https://kubeia.io/sports_analytics_weekly/ Your weekly serving of sports analytics insights. en-us Copyright kubeia.io social@kubeia.io social@kubeia.io Mon, 06 Jul 2026 05:00:07 +0000 Mon, 06 Jul 2026 05:00:07 +0000 Python https://validator.w3.org/feed/docs/rss2.html 1440 https://kubeia.io/favicon-32x32.pnghttps://kubeia.io https://kubeia.io/sports_analytics_weekly/2026-25.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.expectinggoals.com/p/world-cup-update-june-18-england" target="_blank">World Cup Update June 18: England, Colombia, and Explaining Adjusted xG</a> </h4> <p class="source_author ml-1 mr-1"> expectinggoals.com - Michael Caley</p> <blockquote> <p class='mb-2'>Uzbekistan is credited with 0.6 “Adjusted xG” while Opta’s published xG rates their chances at 1.2 xG. While my xG system will differ from others at times,1 this is not the case here. In unadjusted xG I had Uzbekistan with 1.1 xG. </p><p class='mb-2'> </p><p class='mb-2'>The reason that adjusted xG differs so much from published xG here is that Uzbekistan’s equalizing goal was a header into an open net by Abbosbek Fayzullaev off a saved shot from wide in the penalty area. The chance is credited at 0.98 xG by Opta’s system and 0.92 xG by mine.2 If the job of xG is to tell you how many goals are likely to be scored in the match, based on the shooting chances, then this is correct. Uzbekistan were going to score a goal in this match barring an Eric Choupo-Moting situation. </p><p class='mb-2'> </p><p class='mb-2'>But if the job of xG is to evaluate the quality of the two teams in the match and how well they played, that massive xG is misleading.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">👁️ Computer Vision </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://github.com/TrackNetV4/TrackNetV4" target="_blank">TrackNetV4/TrackNetV4</a> </h4> <p class="source_author ml-1 mr-1"> github.com</p> <blockquote> <p class='mb-2'>This repository contains the official TensorFlow implementation of our ICASSP 2025 paper: TrackNetV4: Enhancing Fast Sports Object Tracking with Motion Attention Maps.We provide TensorFlow code demonstrating our Motion-Aware Fusion framework, integrated with TrackNetV2, for training, testing, and visual prediction of sports object trajectories. Our implementation supports three datasets: tennis, badminton, and newly introduced badminton dataset. Additionally, we provide model code for integrating the framework with TrackNetV3.Additionally, feel free to explore the project website for more visualizations, dataset details, and experimental results.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🏛️ Economics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.hulltactical.com/2026/06/17/the-world-cup-and-equity-markets/" target="_blank">The World Cup and Equity Markets</a> </h4> <p class="source_author ml-1 mr-1"> hulltactical.com</p> <blockquote> <p class='mb-2'>No-one now really believes humans are rational, either individually or in aggregate. And the entire field of behavioral finance is predicated on this fact. </p><p class='mb-2'> </p><p class='mb-2'>The effects that relate to the market have often been strange. Researchers have linked stock returns to holidays, Netflix, the doomsday clock, cloud cover, seasonal affective disorder, and lunar cycles. Much of this literature is entertaining but not especially convincing. Many of the effects are small, difficult to replicate, or vulnerable to accusations of data mining. </p><p class='mb-2'> </p><p class='mb-2'>Then there is football (“soccer” if you insist on being wrong). </p><p class='mb-2'> </p><p class='mb-2'>In 2007, Alex Edmans, Diego Garcia, and Øyvind Norli published one of the most famous papers in behavioral finance: Sports Sentiment and Stock Returns. Their findings were simple, surprising, and difficult to explain away. When a national soccer team loses an important match, the country’s stock market tends to fall the next trading day.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-25.html Mon, 22 Jun 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-24.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://link.springer.com/chapter/10.1007/978-3-032-16645-6_5" target="_blank">An Inverse Problem of Social Influence for Spatial Reference Positions in Soccer</a> </h4> <p class="source_author ml-1 mr-1"> springer.com - Ulrik Brandes</p> <blockquote> <p class='mb-2'>We present a complex network approach to depict representative spatial arrangements of association football (soccer) teams. It is designed to mitigate issues that arise with average locations, which are currently the most common choice. Our approach uses proximity networks, affiliation networks, and a network model of social influence to reduce spatial concentration bias and amplify the main positioning signals. Example applications include the display of enacted formations, and the placement of players in passing networks.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.cannonstats.com/p/find-a-players-statistical-twin" target="_blank">Find a Player's Statistical Twin</a> </h4> <p class="source_author ml-1 mr-1"> cannonstats.com - Scott Willis</p> <blockquote> <p class='mb-2'>Explaining and introducing the Cannon Stats Similar Player Tool</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.expectinggoals.com/p/world-cup-projections-introducing" target="_blank">World Cup Projections: Introducing PADDLIN'</a> </h4> <p class="source_author ml-1 mr-1"> expectinggoals.com - Michael Caley</p> <blockquote> <p class='mb-2'>This model seeks an answer to all the problems that bedevil any objective measurement of international team quality. It uses an Elo rating based on actual results and, where available, an xElo rating based on the projected result of adjusted expected goals, to get a first view of team quality that can be compared among teams around the world. The model does in fact increase the weights of matches in which a team put up a significant margin in goal difference or xG difference, actually incorporating the question of whether one team or another did suffer a paddlin’. It makes use of player values from Transfermarkt.com to adjust these Elo ratings, giving teams credit for having better players available beyond the effects they may have had on results in the past. While these crowd-sourced player ratings are far from definitive, work by Paul Johnson among others has demonstrated their utility for statstical projections.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-24.html Mon, 15 Jun 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-23.