12:39.151Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*VyAXQ05WdNl_22oN" /><figcaption>81–83 MPH fastball Stuff+ values by horizontal and vertical break.</figcaption></figure><p>Stuff+ is an attempt to quantify the predicted value of a pitch solely based on its physical characteristics, such as velocity and movement. Fangraphs has an excellent article exploring this concept more deeply (<a href="https://library.fangraphs.com/pitching/stuff-location-and-pitching-primer/">Link</a>). When we developed our original Stuff+ model, we had a massive limitation in our quantity of data. This meant we had to be creative with our model, which led us to calculate a “Stuff Constant” which served as an approximation of run values. The “Stuff Constant” was calculated by regressing various statistics, such as swinging strike % and groundball % against FIP, scaling the predicted results to have an average of 100. However, as our company has expanded, we have enough data to take a more traditional approach, where Stuff+ scores serve as a function of run values.</p><p>To create this model, I filtered our data to have pitches that resulted in a swing exclusively. This was to limit the influence of command on our model; while admittedly not a perfect approach, this was an easy way to create a model that mitigated the impacts of command as much as possible. Future models should explore different ways to continue limiting the impact of command while measuring Stuff. Following this, I used Baseball Savant to find the run values for each respective swing result: foul ball, swinging strike, groundball, pop-up, line drive, and fly ball. While MLB is a different run environment from college baseball, we decided that this would be easier than trying to calculate run values for Division III, especially with data limitations. I grouped fly balls with line drives but not with pop-ups; pop-ups are incredibly valuable for a pitcher, and grouping them with line drives and fly balls introduced significant bias into our model, even though a pitcher’s pop-up rate does not seem to be consistent YoY (<a href="https://blogs.fangraphs.com/is-popup-rate-a-skill/#:~:text=Don&#39;t%20count%20on%20it,a%20boatload%20of%20popups%2C%20rejoice">Fangraphs Article</a>). Pop-ups lead to an out virtually 100% of the time and are even more valuable than a swing and miss.</p><p>For the actual model, the first steps were scaling and doing basic feature selection. First, I took the handedness-dependent features in our model (horizontal release angle, spin axis, release side, and horizontal break) and took their absolute value. While this is not optimal, our dataset was not large enough to employ separate models based on handedness. After that, I removed any pitches that had a feature more than 3 standard deviations away from the mean. I did this to eliminate possible bias resulting from errors in the data. While this means players on the extremes (and thus the very highest and lowest Stuff+ values) would be excluded, it helped keep the model from overfitting to noise. After that, I employed an XGBoost model with all of the features Z-scored (velocity, vertical release angle, horizontal release angle, spin rate, spin axis, release height, release side, induced vertical break, and horizontal break). As opposed to our previous model, this one is much more true to the idea of a “Stuff” model. Finally, I scaled the model so that each pitch would have an average of 100 with a standard deviation of 10; this allows for easy conversion into the traditional 20–80 scouting scale. The overall Stuff+ score combines usage with the differences in the given run value of a pitch; for example, a 100 Stuff+ fastball is much less valuable than a 100 Stuff+ slider. Our model incorporates this by taking the difference in the mean run value and dividing it by the fastball mean run value. Thus, a fastball carries a weight of 1.0, and the other pitches are given more weight, with the weighted versions of the pitches being multiplied by the given usage, which can be set equally for each pitch to get a true Stuff+ value. This value is inflated, and so it was rescaled with the knowledge each pitch has an average of 100 with a standard deviation of 10. Using the mean of the pitches and the given variance, I calculated the overall mean and standard deviation to scale the final, overall Stuff+ to have an average of 100 with a standard deviation of 10.</p><p>To ensure the results matched what I anticipated, I created graphs to view the relationship between velocity and Stuff+. I also performed cross-validation to ensure that the results were not overfitting to noise. The graphs and model results are shown in the appendix. As for the results, they were strong and highly predictive of run value. While the mean cross-validation RMSE was around 11, this is still a strong result, as the range of outcomes in run values ranged from nearly -25 to 25. Since the true run values spanned across positive and negative ranges, this can make predicting the value difficult. The broad variability in actual values means an RMSE of around 11 is still fairly accurate. Additionally, the cross-validated RMSE standard deviations were all low, ranging from around 0.1 to 0.5. This signifies a model that has stable and consistent predictions.</p><p><strong>Appendix</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*NUDJvRU0Y4FYH_R8" /><figcaption>Fastball Velocity vs. Stuff+</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*6_Z1hdoS4abzanNw" /><figcaption>Slider Velocity vs. Stuff+</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*LaisGdpleuJFXA6n" /><figcaption>Curveball Velocity vs. Stuff+</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*tACtwjUu77QmnRWx" /><figcaption>Offspeed Velocity vs. Stuff+</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/754/1*M1N2LqwhHDB7eqGGTgmVFg.png" /><figcaption>Cross-validation RMSE by pitch</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/942/1*6SJF8rTiW_qh-mdimFRyMQ.png" /><figcaption>Mean predicted vs. actual run value</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*e0rR7plOvsrroJF0WhF93g.png" /><figcaption>Full model results.</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d54f239cbbea" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Glossary for the Pitching+ Server]]></title>
            <link>https://medium.com/@reinha3/glossary-for-the-pitching-server-22b09c270626?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/22b09c270626</guid>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Wed, 16 Oct 2024 23:41:29 GMT</pubDate>
            <atom:updated>2024-10-16T23:41:29.653Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XKPmV3P6X1fvG-pdB5LnBQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*jktZcdE8WciH6sos3gImHQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*o-XnEWTnoe0-V_044NoJNg.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*HzTBP_kzYCq0ok7Z3vvreQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*O3MWeKRCX1E3qGCc5YjyzQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/816/1*2T9-5z4gtFiwhEdndHG1Ig.jpeg" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=22b09c270626" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Applying Break Charting for Pitching Development: Introducing DDB and OAB]]></title>
            <link>https://medium.com/@reinha3/applying-break-charting-for-pitching-development-introducing-ddb-and-oab-fd6a61d619e7?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd6a61d619e7</guid>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Mon, 16 Sep 2024 23:13:03 GMT</pubDate>
            <atom:updated>2024-09-16T23:13:03.105Z</atom:updated>
