<?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 Craig P on Medium]]></title>
        <description><![CDATA[Stories by Craig P on Medium]]></description>
        <link>https://medium.com/@cpoeppelman?source=rss-b17925dc1923------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*5LEilEi_Kx-MNX2T.jpg</url>
            <title>Stories by Craig P on Medium</title>
            <link>https://medium.com/@cpoeppelman?source=rss-b17925dc1923------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 17:15:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@cpoeppelman/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[2018 Club Nationals Predictions]]></title>
            <link>https://medium.com/the-flip/2018-club-nationals-predictions-94afadeb2a6b?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/94afadeb2a6b</guid>
            <category><![CDATA[stats]]></category>
            <category><![CDATA[ultimate-frisbee]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Mon, 08 Oct 2018 04:23:33 GMT</pubDate>
            <atom:updated>2018-10-20T16:02:44.925Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Simulated results for 2018 USAU Club Nationals (Men’s and Women’s). See each team’s probability of reaching different stages of the tournament.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/202/0*pSiWdSPv_QG74jnM.jpg" /></figure><p>With the seedings announced for Club Nationals, it’s time to let the predictions and speculation begin. Using my <a href="https://medium.com/the-flip/updated-club-elo-ratings-2018-post-regionals-fe3525a6d6e2">Elo ratings</a>, I simulated the tournament determining probabilities for each team reaching different rounds of the tournament.</p><p>And for the curious, here is a link to <a href="https://medium.com/the-flip/2017-club-nationals-predictions-da9939d9a1ab">last year’s predictions article</a>.</p><h4>Men’s</h4><p>The Elo ratings are much tighter for the men’s division compared to the women’s. Therefore, we see a greater tail of probabilities for teams making the finals and winning the tournament. Far and away Revolver is the favorite to win the tournament (48%) despite not winning either the US Open or the Pro Championships. The simulations say it’s about a toss-up between Revolver and the field. The only other teams with a greater than 10% chance of winning the tournament are PoNY (15%) and Ring (12%).</p><p>One other interesting caveat about the seedings this year. Machine lost a pool play game to Madison at the Elite-Select challenge in Columbus. While it’s not a given by any means, if that game’s outcome were reversed, Machine could have been the 8 seed and Madison the 9 seed. In the end, Machine takes the 9 seed which in my view <a href="https://medium.com/the-flip/hope-for-the-9-seed-fear-the-8-seed-at-club-nationals-353e9ef54f1">is favorable to getting the 8 seed</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mfOIdlSz2JGI3s2oC4ZFZw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fnJ70Pmd07qfmGj93WSFmA.png" /></figure><h4>Women’s</h4><p>The women’s division simulations match conventional wisdom with the group of pool one-seeds likely to capture the title. Coming off a strong regular season, Brute Squad has a 46% chance of winning the tournament. Fury (23%), Riot (16%), and Molly Brown (11%) all have probabilities above 10%. Unlike the men’s division, the probabilities for other teams to make the finals or win the tournament drop off quite dramatically after the top 5–6 teams.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SWgRl8k8N1B94gSBDDZQuw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z7E4selq5FZKgXd8S0u8sg.png" /></figure><h4>Championship Bracket</h4><p>For reference, here is the bracket from 2016 Club Nationals. Shows the cross-over match-ups based on pool finishes that we will see in 2018.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/639/1*xATydMgNpFMKGu3ZjEUZHg.png" /><figcaption>2016 Women’s Championship Bracket — USA Ultimate Club Nationals</figcaption></figure><h4>Methodology</h4><p>Here are some notes on how I determined these probabilities.</p><ul><li>Team ratings developed through my modified Elo rating system for ultimate. Check this <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">article</a> out for background. Also, <a href="https://medium.com/the-flip/updated-club-elo-ratings-2018-post-regionals-fe3525a6d6e2">here</a> are the team’s ratings over the course of the season.</li><li>For each division, I simulated the tournament 20,000 times to determine the probabilities of each team’s finish. One major consideration that I had was whether or not to run the simulation “hot.” This would mean adjusting a team’s rating based on the simulated outcome of each game. I chose not to run the simulation in this manner. Each team’s Elo rating entering the tournament remains the same throughout a simulated tournament.</li><li>The final score of each game was simulated. Each team’s offensive hold probability was calculated based on their Elo rating and subsequent probability for winning a given match-up. The final scores are necessary to determine tie breakers in pool play. Each game simulated included a soft cap at 85 minutes and a hard cap at 105 minutes.</li><li>To determine the win probabilities in pool play games, I use the following formula which is part of the Elo rating system.</li></ul><p>Pr(A) = 1 / (10 ^(-ELO.DIFF/400) + 1)</p><p><em>Example</em>: SF-Revolver (Elo=1840) has a <strong>87%</strong> probability of beating MIN-Sub Zero (Elo=1507) according to the ratings.</p><h4>Actual vs. Predicted Outcomes 2017 Nationals — Elo Ratings</h4><p>For those interested, here are how my Elo ratings did predicting individual games at Club Nationals 2017. A small sample size but still fairly close with the Men’s division and the Elo system overpredicted the number of upsets in the Women’s division. Note that placement games were not included in the data below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/428/1*72BJiHrAc6XDKrSBXGHcIA.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=94afadeb2a6b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/2018-club-nationals-predictions-94afadeb2a6b">2018 Club Nationals Predictions</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Do club teams match up better against certain opponents?]]></title>
            <link>https://medium.com/the-flip/do-club-teams-match-up-better-against-certain-opponents-2f0247a48a96?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f0247a48a96</guid>
            <category><![CDATA[stats]]></category>
            <category><![CDATA[ultimate-frisbee]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Mon, 01 Oct 2018 03:02:50 GMT</pubDate>
            <atom:updated>2018-10-01T03:08:23.539Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A breakdown of club teams’ performance by opponent 2010–2018</em></p><p>Over the years in club ultimate, narratives have emerged that certain teams just do not match up well with other teams. It’s not clear if this is driven by style of play, player match-ups, or some psychological effect. One of the most discussed narratives of bad team match-ups is Machine’s repeated struggles against Ring of Fire over the years (<em>full disclosure: I played on Machine from 2011–2015</em>). This has even evolved to some good twitter trolling between the teams. The first tweet is from Ring of Fire after they upset Machine in pre-quarters at 2014 Nationals. The second is from Machine after we finally broke through and beat Ring of Fire in quarters at 2015 Nationals.</p><h3>Ring of Fire on Twitter</h3><p>Fact: ring does not lose to machine at nationals. 12 10.</p><h3>Mean Machine on Twitter</h3><p>Let it be known, Machine does not lose to the same team 13 times in a row.</p><p>So if we look at the Men’s and Women’s divisions this decade, have there been certain teams that match up with certain opponents better than others? Based on the analysis below, the narrative between Machine and Ring has some statistical grounding. Let’s take a look.</p><h4>How to measure opponent specific performance for club teams?