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.cannonstats.com/p/introducing-the-cannon-stats-finishing" target="_blank">Introducing the Cannon Stats Finishing Skill Model</a> </h4> <p class="source_author ml-1 mr-1"> cannonstats.com - Scott Willis</p> <blockquote> <p class='mb-2'>Moving beyond the simple Goals - xG and hope for the best</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://dtai.cs.kuleuven.be/sports/blog/exploring-how-vaep-values-actions/" target="_blank">Exploring how VAEP values actions</a> </h4> <p class="source_author ml-1 mr-1"> kuleuven.be - Pieter Robberechts</p> <blockquote> <p class='mb-2'>We created an interactive tool to explore how VAEP values player actions in soccer. Try it out!</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-23.html Mon, 08 Jun 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-22.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://dtai.cs.kuleuven.be/sports/blog/tabular-foundation-models-for-xg:-can-tabpfn-score-without-training/" target="_blank">Tabular Foundation Models for xG: Can TabPFN Score Without Training?</a> </h4> <p class="source_author ml-1 mr-1"> kuleuven.be - Jesse Davis, Pieter Robberechts, Timo Martens</p> <blockquote> <p class='mb-2'>Our results suggest that tabular foundation models are an exciting avenue for sports analytics. We found it impressive (and surprising) that the out-of-the-box TabPFN model could marginally outperform a finely tuned XGBoost model. Moreover, TabPFN’s ability to make accurate predictions using only a fraction of the historical data means that analysts could theoretically generate bespoke, highly accurate models for lower-tier leagues, or specific tactical setups using only a handful of matches, rather than waiting years to collect a significant sample size. </p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-22.html Mon, 01 Jun 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-21.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.cannonstats.com/p/do-radar-plots-need-to-die" target="_blank">Do radar plots need to die?</a> </h4> <p class="source_author ml-1 mr-1"> cannonstats.com - Scott Willis</p> <blockquote> <p class='mb-2'>Radar plots become an industry norm after being popularized by Ted Knutson formerly of StatsBomb. They are pretty to look at, and they can (and I believe do) give you a good quick look at a player. </p><p class='mb-2'>They aren’t perfect by any means, no graphic is and this was pointed out in a post from PyMC-Labs called “Radar Plots Must Die.” </p><p class='mb-2'> </p><p class='mb-2'> </p><p class='mb-2'> </p><p class='mb-2'> </p><p class='mb-2'>I use radars, and I plan on continuing to use them going forward and this made me want to directly address a few of the things brought up and what I do to address them or correct some things aren’t quite right.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.natesilver.net/p/pele-international-football-rankings-soccer-ratings-projections" target="_blank">PELE International Football Rankings ⚽</a> </h4> <p class="source_author ml-1 mr-1"> natesilver.net - Nate Silver</p> <blockquote> <p class='mb-2'>What happens when you blend 150+ years of soccer history and player market values into a brand-new model? You get PELE, our insanely detailed, predictive rating system for all 211 FIFA teams.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/does-xg-predict-results-better-than" target="_blank">Does xG Predict Results Better Than EPV?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The paper’s stated contribution is therefore to compare xG and EPV in both settings, asking whether shot-based information or possession-based information better predicts match outcomes when used before and after the game.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.pymc-labs.com/blog-posts/radar-plots-must-die" target="_blank">Football Radar Charts Are Misleading, A Better Way to Visualize Player Data</a> </h4> <p class="source_author ml-1 mr-1"> pymc-labs.com - Chris Fonnesbeck</p> <blockquote> <p class='mb-2'>Radar plots are everywhere in football (soccer) analytics. StatsBomb, The Athletic, FBref, and most analytics accounts on social media use them, and they arrive in my inbox through various football newsletters I subscribe to. They are eye-catching and compress a lot of metrics into one distinctive shape. But they are also, on reflection, a remarkably poor way to read a player: the polygon the reader perceives is driven more by arbitrary choices the analyst made than by the underlying numbers. The problems are not new, but this post summarizes them, surveys what the visualisations that do work have in common, and proposes a replacement built from those principles.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/tabular-ml-is-entering-a-new-benchmark" target="_blank">Tabular ML is entering a new benchmark era</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>From static and narrow benchmarks to live, capability-driven evaluation</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-21.html Mon, 25 May 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-20.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-026-what-actually-gets-teams-promoted" target="_blank">TI #026: What Actually Gets Teams Promoted?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Which technical performance indicators actually increase the chances of promotion to the elite leagues?</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/what-makes-a-pass-well-timed" target="_blank">What Makes a Pass Well Timed?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Quantifying optimal pass timing with tracking data, OBSO, and the PAUSA framework.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-027-the-smallest-margins-often" target="_blank">TI #027: The Smallest Margins Often Reveal the Clearest Football Signals</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The research examines "close matches" in the German Bundesliga where the final goal difference was one or zero to identify what separates winners from losers.Researchers analyzed nearly 8,000 individual player observations over three seasons, categorizing data into five specific positional roles from central defenders to forwards.The study highlights how minute statistical variations in technical and physical performance become the deciding factors when tactical dominance between two teams is roughly equal.Findings reveal that specific high-leverage actions within these positions—such as a midfielder's passing precision or a defender's positioning—carry disproportionate weight in tight games.Ultimately, the article demonstrates that at the elite level, the smallest margins in individual positional efficiency are what consistently dictate the outcomes of the league's most competitive fixtures.