            <content:encoded><![CDATA[<p>by Henry Gliedman</p><h3>Introducing Break Charts</h3><p>Break charts are standard visuals used for analysis of pitching data. Many baseball technologies, including our Pitching + Server, utilize these types of charts to illustrate data for users. Beyond just looking at these charts, we can apply analytical techniques to transform these charts into great development tools. This paper introduces two analysis techniques– Distance Between Breaks (DBB), and Opposed Angle of Breaks (OAB)- which can enhance understanding of pitch shape, and how shaping can be used to refine a quality arsenal of pitches.</p><p>For those unfamiliar with break charts, they display the average vertical and horizontal breaks of each pitch in a player’s arsenal. Vertical breaks are typically displayed in terms of induced vertical break, the spin induced movement when gravity is accounted for. These breaks are generated based on the spin that a player applies, and the angle at which the spin is applied.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*Qawo5zn-I8o-m4Ke" /></figure><p>Figure 1. A typical break as displayed in the Pitching + Server.</p><h3>What are we looking for?</h3><p>Visually, we assess several aspects of the break chart to determine the quality of a player’s arsenal and identify areas for improvement. We begin with the Fastball.</p><p>For the Fastball break charts serve as an evaluation tool rather than a development guide. Typically, Fastball breaks are more strictly influenced by arm slots, and arm actions as a whole. While it is possible to quickly make small tweaks to break profiles. Significantly reshaping a Fastball shape requires overhauling a player’s pitching motion, leading to potential issues with command, or injury proneness. Additionally adjustments of that scale require immense time and practice. Especially below the Division 1 college level, for a lot of cases it makes more sense to focus on adding velocity without worrying much about shape. Below is a chart from a large database of D3 pitchers showing average whiff rates by break shape.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*Xvr529cBRWjgJxx0" /></figure><p>Figure 2. Whiff rates for Fastball shapes for large amounts of D3, indicating that other factors may have more impact on performance.</p><p>This chart is not to dissuade from the importance of breaks amongst pitches, creating more break in both directions has been shown to lead to more whiffs at all levels of college and professional play. However, based on this chart and other analysis at most levels it seems to be more productive to focus on other Fastball traits such as velocity and raw RPM’s instead of spending months making large scale shape refinements, at least until the pitcher reaches a high velocity threshold. However, if one wants to refine shape, the goal is to keep vertical break away from horizontal break. A Fastball profile that is either very vertically dominant or very horizontally dominant. Commonly this is referred to as keeping your Fastball out of the “Dead Zone,” where vertical and horizontal breaks conflict to create a flatter appearing pitch.</p><p>Most secondary pitch shaping work should be done based off of a player’s Fastball movement profile. While specific shapes of breaking balls or changeups can play better than others, their success is often based on being able to work off a pitcher’s fastball.</p><p>Since the birth of Art of W.A.R, we have held one belief as the foundation of all pitching development: you can’t improve something efficiently if you don’t measure it. Thus, quantification is the first step to development. Currently our Stuff + modeling looks at each pitch in a vacuum, but does not consider the value of each pitch when combined with others. We aim to introduce a series of analytics which can measure and evaluate how effectively a pitcher is able to mesh their arsenal together.</p><h3>Introducing Distance Between Breaks (DBB)</h3><p>One method for quantitatively analyzing break charts is to measure the distance between the pitcher’s primary pitch and each secondary pitch. That Distance Between Breaks (DBB) indicates how effective a pitcher is at differentiating their pitches in break. For breaking balls, it is generally understood that you want to shape the pitch in the opposite direction of a pitcher’s primary pitch. The better a player’s raw distance is on the break charts in theory the better they do at tunneling these pitches, and creating swing and miss stuff.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*tTUV8kbKiQLFg50F" /></figure><p>Figure 3. A break chart visualizing the idea of Distances Between Breaks.</p><p>This Distance Between Breaks (DBB) metric can be used as a simple way to quantify a pitcher’s ability to create break separation and tunnel their pitches. While it is not the one metric which leads to more swing and misses, it is a peripheral that can aid with refining arsenals. Here is a chart of the relationship of this DBB metric graphed against Whiff rates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*hxsj6hjQ97AVim_H" /></figure><p>Figure 4. The relationships between the Distance Between Breaks metric and Whiff rates for breaking balls. (Note that alone these regressions are not statistically significant.)</p><p>Although this chart does not control for Stuff + or general pitch quality, which would give a more detailed analysis, the general slope appears to be positively correlated. When applied as a tool for pitch shaping, we would hope it aids a player towards throwing effective breaking balls. While it is not a metric to directly chase in training, it can operate as a good peripheral to contextualize what development is needed.</p><h3>Introducing Opposed Angle of Breaks</h3><p>Beyond just the raw difference in break, the angle at which pitches diverge– the Opposed Angle of Break (OAB)- is another metric that can aid in pitch shape development. By measuring the angle between the fastball and each secondary pitch relative to the origin on the break chart, we can assess how directly a pitch tunnels away from the primary pitch. Here is what that would look like on the break chart.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*SXpJzYtX-RJbZzu6" /></figure><p>Figure 5. A break chart displaying the Opposed Angle of Breaks (OAB) metric, with red lines representing the movement vectors of secondary pitches from the origin.</p><p>The angle of OAB exists in between the red and black lines. This can indicate how directly a pitch tunnels away from the primary pitch. Now this is not to say each pitch would be optimized with an angle of 0. This metric should be further studied, and in time be used to understand more about pitch shaping. A quantification tool of tunneling, showing which angles fare the best for breaking balls of particular movement profiles, or how one should approach shaping a changeup. Here is a preliminary chart showing the linear relationships between OAB and Whiff Rates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/875/0*0TDxGL1UTzTVXO_y" /></figure><p>Figure 6. The relationship between OAB and whiff rates for commonly thrown secondary pitches, color-coded by velocity groupings.</p><p>This chart shows that there is a potential relationship here between generating swings and misses, and having lower values of OAB for each pitch. I would propose these relationships are a lot more complicated than a linear model could show. Additionally, again after normalizing for pitch quality, or Stuff we would hope to see a more clear relation.</p><p>We hope this article is useful to players, coaches, and baseball enthusiasts. Our goal with this is to encourage deeper consideration of what break charts represent, and to strengthen the connection between publicly accessible baseball analytics and player development. Employing simple quantification like DBB or OAB for quality development. While these metrics are no miracle solution in pitch shaping development, hopefully it is a step towards better data driven development. At AoW, we aim to develop a comprehensive model for pitch interaction. Potentially called Arsenal Points, which rewards pitchers for effectively tunneling their pitches, thereby affecting their AoW Stuff +. We hope to eventually introduce these metrics along with further analysis in pitch to pitch interaction into our server throughout the near future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd6a61d619e7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[College Hitter Data-Dive: Part 3]]></title>