</h4><ul><li><em>W-L Record: </em>One way is to measure a team’s W-L record over the years by opponent. I’ve added a matrix similar to the one that you could find on the <a href="http://ultimate.scorereport.net/2007/scores.cgi?page=3&amp;team=1218&amp;team=1201&amp;team=1237&amp;team=1398&amp;team=1268&amp;team=1246&amp;team=1211&amp;team=1348">UPA Score Reporter</a> for each division (Men’s and Women’s).</li><li><em>Elo ratings: </em>While the W-L record is interesting to see a team’s performance vs. other Nationals caliber teams, I prefer to use my <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">Elo ratings</a> to measure performance vs. expectations. The second matrix for each division shows teams’ net Elo change when playing each Nationals qualifying opponent. A positive value (green) shows that a team outperformed Elo expectations in total from 2010-present for all games against that opponent. The ratings do account for the magnitude of the game when exchanging Elo points. For example, a bracket game at Nationals will have 3x the effect on Elo ratings compared to a placement game during the regular season. So situation does matter.</li><li>Note that a team can have a losing record against a team and still have a positive Elo change against that team over the period. For example, Revolver has a 5–3 record against GOAT but has lost Elo points in total to the Toronto squad. The reason is that Revolver is typically the favorite in the match-up and therefore will gain less Elo points in a win compared to GOAT whose wins are upsets against Revolver. A greater breakdown is shown at the end of the article when digging into the Fury &amp; Riot rivalry.</li></ul><h4>Men’s Results</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xRK8MlMre_xSK6YjZC0ORA.png" /><figcaption>Men’s win-loss record by opponent from 2010 through 2018 Regionals.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s0cTVr3q5C4Ri0NZ1abFMQ.png" /><figcaption>Women’s teams’ net Elo change by opponent from 2010 through 2018 Regionals</figcaption></figure><p>A few notes about the Men’s division that caught the eye.</p><ul><li>Machine has lost a net 295 Elo points to Ring this decade. No other teams’ match-up has exchanged more than 200 Elo points over this period.</li><li>Revolver has a 14–0 record against Ring of Fire and a 13–0 record against Truck Stop this decade. While Revolver has been rated higher during this time, it is a bit shocking that neither Ring or Truck has won a game in this match-up.</li><li>Revolver can also claim a winning record against every Nationals team this decade. Quite a feat. Despite this, their 11–7 record against Sockeye and 5–3 record against GOAT resulted in negative Elo points over this period.</li></ul><h4>Women’s Results</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1oq3n2JTa0--90VelFubwQ.png" /><figcaption>Women’s win-loss record by opponent from 2010 through 2018 Regionals</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4-uWwrEhM_szlTOblokwUA.png" /><figcaption>Women’s teams’ net Elo change by opponent from 2010 through 2018 Regionals</figcaption></figure><p>A few notes about the Women’s division that caught the eye.</p><ul><li>Fury and Riot have played a total of 39(!) games in USAU/UPA competition over this period. Fury has the rivalry edge with a 20–19 record. Since Fury has been rated higher over most of this period, Riot has gained a net 131 Elo points despite the losing record. More on the rivalry below.</li><li>The biggest two Elo gains between opponents are both from Pop. Heist and Nemesis have dominated this Midwest match-up over the years. Pop is a combined 1–31 against these two teams.</li></ul><h4>Statistically significant?</h4><p>One point to consider is the potential variation in game results over this time. If a team plays ~20 opponents 10 times each, we should expect that they will have a net positive Elo point change against some opponents and net negative against others just by chance. This means that there may not be match-up advantages, but instead statistical variation.</p><p>I have not tested for this to see the probability that these Elo changes could have occurred by chance. Based on the sample size between opponents, there is a chance that the gains/losses are due to chance. Despite this, it is still interesting to evaluate a team’s performance against opponents over this period.</p><h4>Example: Fury vs Riot (2010-today)</h4><p>To help show how these net Elo changes were calculated between opponents, I’m showing the Fury vs. Riot rivalry in more detail. Fury has a 20–19 W-L record over this period and lost a net 131 Elo points to Riot.</p><p>Keep in mind that depending on the type of game, the Elo change from the outcome will vary. This is based on the K-factor in the Elo rating. Here are the K-factors for my ultimate Elo ratings. A higher K-factor means more Elo points are exchanged based on a game outcome. This is a heuristic input to the Elo ratings. Also the margin of victory is taken into account for Elo changes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/903/1*nCUCrkBB40eB7e76maxmIQ.png" /><figcaption>Ultimate Elo Ratings — K-Factors by club tournament type and game type</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*shNn9lrWOHhM6KL9Q23bOw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Glg0Yv5YMeFucf_50iD-jg.png" /></figure><p>And if you want to get really detailed, Fury has outscored Riot over these 39 games 494–485. Only 9 points separate them over this decade.</p><p>O and to my Machine friends… sorry about writing this article.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/455/1*zwGCTVVVlDnI7k8Y-PN5cw.gif" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2f0247a48a96" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/do-club-teams-match-up-better-against-certain-opponents-2f0247a48a96">Do club teams match up better against certain opponents?</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Updated Club Elo Ratings — 2018 Post-Regionals]]></title>
            <link>https://medium.com/the-flip/updated-club-elo-ratings-2018-post-regionals-fe3525a6d6e2?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe3525a6d6e2</guid>
            <category><![CDATA[ultimate-frisbee]]></category>
            <category><![CDATA[stats]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Sun, 30 Sep 2018 16:53:01 GMT</pubDate>
            <atom:updated>2018-09-30T16:53:01.784Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Men’s and Women’s Divisions updated after 2018 Regional tournaments</em></p><p>This is the last installment of Elo ratings before Club Nationals. While we’re still awaiting seedings, we can start to see which teams may be overseeded or underseeded comparing their Elo rating to their USAU ranking.</p><p>Stay tuned for Club Nationals <a href="https://medium.com/the-flip/2017-club-nationals-predictions-da9939d9a1ab">predictions</a> similar to what I did last year.</p><h4>Men’s</h4><p>The big change here is Furious jumping up to a 1511 rating after winning Regionals over Sockeye. They come in at #8 in the Elo ratings which is still below their #5 USAU ranking.</p><p>A few notes on which teams may be seeded away from their Elo rating at Nationals.</p><ul><li>Sockeye could end up being overseeded with their #2 USAU ranking. It is unclear where they will be seeded at Nationals after a loss to Furious at Regionals.</li><li>Doublewide will likely be underseeded again this year similar to last year. The <a href="https://medium.com/the-flip/2017-club-nationals-predictions-da9939d9a1ab">Elo ratings really liked Doublewide’s chances last year</a> even as a #12 seed. Their run to the final was less of a surprise in the Elo ratings compared to their seed. Let’s see if they can do it again this year.</li><li>Furious is likely to be overseeded with a #5 USAU ranking. Their resume is thin and I imagine a lot of teams will be happy to see them as the #2 team in their pool as opposed to Truck Stop or DiG.</li><li>Bravo will likely be underseeded with their #12 Elo rating but they likely will only fall a few spots to a #14 or #15 seed at Nationals.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Xb9mLedD44hs2DmsbLSxRQ.png" /></figure><h4>Women’s</h4><p>Not many surprises at Regionals in the Women’s division. Still let’s take a look at teams whose ratings vary from their USAU rankings.