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">👁️ Computer Vision </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://learnopencv.com/vision-banana-explained/" target="_blank">Vision Banana: How Image Generators Are Becoming Powerful Vision Models</a> </h4> <p class="source_author ml-1 mr-1"> learnopencv.com - Satya Mallick</p> <blockquote> <p class='mb-2'>Vision Banana turns Nano Banana Pro into a powerful vision model for segmentation, depth estimation, surface normals, image generation, and editing.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/making-tabular-foundation-models" target="_blank">How to make Tabular Foundation Model inference faster</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>The greatest bottleneck with tabular foundation models: Prediction, aka inference, is slow. </p><p class='mb-2'> </p><p class='mb-2'>This post is a collection of tips and tricks to make tabular foundation models much faster. BUT! There is always a price to pay. And you must decide on that bargain. Some improvements are cheaper, some are more expensive.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🧮 Statistics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://valeman.medium.com/if-youre-still-worshipping-pearson-correlation-you-re-not-a-data-scientist-you-re-driving-a-831dc0590de6" target="_blank">If You’re Still Worshipping Pearson Correlation, You’re Not a Data Scientist — You’re Driving a Horse Cart in the Age of AI</a> </h4> <p class="source_author ml-1 mr-1"> medium.com - Valeriy Manokhin, PhD, MBA, CQF</p> <blockquote> <p class='mb-2'>If Pearson correlation is still your default tool for understanding relationships in data, you are not doing modern data science. You are doing statistical sightseeing. </p><p class='mb-2'> </p><p class='mb-2'>Pearson correlation is not useless. That is not the point. </p><p class='mb-2'> </p><p class='mb-2'>The point is that too many people use it like a universal detector of truth, when in reality it is a narrow tool with very specific assumptions and a very long list of weaknesses.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-20.html Mon, 18 May 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-19.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-025-tracking-data-reveals-the" target="_blank">Tracking Data Reveals the Tactical Behaviours That Win Matches?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Width. Depth. Mobility. Support. Superiority. Compactness. </p><p class='mb-2'> </p><p class='mb-2'>Can we translate those principles into measurable features, and do those features actually relate to winning? </p><p class='mb-2'> </p><p class='mb-2'>Using tracking data from 302 professional matches, the authors build a set of features that represent different attacking principles. They then test whether those features can predict match outcomes, both across full matches and within smaller windows of time.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/what-do-players-really-see-on-the" target="_blank">What Do Players Really See on the Pitch?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Using Pose Data, Field-of-View Models, and Pitch Control to Measure Visual Awareness in Soccer.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/disruption-maps-reveal-how-teams" target="_blank">Disruption Maps Reveal How Teams Break The Best Attacks</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>A Graph Neural Network framework combining xReceiver, xPass, and xThreat to quantify player availability, defensive impact, and how teams disrupt attacking decisions in real time</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-19.html Mon, 11 May 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-17.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">🎲 Betting </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://muddywatersresearch.com/research/2026/mw-is-short-srad/" target="_blank">We are short Sportradar Group AG.</a> </h4> <p class="source_author ml-1 mr-1"> muddywatersresearch.com - Muddy Waters Research</p> <blockquote> <p class='mb-2'>At the ICE 2026 gaming conference in Barcelona, our investigators posed as operators of a startup sportsbook. We told SRAD’s sales team — repeatedly and explicitly — that our target markets were Vietnam, Thailand, Indonesia, and China. Every one of these countries bans online gambling. Not one SRAD salesperson told us no. Instead, an Asia-focused sales executive walked us through product offerings tailored to each illegal market, bragged that SRAD “serves everyone,” and offered to introduce us to the Yabo Group — China’s largest illegal gambling operator, whose Cambodian call centers are staffed by trafficked and enslaved workers. He warned us that Yabo’s people didn’t attend ICE because “they would be hunted down.” Then he offered to make the introduction anyway. This is who SRAD is.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.expectinggoals.com/p/the-origins-of-the-set-piece-revolution?hide_intro_popup=true" target="_blank">The Origins of the Set Piece Revolution</a> </h4> <p class="source_author ml-1 mr-1"> expectinggoals.com - Michael Caley</p> <blockquote> <p class='mb-2'>The set piece revolution remains the story of the season in the Premier League and shows no signs of slowing down. Goals from set pieces are still elevated. Corner kicks and long throws continue to account for more or less the entirety of this effect. Since the last Expecting Goals newsletter pinpointed these two tactics as the core of the new set piece vision, more discussions and analyses have focused on these situations.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/the-perfect-xg-model-does-not-exist" target="_blank">The Perfect xG Model Does NOT Exist. Does it?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Expected Goals (xG) models have become a central tool in football analytics, widely adopted across broadcasting, coaching, and performance analysis. The paper highlights that xG is now “a standard part of match facts” and increasingly used by coaches such as Mikel Arteta and Arne Slot to evaluate performance. Among providers, Hudl-StatsBomb positions its model as the “most accurate xG model”, which motivates the authors to critically examine this claim.