            <link>https://medium.com/@reinha3/college-hitter-data-dive-part-2-3bd6f84132a7?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/3bd6f84132a7</guid>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Sun, 14 Jul 2024 14:01:58 GMT</pubDate>
            <atom:updated>2024-07-14T14:02:23.304Z</atom:updated>
            <content:encoded><![CDATA[<p>by Benjamin Reinhard</p><p>As a result of time constraints and limits to data, these writeups are focused on hit tools. I unfortunately do not have easy access to exit velocity data, which is an incredibly important aspect in evaluating hitters. Focusing on hit tool data, such as plate discipline numbers, can help identify hitters with more risk in their profile. Much of the included data that may show risk is dependent on an organization’s player development philosophy and abilities. I will reference swing and miss numbers, chase data, and percentage of batted balls that are pulled in the air. For example, a low percentage of batted balls being pulled in the air potentially signals the ability for a player to improve their power numbers, contingent on an organization helping a player pull the ball in the air more without hindering other abilities. Fangraphs has a ton of excellent articles on pulled fly ball percentage, such as here: <a href="https://blogs.fangraphs.com/the-pulled-fly-ball-revolution-was-always-underway/">Link</a>. Similarly, if an organization is confident in their ability to help a player improve their chase and/or miss numbers, they may not be as concerned. Shown below are the average numbers for the 77 college hitters ranked by MLB Pipeline in their draft rankings. The respective numbers for the hitter covered in this part is included at the end of the article.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gVxaAEYidtXosFqfTmjiHQ.png" /></figure><p><strong>Duce Gourson (21); UCLA</strong></p><p>2B, LHH</p><p>Pipeline Rank: 158</p><p>Elevated breaking ball whiff (37%) and offspeed whiff (35%) raise hit tool questions.</p><p><strong>Khadim Diaw (20); Loyola Marymount</strong></p><p>C/OF, RHH</p><p>Pipeline Rank: 163</p><p>Solid bat-to-ball abilities, but elevated chase, especially on breaking balls, offspeed, and with 2 strikes (36%, 36%, 52%).</p><p><strong>Payton Green (21); Georgia Tech</strong></p><p>SS, RHH</p><p>Pipeline Rank: 169</p><p>Relatively stable profile with no glaring data-based hit tool concerns. Strong pulled balls in the air percentage (23.8%) should enable him to get to power.</p><p><strong>Corey Collins (22); Georgia</strong></p><p>1B, LHH</p><p>Pipeline Rank: 170</p><p>Strong contact abilities on fastballs (9% whiff rate). Pulls the ball in the air at an extreme rate (36.8%).</p><p><strong>Casey Cook (21); North Carolina</strong></p><p>OF, LHH</p><p>Pipeline Rank: 171</p><p>Relatively strong performance coupled with solid across the board whiff and chase numbers.</p><p><strong>Josh Caron (20); Nebraska</strong></p><p>C, RHH</p><p>Pipeline Rank: 172</p><p>Below average whiff rates on breaking balls and offspeed pitches (32%, 38%).</p><p><strong>Emilien Pitre (21); Kentucky</strong></p><p>2B, LHH</p><p>Pipeline Rank: 173</p><p>Strong fastball and breaking ball whiff rates (16%, 21%). Improving his 15.6% pulled air rate may help unlock more power.</p><p><strong>Hunter Hines (21); Mississippi St.</strong></p><p>1B, LHH</p><p>Pipeline Rank: 174</p><p>Struggles mightily with breaking balls and offspeed contact (45%, 50% whiff rates, respectively). Strong pulled air rate (29.0%) which may help mitigate contact issues by getting to all of his power.</p><p><strong>Ryan Stafford (21); Cal Poly</strong></p><p>C, RHH</p><p>Pipeline Rank: 175</p><p>Solid-average profile across the board. Below-average pulled air rate (14.3%).</p><p><strong>Ryan Campos (21); Arizona St.</strong></p><p>C, RHH</p><p>Pipeline Rank: 176</p><p>Relatively average miss and chase rates across the board with a miniscule pulled in the air rate (7.7%).</p><p><strong>Jonathan Vastine (21); Vanderbilt</strong></p><p>SS, LHH</p><p>Pipeline Rank: 188</p><p>Swing-and-miss heavy on breaking balls and offspeed pitches (48%, 40%).</p><p><strong>Derek Bender (21); Coastal Carolina</strong></p><p>1B/C, RHH</p><p>Pipeline Rank: 189</p><p>Slightly below-average chase rates as well as a below-average pulled air rate (13.9%).</p><p><strong>Ivan Brethowr (21); UC Santa Barbara</strong></p><p>OF, RHH</p><p>Pipeline Rank: 194</p><p>Whiffed on breaking balls and offspeed pitches at a high rate (31%, 47%).</p><p><strong>Dub Gleed (21); UC Irvine</strong></p><p>3B, RHH</p><p>Pipeline Rank: 198</p><p>Chased on 43% of 2 strike counts. Solid-average whiff and chase rates.</p><p><strong>Jeremiah Jenkins (21); Maine</strong></p><p>1B, LHH</p><p>Pipeline Rank: 199</p><p>Struggled with breaking balls (34% whiff rate).</p><p><strong>Sam Antonacci (21); Coastal Carolina</strong></p><p>3B, LHH</p><p>Pipeline Rank: 201</p><p>Very strong contact and chase rates across the board. An improvement of his 12.2% pulled air rate may improve his power numbers.</p><p><strong>Will Taylor (21); Clemson</strong></p><p>OF, RHH</p><p>Pipeline Rank: 203</p><p>Swings and misses at offspeed pitches at a high rate (45%).</p><p><strong>Carter Mathison (21); Indiana</strong></p><p>OF, LHH</p><p>Pipeline Rank: 204</p><p>Well-rounded profile with no glaring flaws.</p><p><strong>Sam Petersen (21); Iowa</strong></p><p>OF, RHH</p><p>Pipeline Rank: 205</p><p>Struggles with offspeed pitches (38% whiff).</p><p><strong>Sean Keys (21); Bucknell</strong></p><p>3B, LHH</p><p>Pipeline Rank: 211</p><p>Strong contact abilities with a 15.0% pulled air rate, leaving room for power improvement.</p><p><strong>Jakob Christian (21); San Diego</strong></p><p>OF, RHH</p><p>Pipeline Rank: 216</p><p>Elevated whiff rates, especially on breaking balls (42%) and offspeed pitches (43%).</p><p><strong>Zach Erhard (21); Oklahoma St.</strong></p><p>OF, RHH</p><p>Pipeline Rank: 221</p><p>Strong contact abilities with a pulled air rate of 24.3%.</p><p><strong>Eli Serrano (21); North Carolina St.</strong></p><p>OF, LHH</p><p>Pipeline Rank: 224</p><p>Above-average whiff rates with a below-average 2 strike chase rate (46%).</p><p><strong>Jacob Jenkins-Cowart (21); East Carolina</strong></p><p>OF, LHH</p><p>Pipeline Rank: 228</p><p>High chase rates, particularly on offspeed pitches (46%) and with 2 strikes (46%). Low pulled air rate (10.2%).</p><p><strong>Bobby Boser (21); South Florida</strong></p><p>2B/SS, RHH</p><p>Pipeline Rank: 235</p><p>Struggles heavily with offspeed pitches, whiffing on 46% of them. Pulls the ball in the air well (20.5%).</p><p><strong>Ali Camarillo (21); Texas A&amp;M</strong></p><p>SS, RHH</p><p>Pipeline Rank: 246</p><p>Limited in-game power performance with solid bat-to-ball skills, some elevated chase rates, such as on offspeed (42%).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gZ7aDcvOq1u2GybX" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3bd6f84132a7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[College Hitter Data-Dive: Part 2]]></title>
            <link>https://medium.com/@reinha3/college-hitter-data-dive-part-2-6ed69b56db65?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/6ed69b56db65</guid>
            <category><![CDATA[mlb-draft]]></category>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[college-baseball]]></category>