</p><ul><li>Riot and Molly Brown are swapped in the Elo ratings compared to the USAU rankings. Despite this, it is likely that Riot will take the #3 seed into Nationals.</li><li>Phoenix comes in at #7 in the Elo ratings compared to #11 in the USAU rankings. Winning a lot of close games is better in the Elo ratings compared to the USAU rankings which is why I see the discrepancy.</li><li>Traffic is also likely to be underseeded with their Elo rating of #10 compared to their #15 USAU ranking. After a 1 point loss to Riot at Regionals, Traffic showed they have the potential to be a dangerous team at Nationals.</li><li>Rival is the only team likely to significantly overseeded when comparing Elo ratings to USAU rankings. They come in at #15 in the Elo ratings vs. #10 in the USAU rankings.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QNtRwOsPJIUGX7NGnK2cmw.png" /></figure><p>For a background on how I created these ratings, check out this <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">article</a>. Also, check out historical Elo ratings for club teams back to 2004 for both <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-mens-85343abfd15d">men’s</a> and <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-womens-95b4fc9355bb">women’s</a>. I’m hoping to update the plots in these articles at some point.</p><p><em>Please note that only teams who have made Nationals in their history are included in these ratings and only games between these teams affect Elo ratings. Therefore there are teams that are excluded but are probably true Top 24 teams in the club game today. Also, I don’t advocate using these rankings to allocate Nationals bids. Just another way to measure true team strength empirically in addition to the USAU rankings. Teams with few games against other teams in the rating system may not have their rating correct to their current team strength.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fe3525a6d6e2" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/updated-club-elo-ratings-2018-post-regionals-fe3525a6d6e2">Updated Club Elo Ratings — 2018 Post-Regionals</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Updated Club Elo Ratings — 9/3/18]]></title>
            <link>https://medium.com/the-flip/updated-club-elo-ratings-9-3-18-796e0f155328?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/796e0f155328</guid>
            <category><![CDATA[ultimate-frisbee]]></category>
            <category><![CDATA[stats]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Mon, 03 Sep 2018 23:39:50 GMT</pubDate>
            <atom:updated>2018-09-03T23:39:50.869Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Men’s and Women’s Divisions updated after the 2018 Pro Championships</em></p><p>This is the last installment of Elo ratings for the 2018 club season before the series begins. Obviously these ratings aren’t as important as the USAU rankings which are used for bid allocations. More on the discrepancies below.</p><p>Also, note that WUCC 2018 games are not included in the ratings as well as games between teams on these lists and non-USAU, international opponents.</p><h4>Men’s</h4><p>After winning the Pro Championships, PoNY shoots up to #2 in the Elo ratings. This is the highest PoNY’s Elo rating has been ever (dating back to 2004)! Their previous high was when they reached a 1639 Elo rating after two pool play wins at 2016 USAU Nationals. Unfortunately, they were upset by Furious George in the pre-quarters of that tourney.</p><p>It is interesting to compare the Elo ratings to the estimated USAU rankings from Cody Mills.</p><ul><li>Doublewide, Rhino, and Johnny Bravo are all ranked 4–8 spots higher in the Elo ratings than their estimated USAU rankings.</li><li>Madison and Temper are ranked 5–7 spots lower in the Elo ratings compared to their estimated USAU rankings.</li><li>Another difference is which teams would be considered top 16 teams. The Elo ratings have SoCal Condors and GOAT in the top 16 as well as Johnny Bravo safely in with a #9 rating. Furious George and Chain are not top 16 teams in the Elo ratings.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dTzUzOKzJo5HQndAsOSPdw.png" /></figure><h4>Women’s</h4><p>Not much is new in the Women’s division. There is still the top four of Brute Squad, Fury, Riot, and Molly Brown. Scandal is the closest to the top tier with an Elo rating inching closer to the top 4 after a solid Pro Championships tournament.</p><p>Again, here are the main discrepancies between the Elo ratings and the estimated USAU rankings from Cody Mills.</p><ul><li>Riot and Molly Brown are swapped in the Elo ratings compared to the estimated USAU rankings. While this isn’t a big discrepancy, it has a big effect on what the semis match-ups are if the top 4 teams make it there. Keep in mind that Riot missed the semis last year.</li><li>Showdown is 5 spots higher in the Elo ratings compared to the estimated USAU rankings and falls in the top 16.</li><li>The Elo ratings are much lower on Pop and Wildfire as they are both are outside the top 16 in the Elo ratings compared to the estimated USAU rankings.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dJbl2_UE-AqOqnyzygKHyA.png" /></figure><p>For a background on how I created these ratings, check out this <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">article</a>. Also, check out historical Elo ratings for club teams back to 2004 for both <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-mens-85343abfd15d">men’s</a> and <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-womens-95b4fc9355bb">women’s</a>. I’m hoping to update the plots in these articles at some point.</p><p><em>Please note that only teams who have made Nationals in their history are included in these ratings and only games between these teams affect Elo ratings. Therefore there are teams that are excluded but are probably true Top 24 teams in the club game today. Also, I don’t advocate using these rankings to allocate Nationals bids. Just another way to measure true team strength empirically in addition to the USAU rankings. Teams with few games against other teams in the rating system may not have their rating correct to their current team strength.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=796e0f155328" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/updated-club-elo-ratings-9-3-18-796e0f155328">Updated Club Elo Ratings — 9/3/18</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Flip’s In-Game Win Probability Model (v1.0) for Ultimate]]></title>
            <link>https://medium.com/the-flip/the-flips-in-game-win-probability-model-v1-0-for-ultimate-a382cce76c8c?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/a382cce76c8c</guid>
            <category><![CDATA[ultimate-frisbee]]></category>
            <category><![CDATA[stats]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Mon, 03 Sep 2018 15:43:52 GMT</pubDate>
            <atom:updated>2018-09-03T22:26:46.472Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Applying the win probability model from other sports to ultimate.</em></p><p>In the 2017 USAU Club Nationals final, Brute Squad scored to take half 8–5 over Fury. A comfortable but not insurmountable lead. Brute Squad coach, Ariel Jackson, gave his halftime interview <a href="http://www.youtube.com/watch?v=L51qK8_Nhvo&amp;t=53m0s">saying</a> “8–5 is the most dangerous score in ultimate.” So what were the chances Fury would make the comeback to win the game? And how would their chances change if Fury were not receiving out of halftime? And as they get closer to the cap, how does this reduce Fury’s chances of a comeback? Sounds like a great reason to make a win probability model for ultimate! <em>Spoiler: Brute Squad lost 14–13 to Fury.</em></p><p>Win probability models have become quite popular for other <a href="https://fivethirtyeight.com/features/how-our-2018-world-cup-predictions-work/">sports</a> and <a href="https://projects.fivethirtyeight.com/2018-midterm-election-forecast/house/?ex_cid=rrpromo">politics</a> recently. These models come with their fair share of <a href="https://www.theringer.com/2017/2/13/16077450/nfl-playoffs-super-bowl-new-england-patriots-atlanta-falcons-win-probability-d572b7e7fd">criticism</a> but I still find them really enjoyable and insightful. So why not make one for ultimate?