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-023-what-teams-do-10-seconds-before" target="_blank">TI #023: What Do Teams Do 10 Seconds Before a Goal</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Football analytics has become increasingly sophisticated at measuring shots. We can estimate chance quality, compare finishing skill, model goalkeeper positioning, and understand which teams consistently outperform their xG totals. All of that has genuine value. </p><p class='mb-2'> </p><p class='mb-2'>But goals are rarely created at the instant a player strikes the ball. </p><p class='mb-2'> </p><p class='mb-2'>They are usually built in the seconds beforehand: a forward dragging a centre-back out of position, a midfielder receiving between lines for one touch too long, a full-back arriving unnoticed on the blind side, or a defensive line shifting half a second later than it needed to. </p><p class='mb-2'> </p><p class='mb-2'>The finish is visible. The construction is often hidden.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-024-why-football-is-a-system-not" target="_blank">TI #024: Why Football Is a System, Not a Formation</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>For decades, tactical analysis in football was built on observation. Coaches, scouts, and analysts watched the game, interpreted shapes, and described what they believed they saw. One team looked compact. Another controlled midfield. A striker occupied defenders well. A back line was too deep. Much of it was insightful, but much of it also lived in language rather than evidence. </p><p class='mb-2'> </p><p class='mb-2'>Then tracking data changed the landscape. </p><p class='mb-2'> </p><p class='mb-2'>Once every player could be located multiple times per second, football gained something it had never fully possessed before: a measurable map of collective behaviour. Suddenly, spacing could be quantified. Synchronisation could be tested. Defensive reactions could be timed. Tactical analysis no longer had to rely purely on description. </p><p class='mb-2'> </p><p class='mb-2'>The authors published it at a moment when the game was beginning to move from intuition-led tactical commentary toward data-supported tactical understanding. Long before tracking departments became standard, before pressing metrics entered mainstream discourse, and before clubs openly discussed machine learning, the authors asked a question that still sits at the centre of modern football analytics: </p><p class='mb-2'> </p><p class='mb-2'>What if tactics could be measured through movement patterns rather than explained after the fact? </p><p class='mb-2'> </p><p class='mb-2'>It sounds simple now. At the time, it was a statement about where the game was heading.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">👁️ Computer Vision </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://learnopencv.com/yolo26-pose-estimation-tutorial/" target="_blank">YOLO26 Pose Estimation: Real-Time Keypoint Tutorial</a> </h4> <p class="source_author ml-1 mr-1"> learnopencv.com - Sudip Chakrabarty</p> <blockquote> <p class='mb-2'>Human pose estimation has become a cornerstone of modern computer vision, powering applications from fitness tracking apps and sports analytics to gesture-based interfaces and medical rehabilitation. At its core, keypoint estimation is the task of detecting specific anatomical landmarks on the human body, the nose, shoulders, elbows, wrists, hips, knees, and ankles, and connecting them into a skeleton that represents the person’s pose. </p><p class='mb-2'> </p><p class='mb-2'>In September 2025, Ultralytics announced YOLO26, the next-generation YOLO model optimized for edge computing, robotics, and mobile AI. Among its specialized task heads, YOLO26-pose brings several architectural innovations to keypoint estimation: Residual Log-Likelihood Estimation (RLE) for more accurate keypoint localization, end-to-end NMS-free inference for simpler deployment, and the MuSGD optimizer for more stable training dynamics. In this guide, we walk through the theory, architecture, benchmarks, and a hands-on implementation of YOLO26 keypoint estimation on images and videos.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/context-is-the-new-training" target="_blank">Context is the new training</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>Tabular foundation models such as TabPFN and TabICL don’t need to be trained to perform regression or classification. What they do is called in-context learning. What used to be the training data now becomes the context data at prediction time. </p><p class='mb-2'> </p><p class='mb-2'>This post explores the idea of context data and contrasts it with “classic” training data. Does moving from training data to context data change how we model? Does it enable something new? </p><p class='mb-2'> </p><p class='mb-2'>Let’s dive in.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🎙️ Podcast </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://smartbettingclub.com/blog/episode-100-sbc-podcast-ant-de-rosa-on-sharp-betting-pinnacle-and-market-execution/" target="_blank">Episode 100 SBC Podcast: Ant De Rosa on Sharp Betting, Pinnacle, and Market Execution</a> </h4> <p class="source_author ml-1 mr-1"> smartbettingclub.com - Peter Ling</p> <blockquote> <p class='mb-2'>In Episode 100 of the SBC Podcast, I welcomed a special guest to mark the milestone, in the form of Antonino (Ant) De Rosa, a former Pinnacle trader and now operator of a large scale professional betting group. </p><p class='mb-2'> </p><p class='mb-2'>This episode goes deep into the reality of sharp betting. Not theory, not models in isolation, but how markets actually move, how sportsbooks react, and why execution is often the biggest edge in the modern game. </p><p class='mb-2'> </p><p class='mb-2'>Ant shares his unique journey from elite level Magic: The Gathering player to being recruited by Pinnacle, where he specialised in live NBA trading. He explains how his edge was never about knowing sport better than others, but about predicting behaviour, understanding where the next bet would come from, and how the market would react. </p><p class='mb-2'> </p><p class='mb-2'>The conversation then shifts into the mechanics of running a serious betting operation today. </p><p class='mb-2'> </p><p class='mb-2'>From scaling across multiple sports, managing hundreds of accounts, and handling real world challenges like liquidity, restrictions, and non payment. Ant is clear that having a strong model or idea is not enough, if you cannot get money down, the edge has no value. </p><p class='mb-2'> </p><p class='mb-2'>This is a rare, honest look at the sharp end of betting in 2026, where success comes from combining insight, discipline, and the ability to execute at scale.