            <category><![CDATA[analytics]]></category>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Sat, 13 Jul 2024 14:07:22 GMT</pubDate>
            <atom:updated>2024-07-13T14:07:22.391Z</atom:updated>
            <content:encoded><![CDATA[<p>by Benjamin Reinhard</p><p>As a result of time constraints and limits to data, these writeups are focused on hit tools. I unfortunately do not have easy access to exit velocity data, which is an incredibly important aspect in evaluating hitters. Focusing on hit tool data, such as plate discipline numbers, can help identify hitters with more risk in their profile. Much of the included data that may show risk is dependent on an organization’s player development philosophy and abilities. I will reference swing and miss numbers, chase data, and percentage of batted balls that are pulled in the air. For example, a low percentage of batted balls being pulled in the air potentially signals the ability for a player to improve their power numbers, contingent on an organization helping a player pull the ball in the air more without hindering other abilities. Fangraphs has a ton of excellent articles on pulled fly ball percentage, such as here: <a href="https://blogs.fangraphs.com/the-pulled-fly-ball-revolution-was-always-underway/">Link</a>. Similarly, if an organization is confident in their ability to help a player improve their chase and/or miss numbers, they may not be as concerned. Shown below are the average numbers for the 77 college hitters ranked by MLB Pipeline in their draft rankings. The respective numbers for the hitter covered in this part is included at the end of the article.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8wr4WIbtRvOtdPHc_J3obA.png" /></figure><p><strong>Blake Burke (21); Tennessee</strong></p><p>1B, LHH</p><p>Pipeline Rank: 65</p><p>Has a lot of chase (overall 35%, fastball 37%, breaking ball 34%, 2 strikes 51%). Strong in-game performance, although a slightly lower walk rate than optimal (10.4%).</p><p><strong>Kavares Tears (21); Tennessee</strong></p><p>OF, LHH</p><p>Pipeline Rank: 66</p><p>Tears has quite a bit of hit tool risk. He struck out in 24.7% of plate appearances this season, an extremely high rate. Accompanying this, his overall whiff rate was 30%, with a breaking ball whiff rate of 30%. He pulls the ball in the air well, providing optimism he will be able to reach his power and mitigate hit tool concerns.</p><p><strong>Kyle DeBarge (20); Louisiana-Lafayette</strong></p><p>SS, RHH</p><p>Pipeline Rank: 67</p><p>DeBarge is an aggressive hitter, with a low walk rate (6.0%). This also is shown through his chase rates, especially with a 47% 2 strike chase rate.</p><p><strong>Gage Miller (21); Alabama</strong></p><p>3B, RHH</p><p>Pipeline Rank: 68</p><p>Miller is one of my favorite profiles in this tier of the draft. His low walk rate (9.4%) was accompanied by a miniscule strikeout rate (9.1%). He does not have any concerning whiff rates (breaking ball miss % = 14). He does tend to chase with 2 strikes (45%), but provides a relatively safe profile at this point in the draft.</p><p><strong>Dylan Dreiling (21); Tennessee</strong></p><p>OF, LHH</p><p>Pipeline Rank: 72</p><p>Dreiling has a well-rounded profile with no major concerns. He does K a bit more than in preferable (20.5%), at least some of which can be attributed to his below-average 2 strike chase rate of 38%.</p><p><strong>Peyton Stovall (21); Arkansas</strong></p><p>2B, LHH</p><p>Pipeline Rank: 77</p><p>Stovall had limited power production this season, posting a .528 SLG and .191 ISO. His miss and chase rates range from average to above-average. He pulls the ball in the air 17.4% of the time, so an adjustment there may help unlock more power.</p><p><strong>Ethan Anderson (20); Virginia</strong></p><p>C/1B, SH</p><p>Pipeline Rank: 78</p><p>Anderson’s main questions are regarding his power, with a .514 SLG and .179 ISO. He had strong contact numbers, which, combined with his 16.6% pulled air rate, leads me to believe even if he were to exchange some contact for power, he can still be very successful.</p><p>Cole Mathis (20); College of Charleston</p><p>1B, RHH</p><p>Pipeline Rank: 84</p><p>Mathis has a really well-rounded profile, with no major issues standing out. If one were to nitpick, as a 1B he needs to rake, so his .315 ISO is more good than stellar. At this tier, Mathis is one of the safer bats in my opinion.</p><p><strong>Josh Kuroda-Grauer (21); Rutgers</strong></p><p>SS, RHH</p><p>Pipeline Rank: 90</p><p>Kuroda-Grauer has excellent bat-to-ball skills, running a miniscule 11% miss rate, with his breaking ball and offspeed numbers also being strong. He has a tendency to chase, especially shown with his 48% 2 strike chase rate. He rarely pulls the ball in the air (9.4%), so an improvement there may help unlock more power.</p><p><strong>Carson DeMartini (21); Virginia Tech</strong></p><p>3B, LHH</p><p>Pipeline Rank: 93</p><p>DeMartini has a lot of power (.707 SLG, .378 ISO) with a lot of swing and miss (36% on fastballs and breaking balls). He pulls the ball in the air at an extremely high clip (34.8%) which gives me hope he will get to his power and balance out swing-and-miss concerns.</p><p><strong>Cole Messina (21); South Carolina</strong></p><p>C, RHH</p><p>Pipeline Rank: 100</p><p>Messina ran high swing and miss numbers, particularly on fastballs and offspeed pitches. He had strong performance numbers.</p><p><strong>Jalin Flores (20); Texas</strong></p><p>SS, RHH</p><p>Pipeline Rank: 116</p><p>Flores had high chase rates, with an overall 36% chase. His 2 strike chase rate was 48%, but he was still a strong statistical performer despite these flaws.</p><p><strong>Jared Jones (20); LSU</strong></p><p>1B, RHH</p><p>Pipeline Rank: 119</p><p>Jones’ game features a lot of swing and miss, with a 28% miss rate on fastballs and 44% on breaking balls. Those numbers create concern about his ability to make enough contact to be a productive player. He had excellent power numbers, and will need to hit for a lot of power to make up for hit tool issues, especially at first base.</p><p><strong>Harrison Didawick (21); Virginia</strong></p><p>OF, LHH</p><p>Pipeline Rank: 125</p><p>Didawick has a solid albeit unexciting profile, especially for this stage in the draft. He struck out a lot (24.7%), largely as a result of miss rates (fastball: 22%, breaking ball: 38%, offspeed: 49%). If he can improve some of those issues, he can be a solid outfield option, potentially as a platoon player or fourth outfielder.</p><p><strong>John Spikerman (21); Oklahoma</strong></p><p>OF, SH</p><p>Pipeline Rank: 126</p><p>Spikerman did not hit for a ton of power at Oklahoma, but this is assisted by not having much concern with regards to his miss or chase rates. His pulled air rate was only 13.8%, so an improvement there could potentially unlock more power.</p><p><strong>Casey Saucke (20); Virginia</strong></p><p>OF, RHH</p><p>Pipeline Rank: 127</p><p>Saucke was a strong statistical performer (.349/.412/,587). He had a good fastball whiff rate (15%) but a below-average breaking ball whiff (35%). Improvement there will lift his ceiling immensely.</p><p><strong>Nick McLain (21); Arizona St.</strong></p><p>OF, SH</p><p>Pipeline Rank: 129</p><p>McClain has a really well-rounded profile, not especially excelling in any one area but also with no glaring problems. His fastball whiff rate of 23% is high, but his breaking ball (12%) and offspeed (12%) provide hope that he can improve his fastball whiff rate.</p><p><strong>Woody Hadeen (21); UC Irvine</strong></p><p>SS, SH</p><p>Pipeline Rank: 130</p><p>Hadeen has a strong profile for this point in the draft. While he did not hit for a ton of power (.419 SLG, .061 ISO) he offset this with strong plate discipline and bat-to-ball skills. He only swung and missed on 4% of fastballs and 22% of breaking balls. His 2 strike chase rate of 22% was excellent as well. He only pulled the ball in the air 11.1% of the time.