</p><p>Below is a high-level description of an in-game win probability model for ultimate that calculates both the receiving team and pulling team’s win probability for a certain score and time remaining in the game.</p><p>First, let’s talk through the results. A longer discussion of how the model works is at the bottom of the article.</p><h4>Win Probability Table (Men’s — 70% O-Line Hold Rate)</h4><p>If we assume equal teams that score offensive points with a 70% probability, what is the win probability for every score scenario? The table below shows the win probability for the receiving team for each score scenario.</p><ul><li>Note that the simulations are for games to 15 under the USAU cap rules used at Nationals for semis and finals games.</li><li>The values below are for score scenarios where the <strong><em>soft cap is 85 minutes away. </em></strong>Essentially this is the same as the start of the game.</li><li>I am calling this the “Men’s” table based on historical data for the club and college division (further discussion at the bottom of the article).</li></ul><p>How to read the table below:</p><ul><li>Each square indicates the receiving team’s win probability for a given game score. All 1st half win probabilities assume the receiving team also received to start the game.</li><li>The pulling team’s win probability is (1 -value shown below).</li><li>You can chart the path of a game by looking at corresponding squares based on the result of a point. For example, a team receiving down 5–8 has a 15.5% chance of winning the game (assuming 85 min before soft cap). If they get broken, they are now receiving down 5–9 and have a 7.5% chance of winning. If they hold and are pulling down 6–8, they now have an 18.2% chance of winning (1–81.2%). Note that a true simulation would change the time remaining after each point so this is just an estimation of how to chart a game path with the win probability model.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nxKmS0-cH7Jjsp34LZJuvQ.png" /><figcaption>Win probabilities for receiving team assuming 70% o-line hold rate</figcaption></figure><p>Note that there is a +/-0.2% confidence interval around each value so there may be situations where the win probability skews slightly away from what we’d expect (for example, receiving at 5–5 has a 50.6% win probability while receiving at 4–4 has a 50.8% probability).</p><h4>Win Probability Table (Women’s — 63% O-Line Hold Rate)</h4><p>If we assume equal teams that score offensive points with a 63% probability (closer for the Women’s division), what is the win probability for every score scenario? The table below shows the win probability for the receiving team for each score scenario.</p><ul><li>Note that the simulations are for games to 15 under the USAU cap rules used at Nationals for semis and finals games.</li><li>The values below are for score scenarios where the <strong><em>soft cap is 85 minutes away. </em></strong>Essentially this is the same as the start of the game.</li><li>I am calling this the “Women’s” table based on historical data for the club and college division (further discussion at the bottom of the article).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g3TnhcKBJ6Wo5xEDR0UUhQ.png" /><figcaption>Win probabilities for receiving team assuming 63% o-line hold rate</figcaption></figure><h4>Effect of who received to start the game (70% O-Line Hold Rate)</h4><p>So how do the win probabilities change in the 1st half depending on who receives to start the 2nd half? Below are tables for the 1st half win probabilities showing both scenarios for 70% o-line hold rates. The table on the left shows the receiving team’s win probability when they are <em>pulling after halftime. </em>The table on the right shows the receiving team’s win probability when they are <em>receiving after halftime</em>.</p><p>For example, suppose it is a tied 7–7 game. The receiving team’s win probability is 51.9% if they have to pull out of half. This win probability rises to 60.4% if they get to receive out of half.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XM6lrpWtakA41U33okVXPg.png" /><figcaption>1st half win probabilities for receiving team assuming 70% o-line hold rate. Table on the left shows receiving teams who are pulling out of half and the table on the right shows receiving teams who are receiving out of half.</figcaption></figure><h4>Effect of time and cap on win probability</h4><p>If you have made it this far in the article, I’m sure you are someone who is aware that ultimate games don’t always make it to 15. The cap has a significant impact on the win probability for a given game scenario. Let’s look at the 8–5 game score mentioned in the intro for different times remaining before the cap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1006/1*7R9YFHzmaKZiYlgy9JyoEw.png" /><figcaption>Win probabilities and average expected final score for an 8–5 game with various times remaining until the soft cap</figcaption></figure><p>As you can see from the table, the chance for the receiving team to make a comeback starts to decrease as time gets closer to the cap (as we would expect). With an 8–5 score, the team losing should start to feel anxious with about 40 min to go until the soft cap. At this point, with every minute, your chances of coming back start to decrease.</p><p>Just to reiterate the note from earlier, each simulation has a +/-0.2% confidence interval. This is why the win probability oscillates as the time remaining decreases from 85 min to ~40 min.</p><h4>So what was Fury’s win probability out of half? And how does the win probability model compare to actual game results?</h4><p>Looking at the table above Fury had a 17.4% chance of winning the 2017 Final coming out of half. This assumes the teams score 63% of o-line points and that there was approximately 35 minutes until the soft cap. But Fury did win the game. So is the model wrong? Let’s look at other game results from the club division.</p><p>Based on the point-by-point <a href="https://play.usaultimate.org/teams/events/match_report/?EventGameId=ZWdQdU4xC5iAuJn0%2bJYeit6ybq1NVutReOaT02XYd0c%3d">tables</a> from USAU Club Nationals (2014–2017), we can look at very similar game situations. Unfortunately, we cannot tell who is receiving the point out of half from the USAU website and time remaining to the cap, so I looked at all game situations where the receiving team was down 6–9.</p><ul><li>For the Men’s club division, there were 55 Nationals games from 2014–2017 where the receiving team was down 6–9. The receiving team won 7 of these games or <strong><em>12.7% of the games</em></strong>.</li><li>For the Women’s club division, there were 44 Nationals games from 2014–2017 where the receiving team was down 6–9. The receiving team won 7 of these games or <strong><em>15.9% of the games</em></strong> (including the Fury comeback in the 2017 Final).</li><li>These results are comparable to what the win probability tables show earlier in this article. It would be great to compare the model to a larger data set in the future. One other note from the actual game data is that the teams may not be equal opponents. Likely the team leading 9–6 is a better team and therefore their probability to win will be better than a model which assume both teams have an equal chance of scoring offensive points.</li><li>So is 8–5 the most dangerous lead in ultimate? ¯\_(ツ)_/¯ But it’s probably a good talking point to keep your team motivated for the 2nd half if you’re in the lead.</li></ul><h4>How did the model change the win probability for Brute Squad and Fury over the course of that game?</h4><p>I am showing the win probability over the game in a table form as I think it gives a better representation of how the cap affects the win probability in the second half. Again, the model assumes both teams have a 63% probability of scoring o-line points.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_CJONAgfzb-SW63OuAldow.png" /><figcaption>Win probabilities and average expected final scores for the Brute Squad and Fury USAU Club Nationals Final in 2017</figcaption></figure><h4>What’s next for the model?