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://smartbettingclub.com/blog/phil-smith-on-antepost-betting-betfair-removal-modern-betting-challenges-sbc-podcast-episode-101/" target="_blank">Phil Smith on Antepost Betting, Betfair Removal</a> </h4> <p class="source_author ml-1 mr-1"> smartbettingclub.com - Peter Ling</p> <blockquote> <p class='mb-2'>In Episode 101 of the SBC Podcast, I’m joined by Phil Smith, a full time football bettor specialising in antepost (futures) markets and exchange trading.Phil shares his journey from early losses and trial and error, through to building a profitable approach combining Betfair trading and more recently, antepost betting.We go deep into how antepost betting works in practice, where the real edge comes from, and why volume, structure and discipline are key to long term success.The conversation also explores the growing challenges facing bettors today, including Phil’s own experiences, from being permanently removed from Betfair after years of activity, to being locked out of funds and bets for months during an affordability check with Sky Bet.This is a practical, honest discussion about modern betting, where theory often takes a back seat to execution, access and adaptability.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-17.html Mon, 27 Apr 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-16.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-022-why-a-role-change-can-completely" target="_blank">Why a role change can completely reshape a player's performance</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>What happens when the same player performs differently… simply because his position changes? </p><p class='mb-2'> </p><p class='mb-2'>At first, the answer seems obvious. Of course a fullback runs more than a centre back, and of course a wide midfielder behaves differently from a central one. Positional differences are one of the most established findings in football analytics. </p><p class='mb-2'> </p><p class='mb-2'>But that is not the most interesting question.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/can-clubs-capture-player-evolution" target="_blank">Can Clubs Capture Player Evolution in Football?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The paper begins by situating football analytics within the rise of probabilistic models such as expected goals (xG), which assign likelihoods to actions like shots based on contextual features. These models typically rely on event data and include predictors such as shot distance, angle, body part, and more recently contextual elements like defender positioning or goalkeeper location.However, a key limitation is highlighted: most models do not explicitly include the player as a predictor. As the authors note, “two separate shots that have the same measures for the model predictors will be assigned the exact same xG regardless of who is taking the shot”. This omission contradicts the fundamental assumption that players differ in skill and execution ability.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/the-hidden-cost-of-midweek-matches" target="_blank">The Hidden Cost of Midweek Matches</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Over the past decade, elite football has experienced a steady increase in physical demands, largely driven by the intensification of competitions and the growing number of matches within short timeframes. Congested fixture periods, typically defined as matches played with less than 96 hours of recovery, have become increasingly common and are associated with fluctuations in external load. </p><p class='mb-2'> </p><p class='mb-2'>These periods impose both physiological and psychological strain, with top teams sometimes playing up to six matches in 18 days. This accumulation of load is linked to neuromuscular fatigue, reduced recovery, and increased injury risk. While prior research has documented general declines or adaptations in physical output during congested schedules, it has largely treated these periods as homogeneous, without considering the type of competition involved. </p><p class='mb-2'> </p><p class='mb-2'>The paper highlights that different competitions, such as domestic leagues, national cups, and the UEFA Champions League, may impose distinct contextual demands. These include differences in opponent quality, travel, tactical priorities, and psychological pressure. As a result, understanding external load requires a more granular approach that integrates both competition type and individual playing time.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.americansocceranalysis.com/home/2026/4/15/throw-in-it-back" target="_blank">Throw-in It Back</a> </h4> <p class="source_author ml-1 mr-1"> americansocceranalysis.com - Ben Bellman</p> <blockquote> <p class='mb-2'>Whether you love long attacking throw-ins or hate them, there is no denying that they’ve become both a key feature and flashpoint in men’s soccer in the past year. John Muller likely sparked a renaissance of the tactic (and a soon-to-be Arsenal title) with his 2023 article for The Athletic, and Joe Lowery and I borrowed his method for Backheeled when Minnesota United started longthrowmaxxing in 2025 (Editor’s note: Minnesota work with Mike Imburgio through ASA’s firewalled consulting arm). But while each game has about 40 throw-ins on average, only about 10 of those throws happen close enough to reach the box. But apart from Formerly Called Twitter jokes about consultant Thomas Grønnemark, there hasn’t been much commentary about all the other ones in popular media or public analytics circles. The only exceptions I’m aware of are Eliot McKinley’s 2018 two-part opus on this very website, and some recent academic work on the top 5 European leagues that, if you like in-text citations and interpreting regressions, is an excellent spoiler for the rest of this article.Eliot did that work almost a decade ago (before Game of Thrones jumped the shark), and I thought it was time to replicate and extend those findings with all the amazing infrastructure that ASA has built since the days of CSV files on Dropbox. In addition to models estimating throw completions and retained possession, I also analyze the goals added for possessions following throws to assess the value of throw choices. This allows me to find the MLS throw-in MVPs and offer an expanded set of (very general) rules for approaching these overlooked moments of play.