</p><p><strong>Austin Overn (21); Southern California</strong></p><p>OF, LHH</p><p>Pipeline Rank: 131</p><p>Overn really struggled across the board with whiff and chase rates, only posting a better than average mark in breaking ball whiff (17%) and offspeed whiff (18%). This may signify the potential for him to improve in the other areas.</p><p><strong>Rodney Green (21); California</strong></p><p>OF, LHH</p><p>Pipeline Rank: 132</p><p>Green has a similar profile to Vance Honeycutt and Dakota Jordan, although with less performance and slightly dimmed tools. He struck out in nearly 30% of his plate appearances (28.5%). While his fastball whiff rate of 23% is not awful, his breaking ball whiff rate of 48% is highly concerning, especially as he sees good pitching consistently. He doesn’t chase much on 2 strikes (26%), which could signify the potential for discipline improvements. He only pulls the ball in the air 12.6% of the time, something he will almost assuredly need to improve to hit for enough power to be a potential MLB player.</p><p><strong>Colby Shelton (21); Florida</strong></p><p>SS, LHH</p><p>Pipeline Rank: 133</p><p>Shelton has a risky hit tool profile, not unlike Green. He swung and missed at 30% of pitches, but perhaps more concerning is his 36% whiff rate on breaking balls. He tends to expand with 2 strikes, running a chase rate of 47%.</p><p><strong>Zack Stewart (20); Missouri St.</strong></p><p>OF, LHH</p><p>Pipeline Rank: 134</p><p>While Stewart is only 20, his performance at Missouri St. gives me pause. He struck out in 24.3% of plate appearances and only had a .376 OBP. Additionally, he whiffed on 42% of breaking balls, raising concerns about viability against major league quality breaking pitches. He only pulled the ball in the air 16.4% of the time, and maximizing that may be key to getting to his power, particularly if he has a below-average hit tool.</p><p><strong>Anthony Silva (20); TCU</strong></p><p>SS, RHH</p><p>Pipeline Rank: 137</p><p>Silva has a risky profile with limited performance (.268/.369/.384). He did not have large swing and miss or chase problems outside of a 34% whiff rate on offspeed pitches.</p><p><strong>Cody Schrier (21); UCLA</strong></p><p>SS, RHH</p><p>Pipeline Rank: 138</p><p>Similarly to Silva, Schrier has a risky profile, starting with his limited performance (.250/.323/.366). He swung and missed at 36% of breaking balls and 24% of fastballs, bringing questions about his viability as a bat. He also ran a 50% 2 strike chase rate, concerning me that his limited power and swing and miss problems are not assisted by strong discipline.</p><p><strong>Robert Hipwell (21); Santa Clara</strong></p><p>3B, LHH</p><p>Pipeline Rank: 139</p><p>Hipwell did not have a large sample this year, so his data also includes last year. He ran a below-average 21.9% strikeout rate, but offset that with a 22.7% walk rate. He has elevated swing and miss, with a 38% miss rate on offspeed pitches standing out.</p><p><strong>Zander Darby (21); UC Santa Barbara</strong></p><p>1B/3B, LHH</p><p>Pipeline Rank: 151</p><p>Darby was a below average performer at UC Santa Barbara, putting up a .339 OBP and .415 SLG this season. He really struggled with offspeed pitches, missing on 34% with a 33% chase rate. Improving his 17.0% pulled air rate potentially will help his power numbers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*lT0yjW5-HpPwaIUy" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6ed69b56db65" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[College Hit Tool Data-Dive: Part 1]]></title>
            <link>https://medium.com/@reinha3/college-hit-tool-data-dive-part-1-87b6124897f1?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/87b6124897f1</guid>
            <category><![CDATA[mlb-draft]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[college-baseball]]></category>
            <category><![CDATA[sabermetrics]]></category>
            <category><![CDATA[baseball]]></category>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Fri, 12 Jul 2024 23:18:08 GMT</pubDate>
            <atom:updated>2024-07-12T23:18:54.912Z</atom:updated>
            <content:encoded><![CDATA[<p>By Benjamin Reinhard</p><p>As a result of time constraints and limits to data, this write up (and the following 2) will be focused on hit tools. I unfortunately do not have easy access to exit velocity data, which is an incredibly important aspect in evaluating hitters. Focusing on hit tool data, such as plate discipline numbers, can help identify hitters with more risk in their profile. Much of the included data that may show risk is dependent on an organization’s player development philosophy and abilities. I will reference swing and miss numbers, chase data, and percentage of batted balls that are pulled in the air. For example, a low percentage of batted balls being pulled in the air potentially signals the ability for a player to improve their power numbers, contingent on an organization helping a player pull the ball in the air more without hindering other abilities. Fangraphs has a ton of excellent articles on pulled fly ball percentage, such as here: <a href="https://blogs.fangraphs.com/the-pulled-fly-ball-revolution-was-always-underway/">Link</a>. Similarly, if an organization is confident in their ability to help a player improve their chase and/or miss numbers, they may not be as concerned. Shown below are the average numbers for the 77 college hitters ranked by MLB Pipeline in their draft rankings. The respective numbers for the hitter covered in this part is included at the end of the article.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KnvXGMGAif0q_6K7oTJrgg.png" /></figure><p><strong>Travis Bazzana (21), Oregon St</strong></p><p>2B; LHH</p><p>Pipeline Rank: 1</p><p>Bazzana excels in virtually every area. He is above average in every single statistic listed above. If there is any area to nitpick, he chases with 2 strikes on 34% of pitches, and his fastball whiff rate is 14%. He seems to be a sure bet to hit.</p><p><strong>Charlie Condon (21), Georgia</strong></p><p>OF/3B; RHH</p><p>Pipeline Rank: 2</p><p>Condon has ridiculous power. This year, he hit 37 home runs and has a slugging % over 1000. However, there are some questions that remain to be answered on the hit tool. Specifically, he whiffed on 31% of breaking balls, which concerns me that he will be exposed in pro ball as he faces higher quality breaking pitches more frequently. In the same vein, he whiffed on 32% of offspeed pitches. The rest of his profile does not have real hit tool concerns, but those two areas do raise some questions.</p><p><strong>Jac Caglianone (21), Florida</strong></p><p>1B; LHH</p><p>Pipeline Rank: 3</p><p>Caglianone is probably the most famous player in this class, a two-way star who hits prodigious home runs and has thrown in the upper 90s off the mound. He has a very peculiar profile; his whiff rates range from solid to good, but his chase rates raise a lot of questions. His overall and fastball chase rates are 39%, with a 38% chase rate on breaking balls, 42% chase on offspeed, and 45% chase with 2 strikes. He presents an incredibly interesting developmental case; if a team believes his chase issues are largely fixable, he has the potential to be the best bat in this class.</p><p><strong>JJ Wetherholt (21), West Virginia</strong></p><p>SS/2B; LHH</p><p>Pipeline Rank: 4</p><p>Wetherholt and Bazzana have similar profiles, although it seems as if Wetherholt has a better chance to stay at shortstop. Similarly to Bazzana, Wetherholt has great to solid chase and whiff rates, but his in-game power production leaves much to be desired, with a .258 ISO this past season. However, Wetherholt ran a far below-average pulled-air %, with only 10.8% of his batted balls being pulled in the air. If a team believes they can help him pull the ball in the air more frequently and/or can stay at shortstop, he and Bazzana are virtually equivalent.