</h4><p>I’m hoping to use the model to write articles about in-game strategy as well as share insights about how specific plays in games swung a team’s win probability. Hopefully some more tech savvy individuals will enjoy the idea enough to incorporate it into some live broadcasts or websites that give live updates for big games. Charles Kerr already played around with the model with his JinxCam coverage of D-1 College Nationals this year. Please reach out to me if you have any interest or questions about the model.</p><h3>How the model works</h3><p>The model works by simulating game results based on a certain game scenario. These are the game scenario inputs.</p><ul><li>Score of the game</li><li>Team that is receiving</li><li>Team that received to start the game</li><li>Cap on or time remaining before the cap</li><li>Winning score required (e.g. game to 15)</li></ul><p>The model then simulates the remainder of the game based on the following assumptions.</p><ul><li><strong>Offensive hold rate</strong>: This is the probability that each team scores offensive points. Note that this does not indicate the scoring rate of each <em>possession. </em>The model can input different hold rates for each team (unequal opponents) but the default model shows results for teams with equal hold rates. The model assumes that a team’s probability to score an offensive point is the same regardless of what the score is in the game, the outcome of the previous point, or how early or late in the game it is. The model simulates every point of the game with this probability. For example, if 66% is the hold rate probability, every point of the game would be simulated like rolling a die. If 1–4 are rolled, the offensive team scores the point. If 5–6 are rolled, the defensive team breaks. This repeats until a team wins the simulated game.</li><li><strong>Offensive hold rate (by division)</strong>: The default model uses an offensive hold rate of <strong><em>70% for the Men’s division</em></strong> (college &amp; club) and <strong><em>63% for the Women’s division</em></strong> (college &amp; club). The rates are based on average hold rates between winning and losing teams from USAU Club Nationals and USAU D-1 College Nationals. See the following articles (<a href="https://medium.com/the-flip/offensive-hold-rates-college-nationals-2014-2017-5b9caa180e41">college</a>, <a href="https://medium.com/the-flip/offensive-hold-rates-club-nationals-2014-2017-be2eab59afb">club</a>) with more information.</li><li><strong>Wind</strong>: The default model does not factor in wind. The offensive team’s probability to score a point is the same regardless of which direction they are attacking.</li><li><strong>Time per point</strong>: The duration of each point is randomized based on a log normal distribution. The log normal distribution was selected based on data from the AUDL and WUGC 2016. See this <a href="https://medium.com/the-flip/how-long-do-ultimate-points-take-71d60dd1ef33">article</a> for a detailed breakdown. I used a log normal distribution with mean=4.75 and std.dev=0.75 in the natural logarithm. In English, this means that the average point duration is ~2.5 minutes and the median point duration is ~2 minutes. In addition, I added 1 minute between points. The fitted data below to the WUGC 2016 Men’s data shows a log normal distribution of mean=4.73 and std.dev=0.75 with an added 1 minute for time between points to give you a feel for the distribution of time per point in the simulations. <em>Note that timeouts and injuries are reflected in the time per point distribution and therefore are not inputs to the model.</em></li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0Zdcp-FEw2TFya7t1HH5hw.png" /><figcaption>Distribution of time per point and fitted log normal distribution to WUGC 2016 — Men’s game data, 2,766 observations</figcaption></figure><ul><li><strong>Halftime length</strong>: This can be changed based on the tournament format. The default win probability model is based on USAU Nationals where halftime is 10 minutes.</li><li><strong>Cap rules</strong>: In my opinion this is the most annoying part about applying a win probability model to ultimate. Almost every tournament has different cap rules and USAU Nationals has different cap rules within the same tournament! Because of this, a new win probability model has to be built for each cap rule format. The default win probability model uses the cap <a href="https://play.usaultimate.org/assets/1/7/2017_Nationals_Guide_LR.pdf">rules</a> for USAU Nationals semifinal and final games. Games are played to 15 with the soft cap going on after 85 minutes. Add 2 points to the high score and play to that number.</li><li><strong>Number of simulations</strong>: This is number of simulations run for each game scenario including time remaining. Since the win probabilities are simulated data, the more simulations the lower the standard error in the results. For the default win probability model, 250,000 simulations were run for each game scenario at 5 minute increments (it was a lot of computer simulations!). This results in a 95% confidence interval of +/- 0.2% for 50% win probability scenarios. The confidence interval becomes smaller as the win probability moves closer to 0%/100%.</li></ul><p><strong><em>Notes</em></strong></p><ul><li>Huge shout out to Charles Kerr who helped run these simulations that took weeks. And to his enthusiasm in adding win probabilities to the his JinxCam coverage. He ran a <a href="https://www.youtube.com/watch?v=kM7E90-dHgk&amp;feature=youtu.be&amp;t=58m58s">beta</a> of the calculation in the ticker of his USAU D-1 College Nationals 2018 coverage.</li><li>There are a few others (that I’m aware of) that have built in-game win probability models for ultimate. Tom Murray had an <a href="https://ultiworld.com/2016/09/28/nationals-probabilities-every-teams-chances-to-win-a-title/">article</a> in Ultiworld that touches on this topic when predicting Club Nationals outcomes. In addition, Scott Grindy wrote an <a href="https://scottcgrindy.wordpress.com/2016/11/23/absorbing-markov-chains-and-ultimate-part-i/">article</a> using Markov Chains to estimate win probability. I used a different approach then they did but I still want to give some credit to others who have done this before.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a382cce76c8c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/the-flips-in-game-win-probability-model-v1-0-for-ultimate-a382cce76c8c">The Flip’s In-Game Win Probability Model (v1.0) for Ultimate</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Offensive Hold Rates: Club Nationals 2014–2017]]></title>
            <link>https://medium.com/the-flip/offensive-hold-rates-club-nationals-2014-2017-be2eab59afb?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/be2eab59afb</guid>
            <category><![CDATA[stats]]></category>
            <category><![CDATA[ultimate-frisbee]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Sun, 02 Sep 2018 18:40:54 GMT</pubDate>
            <atom:updated>2018-09-02T18:40:54.822Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A dive into offensive hold rates at USAU Club Nationals 2014–2017</em></p><p>This is a follow-up to an <a href="https://medium.com/the-flip/offensive-hold-rates-college-nationals-2014-2017-5b9caa180e41">article</a> I wrote last year about hold rates in the college division at USAU Nationals. Below are data cuts of hold rates for the Men’s and Women’s divisions at USAU club Nationals. Data were scraped from the USAU Club Nationals web pages.</p><h3>Club Division</h3><p>The table below shows the offensive hold rates for each division at USAU Club Nationals from 2014–2017. I broke down the rates by winning teams and losing teams from each game. The reason is that the lesser of two opponents will end up playing more offensive points and skewing the total number for the division, so this should be taken into account when analyzing the results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/654/1*Nc2k1IBZm72m4Aef4Bq4bQ.png" /></figure><h3>Comparison to College Division</h3><p>The table below shows the offensive hold rates for each division at USAU D-I College Nationals from 2014–2017.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/774/1*upnxICZ9YDgF-LJgywDmGQ.png" /></figure><p>For the Men’s division, the hold rates are very comparable for the club and college divisions. The Women’s division has slightly higher hold rates in the club division compared to the college division.