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🌩 Forecasting </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.argmin.net/p/calibrated-games" target="_blank">Calibrated Games</a> </h4> <p class="source_author ml-1 mr-1"> argmin.net - Ben Recht</p> <blockquote> <p class='mb-2'>One of the main uses of simulation and forecasting in designed feedback systems is for deciding how to act. If I can map what will happen next, I can choose actions that steer me toward good outcomes. This mindset seems perfectly sensible, and it’s the backbone of statistical decision theory, tree search in game play, optimal control, and model predictive control. Moreover, people who are good at prediction get clout. You can even win money in markets. It seems like forecasting is a skill and talent, and one that requires deep knowledge of how the world works. And yet, in class on Monday, I discussed how you can make excellent forecasts by simple, strategic accounting.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-16.html Mon, 20 Apr 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-15.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://arxiv.org/abs/2601.19944" target="_blank">Classifier Calibration at Scale: An Empirical Study of Model-Agnostic Post-Hoc Methods</a> </h4> <p class="source_author ml-1 mr-1"> arxiv.org - Valery Manokhin, Daniel Grønhaug</p> <blockquote> <p class='mb-2'> Abstract:We study model-agnostic post-hoc calibration methods intended to improve probabilistic predictions in supervised binary classification on real i.i.d. tabular data, with particular emphasis on conformal and Venn-based approaches that provide distribution-free validity guarantees under exchangeability. We benchmark 21 widely used classifiers, including linear models, SVMs, tree ensembles (CatBoost, XGBoost, LightGBM), and modern tabular neural and foundation models, on binary tasks from the TabArena-v0.1 suite using randomized, stratified five-fold cross-validation with a held-out test fold. Five calibrators; Isotonic regression, Platt scaling, Beta calibration, Venn-Abers predictors, and Pearsonify are trained on a separate calibration split and applied to test predictions. Calibration is evaluated using proper scoring rules (log-loss and Brier score) and diagnostic measures (Spiegelhalter's Z, ECE, and ECI), alongside discrimination (AUC-ROC) and standard classification metrics. Across tasks and architectures, Venn-Abers predictors achieve the largest average reductions in log-loss, followed closely by Beta calibration, while Platt scaling exhibits weaker and less consistent effects. Beta calibration improves log-loss most frequently across tasks, whereas Venn-Abers displays fewer instances of extreme degradation and slightly more instances of extreme improvement. Importantly, we find that commonly used calibration procedures, most notably Platt scaling and isotonic regression, can systematically degrade proper scoring performance for strong modern tabular models. Overall classification performance is often preserved, but calibration effects vary substantially across datasets and architectures, and no method dominates uniformly. In expectation, all methods except Pearsonify slightly increase accuracy, but the effect is marginal, with the largest expected gain about 0.008%. </p> </blockquote> </div> <h3 class="category mb-3 mt-5">🕰️ Blast From the Past </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.youtube.com/watch?v=iN-QqnbrDYA" target="_blank">Radar Wars - CASSIS Presentation Summer 2018 - YouTube</a> </h4> <p class="source_author ml-1 mr-1"> youtube.com - Ted Knuston</p> <blockquote> <p class='mb-2'>This is my keynote presentation from the CASSIS conference. It focuses on different considerations you have when creating sports data visualisations, while poking a tiny bit of fun at Luke Bornn and Daryl Morey (and myself). </p><p class='mb-2'></p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-15.html Mon, 13 Apr 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-13.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/who-deserves-credit-for-good-defending" target="_blank">Who Deserves Credit for Good Defending?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>This paper starts from a simple but still unresolved problem in football analytics: defending remains much harder to value than attacking. Offensive models have evolved quickly because passes, shots, and goals naturally create measurable events. Defensive contribution is different because many of the most important actions never appear directly in event logs. A defender may force a backward pass, remove a dangerous lane, delay an attack, or make a threatening option disappear entirely without recording a tackle or interception. That is why the authors build the whole paper around Paolo Maldini’s famous line: “If I have to make a tackle, then I have already made a mistake.” </p><p class='mb-2'> </p><p class='mb-2'>The paper reviews how existing work still leaves important gaps. Metrics such as VAEP-style action valuation capture offensive actions well, while defensive methods often either work only at aggregate match level, focus only on explicit actions like tackles and interceptions, or evaluate team defense without distributing value to individuals. Even recent graph-based defensive models mostly study passing pressure qualitatively or restrict the analysis to pass prevention only. </p><p class='mb-2'> </p><p class='mb-2'>To address this, the authors introduce DEFCON (DEFensive CONtribution evaluator), a framework designed to assign defensive credit in every game situation, not only when a defender touches the ball. The key idea is to estimate all available attacking options at a given moment, quantify how dangerous they are, estimate how likely they are to succeed, and then determine how much each defender is responsible for suppressing or allowing those options. Defensive value is then defined as the reduction in opponent scoring potential before and after an action. In practical terms, defenders are rewarded when they lower expected attacking value, penalized when they allow dangerous progression, and also credited when they successfully force opponents toward less threatening choices rather than simply stopping the action altogether.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-13.html Mon, 30 Mar 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-12.