</p><p><strong>Nick Kurtz (21), Wake Forest</strong></p><p>1B; LHH</p><p>Pipeline Rank: 7</p><p>Kurtz has a strong, well-rounded profile. He has one of the best eyes at the plate, with low chase rates on all pitches and with 2 strikes, which is a good measure of a player’s actual eye. He does have a slightly concerning 33% whiff rate on breaking balls, which is below average, but not significantly. As long as that does not become a hindrance, Kurtz has a safe profile, which is important for seemingly a 1B only prospect.</p><p><strong>Braden Montgomery (21); Texas A&amp;M</strong></p><p>OF, SH</p><p>Pipeline Rank: 8</p><p>Reading about Montgomery, his batted ball data sounds like it is excellent, which should help mitigate some of the slight concerns about his hit tool. His overall whiff rate of 26% is high, although not extremely. Similarly, his breaking ball whiff rate of 29% is below average, but not majorly. His offspeed whiff is 38%, another high mark, but his chase numbers are pretty solid.</p><p><strong>James Tibbs (21); Florida St.</strong></p><p>OF, LHH</p><p>Pipeline Rank: 12</p><p>Tibbs has a relatively safe profile, data-wise, among players in the second tier of college players. While his whiff and chase rates trend more solid average than plus, his only real concern is a breaking ball whiff rate of 28%, only 1% below the average,</p><p><strong>Christian Moore (21); Tennessee</strong></p><p>2B, RHH</p><p>Pipeline Rank: 13</p><p>Moore has shot up draft boards on the back of strong batted ball data and great performance throughout the NCAA tournament. His one real concern is a fastball whiff rate of 21%. He could potentially improve his pulled air percentage, which at 14.8% is below average. None of his whiff or chase rates stand out as especially strong, but perhaps more importantly, none stand out as especially poor.</p><p><strong>Cam Smith (21); Florida St.</strong></p><p>3B, RHH</p><p>Pipeline Rank: 14</p><p>Smith has a strong, relatively low-risk profile. Virtually all of his whiff and chase data are at least average to above-average. He didn’t have a super high ISO (.269), which can likely be improved with an improvement in pulled air rate (13.6%).</p><p><strong>Seaver King (21); Wake Forest</strong></p><p>UTL, RHH</p><p>Pipeline Rank: 17</p><p>King has an 80 grade baseball name, and with only 1 year at D1 after 2 years at D2, it isn’t far-fetched to think some of the concerns in his profile will be alleviated with experience. His whiff rates are not very concerning, but he does run high chase rates in nearly every area, from fastballs, to breaking balls, and especially with two strikes (33, 39, 53). If I had to guess, I think organizations will be less concerned with elevated chase rates than with whiff rates.</p><p><strong>Carson Benge (21); Oklahoma St.</strong></p><p>OF, LHH</p><p>Pipeline Rank: 18</p><p>Benge is another well-rounded college performer in this class. He had really strong overall and fastball whiff rates (16, 13). He does tend to chase breaking balls (31%), but other than that, his plate discipline numbers are strong. His pulled air rate is only 10.5%, so an improvement there offers potential for his power to really show.</p><p><strong>Tommy White (21); LSU</strong></p><p>CIF, RHH</p><p>Pipeline Rank: 20</p><p>Tommy Tanks is another famous player in this class, well-known after raking at NC St. as a freshman and then going to LSU and winning a national championship with Skenes and co. While he may seem to be a safe bat, there are some concerns. Namely, his walk rate of 7.1% shows his overall aggressiveness, accompanied by his overall 34% chase rate. Throughout his profile, he shows a tendency to chase, although his whiff rates are very solid. To maximize his power, he will probably need to improve his 15.1% pulled air rate.</p><p><strong>Vance Honeycutt (21); North Carolina</strong></p><p>OF, RHH</p><p>Pipeline Rank: 22</p><p>Honeycutt has a high ceiling with massive hit tool concerns, reminding me of Jeren Kendall from a few years ago. Honeycutt struck out in 27.9% of his plate appearances this season, and he also whiffed on 32% of swings. On breaking balls, he put up a 40% whiff rate. He pulls the ball in the air well, but with his massive hit tool concerns, he will need to get to every last ounce of his power.</p><p><strong>Ryan Waldschmidt (21); Kentucky</strong></p><p>OF, RHH</p><p>Pipeline Rank: 23</p><p>Waldschmidt is another solid college outfield bat in this class. His biggest concern is his 31% whiff rate on breaking balls, which is elevated, but not egregiously so. Otherwise, his overall profile is very solid, feeling relatively safe, albeit somewhat unexciting.</p><p><strong>Walker Janek (21); Sam Houston</strong></p><p>C, RHH</p><p>Pipeline Rank: 24</p><p>Janek’s largest concerns are about his ability to handle quality breaking balls. He whiffed on 32% of breaking balls and chased 31%, while playing outside of a top-tier conference. He didn’t swing through fastballs much (16%), so there is hope for his bat. He only pulled the ball in the air 9.0% of the time, leading me to believe there is a lot more power potential than he showed.</p><p><strong>Malcolm Moore (20); Stanford</strong></p><p>C, LHH</p><p>Pipeline Rank: 26</p><p>Moore has probably my favorite profile of the catchers in this tier. While his average was quite low (.255), the underlying data points to that being mostly a fluke. His fastball whiff rate of 7% was exceptional, and his breaking ball whiff rate was 25%, a mark slightly above average. His 2 strike chase rate was 41%, slightly elevated, but not a massive concern.</p><p><strong>Kaelen Culpepper (21); Kansas St.</strong></p><p>SS, RHH</p><p>Pipeline Rank: 31</p><p>Culpepper had quite a bit of struggles with breaking balls, whiffing on 29% of them, while chasing 34% of them. He only swung and missed at 12% of fastballs, so the hope is that his bat control is not the issue. He did chase on 45% of 2 strike counts, cause for at least some concern.</p><p><strong>Billy Amick (21); Tennessee</strong></p><p>3B, RHH</p><p>Pipeline Rank: 32</p><p>Amick has a pretty risky profile in my opinion. His fastball whiff rate (24%) and breaking ball whiff rate (33%) were both below average, giving me hit tool concerns. His 2 strike chase rate of 39% was relatively average, exemplifying a lot of his profile. He excelled at pulling the ball in the air (24.6%), so I do worry about coaxing more power out of his bat, although his power numbers were more than fine.</p><p><strong>Caleb Lomavita (21); California</strong></p><p>C, RHH</p><p>Pipeline Rank: 33</p><p>Lomavita had a bit of an up and down season at Cal, posting only a 4.1% walk rate, with solid-average OBP and SLG numbers. His overall chase rate of 37% was a bit high, especially shown with a 43% breaking ball chase and a 49% 2 strike chase rate. His whiff numbers were relatively average.</p><p><strong>Dakota Jordan (21); Mississippi St.</strong></p><p>OF, RHH</p><p>Pipeline Rank: 34</p><p>Dakota Jordan has been a really popular player as a ceiling choice, with supporters crediting his batted ball data and his limited time focusing on baseball. Even with that, I find his profile to be exceptionally risky for a potential first round pick. His overall whiff rate is 35%, with a fastball whiff of 33% and a breaking ball whiff rate of 37%. He also has a 42% chase rate on breaking balls, leading me to seriously question the chances he reaches his ceiling. He does not pull the ball in the air well (11.4%), but that is a secondary problem to the whiff rates.