</p><h3>Club Division by Year &amp; Weather</h3><p>To add a little more detail for each year, I show the average hold rate by division and the average wind speed each year. The wind data was collected from weatherunderground.com and averaged over all the round times for Club Nationals.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/812/1*vT74kfkEi8uEwQrLX-gN-A.png" /></figure><p>Note that stronger wind correlates with slighter lower hold rates for the Men’s division and slightly higher hold rates for the Women’s division. I wouldn’t make any sweeping conclusions from these data though. It’s hard to know which direction the fields were in relation to wind direction. For instance a heavy upwind/downwind game would likely increase the hold rate whereas a heavy crosswind game would likely decrease the hold rate.</p><h3>Club Division vs. Simulated Results</h3><p>Another way to think about the hold rates for winning teams and losing teams is to compare hold rates for simulated games where each team has the same probability to score each offensive point.</p><p>For the Men’s division, I used a 70% offensive hold probability in the simulations and a 63% probability for the Women’s division. As you can see, the club division hold rates from Nationals are similar to what you see from the simulations. The actual club game data have a bit lower rates for losing teams and a bit higher rates for winning teams. This is to be expected as teams aren’t equal in every game. As teams are more unequal, you would expect a bigger difference in hold rates between winning and losing teams. Comparing to the Club division, we see that the Women’s division has more unequal teams playing each other than the Men’s division.</p><p><em>Simulated Hold Rates</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cBJHElX0wqGL1q-j-1BYsg.png" /></figure><p><em>Club Nationals Hold Rates (same table as above)</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/654/1*Nc2k1IBZm72m4Aef4Bq4bQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=be2eab59afb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/offensive-hold-rates-club-nationals-2014-2017-be2eab59afb">Offensive Hold Rates: Club Nationals 2014–2017</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Updated Club Elo Ratings — 8/6/18]]></title>
            <link>https://medium.com/the-flip/updated-club-elo-ratings-8-6-18-a5193b18b078?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/a5193b18b078</guid>
            <category><![CDATA[ultimate-frisbee]]></category>
            <category><![CDATA[stats]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Fri, 10 Aug 2018 18:37:49 GMT</pubDate>
            <atom:updated>2018-08-10T18:41:51.043Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Men’s and Women’s Divisions updated after the 2018 US Open</em></p><p>This is the first installment of Elo ratings for the 2018 club season. We’ll see how they evolve over the next couple of months. Note that WUCC 2018 games are not included in the ratings as well as games between teams on these lists and non-USAU, international opponents.</p><h4>Men</h4><p>Despite the loss to Sub Zero, Revolver remains the clear #1 in the Elo ratings even without their WUCC games included. Ring of Fire is the clear #2 in the ratings and Sockeye is the only other team with a rating above 1600.</p><p>PoNY comes in at #7 in the ratings after winning the US Open. You may expect them to be higher but their losses at Colorado Cup are still weighing on their Elo rating.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s4ehKzUIbOXUUgzY7GWSqg.png" /></figure><p>One other tidbit regarding the US Open is ranking Sub Zero’s upset of Revolver historically. For regular season bracket games, this game is the 3rd biggest upset in the Men’s division for all tournaments back through 2004. Heading into the game, Sub Zero had a 1511 rating and Revolver had a 1877 rating. The biggest upset for a regular season bracket games in the Men’s division was Dig’s <a href="https://ultiworld.com/2017/07/05/club-update-dig-beats-ironside-wild-card-leads-ne-mixed-riot-starts-strong/">upset</a> of Ironside last year at the Boston Invite. This year’s Sub Zero victory over Revolver is similar to Sub Zero’s <a href="https://ultiworld.com/2013/08/19/frostbitten-top-seed-ironside-falls-to-surging-sub-zero-in-chesapeake-finals/">upset </a>of Ironside in the final of the 2013 Chesapeake Invite (1518 vs. 1883). This was after Sub Zero also upset Ironside in pool play at the same tournament. Maybe there’s something about Sub Zero and upsets in the month of August.</p><p>The data are below for the top 5 upsets in the regular season for both pool play and bracket games. I would take all these upsets with a grain of salt as teams don’t often field full rosters for regular season events especially in the early season.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E-ThFWcj37try0VaHja2HQ.png" /><figcaption>Note: Ronin never qualified for Nationals but are included in the Elo ratings as part of the historic team performance for FL-United. See this <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">article </a>for more information on multiple teams included in a franchise’s history.</figcaption></figure><h4>Women</h4><p>The top women’s teams continue to stay in a tier of their own in the Elo ratings. Brute Squad tops the list at #1 after their US Open title with Riot taking the #2 spot even without their WUCC results included in the ratings. Fury and Molly Brown fill out the top tier. Scandal is the team closest to the top 4 with Ozone the only other team above a 1600 rating.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZQ5Ky_G0r2PHdMH2qPoZEg.png" /></figure><p>For a background on how I created these ratings, check out this <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">article</a>. Also, check out historical Elo ratings for club teams back to 2004 for both <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-mens-85343abfd15d">men’s</a> and <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-womens-95b4fc9355bb">women’s</a>. I’m hoping to update the plots in these articles at some point.</p><p><em>Please note that only teams who have made Nationals in their history are included in these ratings and only games between these teams affect Elo ratings. Therefore there are teams that are excluded but are probably true Top 24 teams in the club game today. Also, I don’t advocate using these rankings to allocate Nationals bids. Just another way to measure true team strength empirically in addition to the USAU rankings. Teams with few games against other teams in the rating system may not have their rating correct to their current team strength.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a5193b18b078" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/updated-club-elo-ratings-8-6-18-a5193b18b078">Updated Club Elo Ratings — 8/6/18</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Offensive Hold Rates: College Nationals 2014–2017]]></title>
            <link>https://medium.com/the-flip/offensive-hold-rates-college-nationals-2014-2017-5b9caa180e41?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/5b9caa180e41</guid>
            <category><![CDATA[stats]]></category>
            <category><![CDATA[ultimate-frisbee]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Fri, 29 Dec 2017 16:31:38 GMT</pubDate>