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-020-when-changing-the-manager" target="_blank">TI #020: When Changing the Manager Changes Nothing</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Does changing manager actually improve performance, or do teams often recover anyway? </p><p class='mb-2'> </p><p class='mb-2'>That is what this paper set out to examine across fifteen seasons of Premier League football, using a methodology that is more demanding than the usual before-and-after comparison. Rather than simply measuring points gained after a dismissal, the authors compared each managerial change to a carefully matched counterfactual: a team in a similar competitive situation, with a similar recent performance trajectory, but without changing coach.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/can-football-models-separate-player" target="_blank">Can Football Models Separate Player Skill from Team Context?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The paper starts from one of the most persistent problems in football analytics: separating what belongs to the player from what belongs to the environment around him. A player’s observable production, whether measured through pass completion, xG contribution, or other event-based outputs, is always entangled with tactical context, teammate quality, opposition level, and game state. As the paper puts it, “a player’s observable metrics… are not a pure function of their individual ability”. This immediately creates a practical difficulty for recruitment and projection. A striker scoring regularly inside a dominant possession structure may not reproduce the same output elsewhere, while a midfielder whose numbers look modest could in fact be constrained by system effects.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">👁️ Computer Vision </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.youtube.com/watch?v=aBVGKoNZQUw" target="_blank">Football AI Tutorial: From Basics to Advanced Stats with Python -</a> </h4> <p class="source_author ml-1 mr-1"> youtube.com</p> <blockquote> <p class='mb-2'>Let's build a Football AI system to dig deeper into match stats! We'll use computer vision and machine learning to track players, determine which team is which, and even calculate stuff like ball possession and speed. This tutorial is perfect if you want to get hands-on with sports analytics and see how AI can take your football analysis to the next level.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/im-betting-on-tabular-foundation" target="_blank">I’m betting on tabular foundation models</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>All data modalities and tasks are occupied by foundation models. </p><p class='mb-2'> </p><p class='mb-2'>All? No! One small modality still holds out against them: tabular data. </p><p class='mb-2'> </p><p class='mb-2'>But this resistance is crumbling. </p><p class='mb-2'> </p><p class='mb-2'>TFMs are a fundamental shift, not just a performance trade-off </p><p class='mb-2'>TabPFN opened the era of foundation models for tabular data. For small and mid-sized data, tabular foundation models now outperform other ML algorithms (see TabArena).</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-12.html Mon, 23 Mar 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-11.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/not-every-league-demands-the-same" target="_blank">Not Every League Demands the Same Athlete</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>This study investigates how physical contexts differ across Europe’s top five leagues and how athletic ability is leveraged differently to generate player output. The authors develop models designed to project player performance when transferring between leagues based on physical metrics, age, team quality, and position. The central premise is that league environments impose distinct physical demands, and that these contextual differences materially influence performance translation.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/the-defensive-actions-behind-the" target="_blank">The Defensive Actions Behind the Passes That Never Happen</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The paper starts from a familiar limitation in football analytics: most defensive value remains invisible because many important defensive actions happen without direct contact with the ball. A defender may never register a tackle, interception, or clearance, yet still prevent danger simply by occupying the right space and discouraging a pass. The authors frame this around cover shadows, the defensive occupation of passing lanes that forces opponents away from valuable options. They argue that traditional event statistics fail precisely because they only record realized actions, while a successful cover shadow often removes an action before it exists.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-019-why-the-most-common-formation" target="_blank">TI #019: Why The Most Common Formation Was No Longer The Winning One</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>For years, football formations were treated like labels. </p><p class='mb-2'> </p><p class='mb-2'>4-4-2. </p><p class='mb-2'>4-3-3. </p><p class='mb-2'>4-2-3-1. </p><p class='mb-2'> </p><p class='mb-2'>A quick graphic before kickoff, a TV overlay, a line in the match report. </p><p class='mb-2'> </p><p class='mb-2'>But formations are rarely just shapes. They are tactical choices shaped by context, squad quality, physical demands, and even by what coaches believe the modern game now requires. </p><p class='mb-2'> </p><p class='mb-2'>And if you zoom out long enough, you are able to see interesting things. Not one formation replacing another, but an entire league slowly changing how it distributes risk, control, and attacking presence. </p><p class='mb-2'> </p><p class='mb-2'>That is what makes this LaLiga study particularly interesting. </p><p class='mb-2'> </p><p class='mb-2'>Rather than focusing on one team, one coach, or one season, it looked across 3,420 matches and 6,840 starting formations over nine seasons, asking a simple question: </p><p class='mb-2'> </p><p class='mb-2'>How has elite football actually changed structurally over time? </p><p class='mb-2'> </p><p class='mb-2'>And the answer is more nuanced than the usual “everyone moved from 4-2-3-1 to 4-3-3.”</p> </blockquote> </div> <h3 class="category mb-3 mt-5">💰 Quantitative Finance </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://substack.com/home/post/p-190257777" target="_blank">AI Will Create Millions of Quants</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Kris Longmore</p> <blockquote> <p class='mb-2'>Nice looking backtests are cheap now. </p><p class='mb-2'> </p><p class='mb-2'>This is worth sitting with for a moment. </p><p class='mb-2'> </p><p class='mb-2'>In the age of AI, a beautiful backtest proves almost nothing. </p><p class='mb-2'> </p><p class='mb-2'>The probability that some parameter combination produces an amazing equity curve approaches certainty as the number of combinations you try increases.