</p><p><strong>Griff O’Ferrall (21); Virginia</strong></p><p>SS, RHH</p><p>Pipeline Rank: 38</p><p>O’Ferrall has a really interesting profile. He is a very aggressive hitter, with a really high chase rate (35%). His walk rate is low at 6.6%, but he only strikes out 7.5% of the time. Likely closely related here, his ISO was only .132. He pulls the ball in the air at a roughly average rate, but my understanding is that he does not hit the ball especially hard, although it seems as if he should stick at shortstop, if not, up the middle. His whiff rates are excellent, with a fastball whiff of only 7% and a breaking ball whiff rate of 20%. He has high chase rates, especially shown through his two strike chase of 61%.</p><p><strong>Jacob Cozart (21); North Carolina St.</strong></p><p>C, LHH</p><p>Pipeline Rank: 42</p><p>Cozart is another in a long list of college catchers in a similar tier, along with Bazzell, Lomavita, Janek, and Moore. Of this group, Cozart has a relatively concerning profile, as he has high whiff rates across the board. His breaking ball whiff was 41%, giving me concern about how he will handle consistently good breaking pitches. While he had a solid walk rate of 16.7%, his 2 strike chase rate of 47% leads me to believe that is elevated above his true ability.</p><p><strong>Mike Sirota (21); Northeastern</strong></p><p>OF, RHH</p><p>Pipeline Rank: 50</p><p>Coming into this season, Sirota was discussed as a potential early selection, but a down year has pushed his stock down. In a lesser-conference, his raw performance numbers, particularly his slugging (.511), were below average. His fastball whiff of 21% is slightly below average, although his breaking ball whiff is above average at 24%. His pulled in the air rate is only 10.7%, and an improvement in that aspect should hopefully improve his power numbers.</p><p><strong>Kevin Bazzell (21); Texas Tech</strong></p><p>C, RHH</p><p>Pipeline Rank: 55</p><p>Bazzell is a well-rounded profile, with relatively dimmed in-game production. His average (.307), OBP (.396) and SLG (.475) are all below average. However, he has excellent whiff rates, with an 11% overall rate, a fastball whiff of 7%, breaking ball whiff of 14%, and offspeed of 19%. His pull in the air rate is below average at 12.6%. If an organization believes this will enable his power to play up without harming his whiff rates, he has the potential to be a very solid bat at the catcher position (contingent on staying at C).</p><p><strong>Jared Thomas (21); Texas</strong></p><p>1B/OF, LHH</p><p>Pipeline Rank: 60</p><p>This is a relatively risky profile in my opinion. Thomas had a 10.1% walk rate and 20.8% strikeout rate, both of which are below average. On the other hand, Thomas’ overall, fastball, and breaking ball whiff rates are all good (20, 16, 24, respectively). He pulls the ball in the air at a strong clip (20.9%). He does run concerning chase rates, chasing 37% of breaking balls and 46% of the time with 2 strikes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vSfqkyWJ_Q1ABFxl" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=87b6124897f1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Creating an Accessible Research and Development Software For Pitching Development]]></title>
            <link>https://medium.com/@reinha3/creating-an-accessible-research-and-development-software-for-pitching-development-ffc8746e58d5?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/ffc8746e58d5</guid>
            <category><![CDATA[baseball]]></category>
            <category><![CDATA[college-baseball]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[sabermetrics]]></category>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Sat, 15 Jun 2024 15:25:50 GMT</pubDate>
            <atom:updated>2024-06-15T15:25:50.982Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Creating an Accessible Research and Development Software For Pitching Development</strong></p><p><em>The Need For a Baseball R&amp;D System</em></p><p>Data drives success in all aspects of the world today, especially in sports. Virtually all professional organizations and large college programs have an internal research and development (R&amp;D) software system to assist player development and scouting. This technology and access to data are becoming necessary for a successful club. These systems supply the foundation for sophisticated analytics. However, they often require a lot of money to construct and maintain, making them financially inaccessible to many smaller college and high school programs. To level the playing field and democratize data-driven player development, we have created an affordable, efficient internal Research and Development system for pitching development.</p><p>Creating a good R&amp;D department requires at least one of two things: a full-time analyst who understands advanced analytics and modeling, or enough money to pay for a technology service to do that for you. Those costs can be inaccessible to smaller programs. Beyond technology costs, teams also face issues with the quantity of data they have access to. A strong analytical foundation requires a vast amount of data. MLB organizations accumulate 162 games of regular season data, in addition to the playoffs and spring training. It would take years for smaller college programs and high schools to accumulate enough data to fulfill even one MLB season’s worth. Despite these constraints, we have designed a web application that uses in-depth statistical modeling and the power of the cloud to work through the issues smaller programs may face.</p><p><em>Outlining Our R&amp;D System: The Pitching + Server</em></p><p>Our R&amp;D database dubbed “The Pitching + Server” operates as an R Shiny application. We utilized the R programming language to code a series of modules that help create a data foundation. To ensure that large amounts of data are saved safely on the cloud with user privacy ensured, we constructed an SQL database supported by Amazon Web Services. This database is crucial to the success of this server. Since the technology required for this server to operate is minimal, we need to make up for it with good modeling and a large sample size. This shared SQL database allows for the server to refine our models and show detailed data visualizations. Additionally, we use an AWS password system to ensure that each team’s data is secure and confidential.</p><p>The first module in the Pitching + Server is our Stuff + generator. The module has a series of input boxes allowing the user to input thrown ball metrics from a Rapsodo, Trackman, or other similar device after a bullpen or live outing. These thrown ball metrics are used to calculate Stuff + values for each pitch in the player’s arsenal. Our Stuff + models which we outline here (The ART of W.A.R Pitching + Model) use machine learning to quantify a player’s “stuff”. This module also utilizes the model to provide players with development advice by indicating which thrown ball metrics most negatively affect their Stuff values for each pitch. The Stuff+ generator is an easy and efficient way to track a player’s development in pitch quality.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8-iVJNY1qYTLAJ9I" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*KpojASZH4MEnxOs8" /></figure><p>The second module in the Pitching + Server is our charting database. To work this module you simply input the pitch type (Fastball, Slider …), result (Swinging Strike, Ball, Groundball…), and velocity, if available. Our server tracks those results along with counts, outs, and innings. After the game, our server outputs a large data table with in-depth information about said outing. This large database is extremely useful for a team’s internal use. The number of statistics shows programs with detail on how their pitcher did, supplying a foundation for analysis and development.