            <atom:updated>2017-12-29T16:31:38.569Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A dive into offensive hold rates at USAU D-I College Nationals 2014–2017</em></p><p>If you have been a player, coach, or spectator in ultimate, it’s likely you’ve heard some variation of the following after a game.</p><ul><li><em>“Our o-line was terrible that game. The d-line got the breaks we needed but we just got broken too many times to win.”</em></li><li><em>“Our d-line had their chances but we just couldn’t put them in. You can’t ask the o-line to do anything more. They only got broken a few times and we still lost.”</em></li></ul><p>These statements aren’t always the best way to build a sense of unity within a team, but they get at a question about how frequently teams score when starting a point on offense. The rates that an offense will score is highly dependent on the division, the weather, and the opponent that a team is facing. But if we aggregate data for many games, <strong>at what rate do teams score offensive points?</strong></p><p>For this article, I want to share what rates offensive lines score points in the D-I college division. To do this, I scraped data from USAU’s website which includes point-by-point score totals for most games at D-1 College Nationals from 2014–2017. From the point-by-point data, we can determine if each point is a break or an offensive hold for each team. Below I’ve aggregated these data for 208 games in the Men’s division and 198 games in the Women’s division. There were some games missing from the USAU website or games had missing or incorrect point-by-point entries. Regardless, this is still a nice data set to gather some high-level data for the college game.</p><h3>Offensive Hold Rates by Division</h3><p>The table below shows the offensive hold rates for each division at USAU D-I College Nationals from 2014–2017. I broke down the rates by winning teams and losing teams from each game. The reason is that the lesser of two opponents will end up playing more offensive points and skewing the total number for the division, so this should be taken into account when analyzing the results.</p><p>One other note for clarification. I am defining the term “offensive hold” to be when the team who receives the disc (offensive line or offensive point) to start a point is the team that scores that point. It does not matter how many turnovers there are during the point.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/774/1*upnxICZ9YDgF-LJgywDmGQ.png" /></figure><p>For strategy, I believe these data reveal that a team should aim to score on 75% of its offensive points and break its opponents on 33% of its defensive points. Obviously any team that holds offensive points at a higher rate than their opponent will win, but I think it’s useful to have a target for o-lines and d-lines to set as goals for a day or a tournament. Also, you can use these rates to rebut any claims that a d-line or o-line isn’t pulling their weight.</p><h3>Offensive Hold Rates &amp; Defensive Break Rates by Team</h3><p>Below are scatter plots for each team’s offensive hold rate and defensive break rate for each Nationals. On average, each team in the data set has 53 points on offense and 53 on defense. Better teams will have more games and points in total and skew towards more points on defense.</p><p>I call out the champions as well as a few outliers. Overall the data are much more concentrated for the Men’s division.</p><p><strong><em>Men’s Division</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*ifDx92ki36aklDMYdt_H5A.png" /></figure><p><strong><em>Women’s Division</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*40lP9kDKs36WYXhGQIcHbA.png" /></figure><h3>Top 10 Team Performances by Division (2014–2017)</h3><p>Here are the top 10 team tournament performances by offensive hold rate and defensive break rate. These only include data from my data set and will be missing at least a few points from each game. More on this in the section below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*E46kidue9BIdZKKjhW7WkQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tkHUF8l3eTNitrTUziCIbw.png" /></figure><h3>Data</h3><p>The data used here were scraped from the match report web pages on USAU’s website (<a href="https://play.usaultimate.org/teams/events/match_report/?EventGameId=B6uCqztNHp6t%2f3Lc67M5xsEnJQSeSeoHDijqURDEaQg%3d">example</a>). The data include pool play, bracket, and consolation games for each division. There are 208 games in the Men’s division and 198 games in the Women’s division included in the data set.</p><p>It’s hard to tell how accurate the point-by-point scoreboards are at the bottom of each webpage, but I’ll take a leap of faith and trust their accuracy. I’m sure there are some small mistakes in the data but I have no reason to think there are any systemic issues that muddy the high-level results above.</p><p>Since it is unclear from the data who received to start each game and half, these initial points each half were discarded. In addition, several games were removed from the data set because the point-by-point match report data did not sync with the final score. Here are a <a href="https://play.usaultimate.org/teams/events/match_report/?EventGameId=hFhMN9jpHDPi1QBp9Ar98qjkXY%2fmsM9Uhza5a8NrjLk%3d">few</a> <a href="https://play.usaultimate.org/teams/events/match_report/?EventGameId=s7CZqYzz0t96XKANekxaNXQZU2kVrVnI8KHd%2bpuZqVc%3d">examples</a> of games that I discarded.</p><p>For the team by team data above, it is important to keep in mind that some teams appeared to rest their top players for their last pool play game. This would impact their hold rates and break rates for the overall tournament. For example, in 2015 North Carolina (M) lost their meaningless last pool play game 15–5 before going on to win the whole tournament.</p><p>All told, the data aren’t perfect but they are pretty interesting to analyze.</p><p>As always, let me know what you think of these data or if there are any analyses you’d like to me to run. I’m always looking for good ideas to run with.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5b9caa180e41" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/offensive-hold-rates-college-nationals-2014-2017-5b9caa180e41">Offensive Hold Rates: College Nationals 2014–2017</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[By The Numbers: Club Nationals 2004–2017]]></title>
            <link>https://medium.com/the-flip/by-the-numbers-club-nationals-2004-2017-deaabacf5ba3?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/deaabacf5ba3</guid>
            <category><![CDATA[stats]]></category>
            <category><![CDATA[ultimate-frisbee]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Mon, 30 Oct 2017 21:53:45 GMT</pubDate>
            <atom:updated>2017-10-30T22:39:49.219Z</atom:updated>
            <content:encoded><![CDATA[<p><em>How did 2017 stack up? A look at the biggest upsets, best match-ups, and more for all Club Nationals (M &amp; W) back to 2004</em></p><p>So I took a deep dive into the numbers of Club Nationals back to 2004 to see how the year 2017 stacked up. The data reveal several statements about the 2017 Club Nationals tournament.</p><ol><li>The Ozone upset over Riot was the biggest upset in an elimination game for both Women’s and Men’s dating back to 2004.</li><li>Doublewide’s performance is the biggest overachievement for a team at Club Nationals when considering seed and making finals as the measurement.</li><li>Sockeye’s early exit is the biggest underperformance for a team at Club Nationals when considering seed, Elo rating, and tournament structure.</li><li>Fury’s championship came from the Women’s team with the highest seed and lowest Elo rating.</li><li>Machine’s 15–0 win over Patrol in a placement game was the only shutout in any Club Nationals Men’s or Women’s game.</li></ol><p>Let’s dig into the numbers.</p><h4>Biggest Upsets (Bracket Play)</h4><p><em>Top 10 upsets as measured by lowest expected Elo win probability</em></p><p>In case you were wondering, the Ozone upset of Riot this year is the biggest upset per Elo ratings for both divisions dating back to 2004. And it is by a wide margin. Also, Fury’s upset of Brute Squad comes in at #10. On the Men’s side, GOAT’s upset of Revolver in 2014 takes the top spot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Agoa4aa0K_c0-DUcEM09Fg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2xmZlHxx4HssXD4sU7UiDw.png" /></figure><h4>Biggest Upsets (Pool Play)</h4><p><em>Top 10 upsets as measured by lowest expected win probability of winning team</em></p><p>Florida United’s upset of Sockeye this year comes in at #7 on the list for the Men’s division. Sockeye has been on the losing end of two of the top 3 upsets in pool play as well.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AOUaU527kx37SJQn8MwJcA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wuu6uRAKUUM_yZa5L6SZbQ.png" /></figure><h4>Greatest Match-ups</h4><p><em>Top 10 strongest match-ups as measured by combined Elo ratings between opponents</em></p><p>The final between Fury and Brute Squad comes in as the #8 strongest matchup in the Women’s division. No Men’s division games crack the top 10 dating back to 2012. I think this is due to the parity in the division today compared to years past when there were a few dominant teams in the division.</p><p>Note that there is some Elo rating inflation as the years have progressed, so it is hard to compare absolute ratings from today to years past. This is more so the case in the Women’s division versus the Men’s. Keep this in mind when comparing Furious vs Sockeye 2006 to Revolver vs Ironside 2011 or Fury vs Riot 2008 to Riot vs Brute Squad 2016.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Rza5el-xHjdptDwlr_OmDw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*T_0tCK-kBV5JvJFeYWHTrQ.png" /></figure><h3>OVERACHIEVERS</h3><h4>Highest Seeds — Made Semis</h4><p><em>9-Seed or higher teams that made semis. Elo rating is team rating entering tournament.</em></p><p>Only Ring of Fire in 2016 had a higher seed for a semis team than Doublewide in 2017.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xj9RNVvLeKN_NVbmV-kxew.png" /></figure><h4>Highest Seeds — Made Finals</h4><p><em>3-Seed or higher teams that made finals. Elo rating is team rating entering tournament.</em></p><p>Doublewide is the highest seed by far for both divisions to make the finals of Club Nationals. Fury is one of only five 4-seeds to make the finals in the Women’s division.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7j7PQ9736bjysDxCA3qsDw.png" /></figure><h4>Highest Seeds — Won Nationals</h4><p><em>2-Seed or higher teams that won Nationals. Elo rating is team rating entering tournament.</em></p><p>Fury is one of only two 4-seeds to win Club Nationals in the Women’s division. The last one was Scandal back in 2014. Fury entered 2017 Nationals with a lower Elo rating than Scandal did in 2014.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RlysO1CRbLSb5EDJMlTbuA.png" /></figure><h3>UNDERPERFORMERS</h3><h4>Lowest Seeds — Missed Quarters</h4><p><em>Lowest seeded teams per division that missed quarters. Elo rating is team rating entering tournament.</em></p><p>Sockeye 2017 was one of three 2-seeds in either division to not make quarters at Nationals in the last 14 years. There is a case that Sockeye’s performance is the worst due to the tournament structure in 2017 compared to 2014 and 2013. In 2013, GOAT took 3rd in their pool then lost to the tournament’s 3-seed Doublewide in prequarters. In 2014, Machine went 3–0 in their pool before losing to Ring of Fire in prequarters. Ring was one marathon point away from making finals that year (<em>Note: I was a player and captain on that 2014 Machine team… sigh).</em></p><p>Sockeye this year went 0–3 in pool play and missed out on prequarters all together. Ouch.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gS7AfYL4TJ-FJ4lh78RlXg.png" /></figure><h4>Lowest Seeds — Missed Semis</h4><p><em>Lowest seeded teams per division that missed semis. Elo rating is team rating entering tournament.</em></p><p>Unfortunately the Seattle troubles continue here. Riot is the only 1-seed in the Women’s division to not make semis. In addition, they have the highest Elo rating of any team not making semis in either division.</p><p>On the Men’s side, Sockeye is one of only four 2-seeds to not make Semis. Three of these four were highlighted in the previous discussion.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7qWDinDx_9lDw_9pl24NrA.png" /></figure><h3>OTHER</h3><h4>Shutout</h4><p>Machine’s 15–0 win over Patrol was the only recorded shutout in the Men’s or Women’s divisions at Club Nationals dating back to 2004. There is no Score Reporter data before 2004, so it’s unclear if this has happened before 2004. There have been four Women’s games that finished 15–1 at Club Nationals. The most recent one was Fury beating Hot Metal 15–1 in pool play in 2012.</p><h4>Actual vs. Predicted Outcomes 2017 — Elo Ratings</h4><p>And finally for the curious ones (aka nerds) out there, here are how my Elo ratings did predicting individual games at Club Nationals 2017. A small sample size but still fairly close with the Men’s division and the Elo system overpredicted the number of upsets in the Women’s division. Note that placement games were not included in the data below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/428/1*72BJiHrAc6XDKrSBXGHcIA.png" /></figure><p>Finally, let me know if there are any other ways you would like to see the data cut. I’m always open to ideas about how to analyze these results.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=deaabacf5ba3" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/by-the-numbers-club-nationals-2004-2017-deaabacf5ba3">By The Numbers: Club Nationals 2004–2017</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[End of 2017 Club Elo Ratings]]></title>
            <link>https://medium.com/the-flip/end-of-2017-club-elo-ratings-ef222c437134?source=rss-b17925dc1923------2</link>
            <guid isPermaLink="false">https://medium.com/p/ef222c437134</guid>
            <category><![CDATA[stats]]></category>
            <category><![CDATA[ultimate-frisbee]]></category>
            <dc:creator><![CDATA[Craig P]]></dc:creator>
            <pubDate>Wed, 25 Oct 2017 02:29:22 GMT</pubDate>
            <atom:updated>2017-10-25T02:36:04.766Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Men’s and Women’s Divisions updated after Club Nationals</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/340/1*qf9HkTmK-wDveQAvNNK_fw.jpeg" /></figure><p>While the sock lines are still fresh and the voices are still hoarse from Sarasota, it’s time to put to rest the 2017 club season. Here are the final Elo ratings for each team in the Men’s and Women’s divisions.</p><p><em>Note that very little weight is put on placement games at Nationals for these Elo ratings. Ratings mostly impacted by Pool and Bracket games. Only individual game results impact ratings as opposed to where a team finishes in the tournament. Check out this </em><a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41"><em>article</em></a><em> explaining the rating system if you’re curious.</em></p><h4>Men</h4><p>Revolver heads and shoulders above the competition as we would expect. Doublewide climbs 160 points to 4th in the rating system. Sockeye closes the season almost 200 points lower than where they entered Nationals.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jzhiKtWnBus26RnAsEPCPg.png" /></figure><h4>Women</h4><p>Despite the result of the double-game point Finals, Brute Squad still retains the top spot in the Elo ratings. Ozone sees a big jump up from their huge upset of Riot. The ratings of the top 5 teams look pretty close to what they were a year ago. Not much turnover in the division.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bYGuS4dAqSI9eZm9B-KxMA.png" /></figure><p>For a background on how I created these ratings, check out this <a href="https://medium.com/the-flip/introducing-club-ultimate-elo-ratings-v-1-0-mens-49baaff28c41">article</a>. Also, check out historical Elo ratings for club teams back to 2004 for both <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-mens-85343abfd15d">men’s</a> and <a href="https://medium.com/the-flip/historical-elo-ratings-by-team-womens-95b4fc9355bb">women’s</a>.</p><p><em>Please note that only teams who have made Nationals in their history are included in these ratings and only games between these teams affect Elo ratings. Therefore there are teams that are excluded but are probably true Top 24 teams in the club game today. Also, I don’t advocate using these rankings to allocate Nationals bids. Just another way to measure true team strength empirically in addition to the USAU rankings. Teams with few games against other teams in the rating system may not have their rating correct to their current team strength.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ef222c437134" width="1" height="1" alt=""><hr><p><a href="https://medium.com/the-flip/end-of-2017-club-elo-ratings-ef222c437134">End of 2017 Club Elo Ratings</a> was originally published in <a href="https://medium.com/the-flip">The Flip</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>