</p><p>Our server continuously updates season statistics on save. These data tables supply programs with information not only on the pitcher’s outings individually but cumulative season statistics. These cumulative statistics also make use of our Pitching + model in the form of Location +, a statistic that quantifies a player’s ability to command the strike zone. This stat is an important evaluation and development tool. When Stuff + numbers and a Location + value are both present for our player the module will also calculate a Pitching + value, a metric that quantifies a player’s overall pitching ability. Pitching + has been strongly correlated with future ERA in the models we have run.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jJtwhhku5gVP7MVd" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ixDndP6xf45AyC3Z" /></figure><p>The final module in the Pitching + Server is our Strike % and Stuff + calculator. In this module, the user inputs Strike%, Stuff +, and usage rate for each pitch in their arsenal. The module will calculate the player’s overall strike percentage and overall Stuff +. The sliding bars in the module are used to adjust the usage rate. This simple but effective module helps coaches refine pitch-calling strategies to optimize usage rates for their players’ arsenals. This could be a helpful tool for any pitching coach who is looking for an edge in their pitch-calling strategy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/890/0*IveTv9QbzcrqCm8B" /></figure><p>We also constructed a database of baseball development writing from across the internet. Most of this writing is simply a link to other excellent work in the field, it makes research for players a lot more efficient as they get directed to fantastic information instantly. When paired with our Stuff + Generator, this database can be used as a resource to make development progress as efficiently as possible.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*0EZIwKPVP0JwxeLv" /></figure><p><em>Moving Forward</em></p><p>Looking ahead, we plan to enhance the Pitching + Server with the use of data visualizations. Visualizations make data more understandable, especially for coaches and players. The first source of inspiration for these would be to replicate some of the pitching charts from MLB’s Baseball Savant, such as the percentiles charts. These plots very clearly show the statistical value of each pitcher through comparison to the rest of the league. Our version of this would allow you to select each player on your roster. The chart would be rendered using all of the statistics in the entire database to create averages and display where the selected player ranges compared to the rest of the level of their competition on a handful of statistics.</p><p>We also plan for increased use of machine learning models. Currently, we have our Stuff + and Location + models, which work very well. However, they can only say so much about a pitcher and do not quite represent all aspects of pitching success. We will continue to build useful models to further help smaller programs. We think that models focusing on a pitcher’s mechanical movements or aspects of their delivery that create deception could be very interesting next steps. We would also love to find a way to implement a generative AI system into our server, serving as an AI pitching coach.</p><p>Our journey to democratize baseball data to programs across the country has only begun. If you have any interest in being a part of our journey or would like to speak with us, please feel free to reach out.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ffc8746e58d5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Brief Company Update]]></title>
            <link>https://medium.com/@reinha3/a-brief-company-update-6c3fb7925c3a?source=rss-8dcb15e5d8e2------2</link>
            <guid isPermaLink="false">https://medium.com/p/6c3fb7925c3a</guid>
            <dc:creator><![CDATA[RG Pitching -- Art of W.A.R. ]]></dc:creator>
            <pubDate>Tue, 09 Apr 2024 16:54:42 GMT</pubDate>
            <atom:updated>2024-04-09T16:54:42.353Z</atom:updated>
            <content:encoded><![CDATA[<p>by Henry Gliedman and Benjamin Reinhard</p><p>We thought it would be useful for us to create a short piece of writing to give a company update on Art of W.A.R. We have been fairly busy over the past few months and would like to share what we have been up to. There is some exciting news and even more good stuff to come soon. First, Art of W.A.R, had the opportunity to attend two sports analytics conferences recently. We attended and presented at the annual SABR Baseball Analytics Conference in Phoenix, Arizona. This was a great opportunity for us to share our Pitching + Model. We received great feedback and advice from some leaders in the baseball analytics industry. This event gave us great ideas on the direction of our company going forward. We also attended the MIT Sloan Sports Analytics Conference in Boston, Massachusetts. We heard from some of the most influential names in the sports business. These two events helped us gain great experience and advice that we feel can help us properly direct our company with our next steps.</p><p>The first of those major steps for Art of W.A.R has been creating our first full software product: The Art of W.A.R Pitching + Server. The purpose of this server is to create a large database of statistics for a team, using our Pitching + model and a computer charting system. This database includes a variety of stats, ranging from traditional stats such as Innings Pitched to advanced statistics such as Location + and FIP. Our hope is that this database on its own could be useful to both coaches and players. This database can serve a similar purpose to proprietary databases used by teams across the Major Leagues, but at an affordable price. Additionally, we have compiled a set of supplementary writings from across the web. These are some of the best online resources out there and are another good resource for development. We plan to further develop our app through the incorporation of AI and machine learning to the databases. We plan to get the server in a place where one day it could act as a sort of computer pitching coach or analytics advisor.</p><p>To elaborate on our app, the most important feature is the three quantitative models we have developed: Stuff +, Location +, and Pitching +. These models attempt to quantify various aspects of a pitcher’s abilities: specifically, their nastiness (movement, velocity, etc), location (ability to throw quality strikes consistently), and their overall pitching ability, which looks at both stuff and location. We calculated Stuff + by first developing a “Stuff Constant,” which is a composite of various stats that can be considered success for a pitcher, such as whiff rate and groundball rate. After calculating these “Stuff Constants’’, we used Boruta, a feature selection algorithm to determine which ball flight metrics contributed most to a pitcher’s success, as measured by the “Stuff Constant.” Finally, we used XGBoost, a gradient boosted decision tree, to determine precisely how to weight the various ball flight metrics.</p><p>As time has progressed following our initial model, this process has stayed consistent. Our app enables us to collect vast amounts of data efficiently, helping us to refine our models. The main changes in our model have been behind the scenes, with various different weights changing. We have greatly increased the quantity of our data, and this increased quantity has helped our model become even more predictive. We are continuing to refine our model as we gather more data. We are currently working with a few partner schools to beta-test our app in exchange for increased data access. This increased access from a variety of playing levels will help our model continue to become more predictive, further enabling teams to help evaluate their pitching staff.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6c3fb7925c3a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>