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://jonathankinlay.com/2026/03/reinforcement-learning-for-portfolio-optimization-from-theory-to-implementation/" target="_blank">Reinforcement Learning for Portfolio Optimization: From Theory to Implementation</a> </h4> <p class="source_author ml-1 mr-1"> jonathankinlay.com - Jonathan</p> <blockquote> <p class='mb-2'>The quest for optimal portfolio allocation has occupied quantitative researchers for decades. Markowitz gave us mean-variance optimization in 1952,¹ and since then we’ve seen Black-Litterman, risk parity, hierarchical risk parity, and countless variations. Yet the fundamental challenge remains: markets are dynamic, regimes shift, and static optimization methods struggle to adapt. </p><p class='mb-2'> </p><p class='mb-2'>What if we could instead train an agent to learn portfolio allocation through experience — much like a human trader develops intuition through years of market participation? </p><p class='mb-2'> </p><p class='mb-2'>Enter reinforcement learning (RL). Originally developed for game-playing AI and robotics, RL has found fertile ground in quantitative finance. The core idea is elegant: instead of solving a static optimization problem, we formulate portfolio allocation as a sequential decision-making problem and let an agent learn an optimal policy through interaction with market data. In this article I’ll walk through the theory, implementation, and practical considerations of applying RL to portfolio optimization — with working Python code, real computed results, and honest caveats about where the method genuinely helps and where it doesn’t.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/the-random-forest-of-the-2030s" target="_blank">The Random Forest of the 2030s?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>Tabular foundation models (TFMs) are a paradigm shift from traditional tabular ML: They are transformer-based architectures pre-trained on synthetic data. There is no classic training step. Instead, TFMs predict the test data in a single forward pass of combined training and test data without any parameter updates (in-context learning). </p><p class='mb-2'> </p><p class='mb-2'>These last few weeks of deep-dive have reshaped how I think about TFMs and tabular ML as a whole. I won’t claim I can predict the future. I’ve been completely wrong before, like about how good AI would become at coding. Instead of predictions, here are a few scenarios of increasing impact of TFMs (levels) on everyday tabular ML work. </p><p class='mb-2'> </p><p class='mb-2'>Let’s dive in.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://valeman.medium.com/why-im-writing-a-new-book-on-foundation-models-for-tabular-data-4d67236401fe" target="_blank">Why I’m Writing a New Book on Foundation Models for Tabular Data</a> </h4> <p class="source_author ml-1 mr-1"> medium.com - Valeriy Manokhin, PhD, MBA, CQF</p> <blockquote> <p class='mb-2'>Foundation models for tabular data are powerful. </p><p class='mb-2'> </p><p class='mb-2'>They are also dangerously easy to misjudge. </p><p class='mb-2'> </p><p class='mb-2'>And in structured domains — especially high-stakes ones like finance, insurance, healthcare — misjudgment isn’t harmless. </p><p class='mb-2'> </p><p class='mb-2'>It leads to fragile systems. </p><p class='mb-2'> </p><p class='mb-2'>It leads to misplaced confidence. </p><p class='mb-2'> </p><p class='mb-2'>It leads to real consequences.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-11.html Mon, 16 Mar 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-10.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://arxiv.org/abs/2602.01736" target="_blank">Position: The Inevitable End of One-Architecture-Fits-All-Domains in Time Series Forecasting</a> </h4> <p class="source_author ml-1 mr-1"> arxiv.org - Qinwei Ma, Jingzhe Shi, Jiahao Qiu, Zaiwen Yang</p> <blockquote> <p class='mb-2'> Abstract:Recent work has questioned the effectiveness and robustness of neural network architectures for time series forecasting tasks. We summarize these concerns and analyze groundly their inherent limitations: i.e. the irreconcilable conflict between single (or few similar) domains SOTA and generalizability over general domains for time series forecasting neural network architecture designs. Moreover, neural networks architectures for general domain time series forecasting are becoming more and more complicated and their performance has almost saturated in recent years. As a result, network architectures developed aiming at fitting general time series domains are almost not inspiring for real world practices for certain single (or few similar) domains such as Finance, Weather, Traffic, etc: each specific domain develops their own methods that rarely utilize advances in neural network architectures of time series community in recent 2-3 years. As a result, we call for the time series community to shift focus away from research on time series neural network architectures for general domains: these researches have become saturated and away from domain-specific SOTAs over time. We should either (1) focus on deep learning methods for certain specific domain(s), or (2) turn to the development of meta-learning methods for general domains. </p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-10.html Mon, 09 Mar 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-9.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/when-context-changes-so-does-threat" target="_blank">When Context Changes, So Does Threat</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The following summary critically reviews the research paper titled “Dynamic Expected Threat (DxT) Model: Addressing the Deficit of Realism in Football Action Evaluation” by Karim Hassani, Mohammed Ramdani and Marwane Lotfi. All data, figures, and analysis presented here are drawn from their original work; I do not claim any authorship or ownership of the content. This summary has been written to provide a concise and technically informed synthesis of the paper’s findings, methodologies, and implications, while maintaining fidelity to the authors’ intellectual contributions.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-018-before-xt-before-vaep-before" target="_blank">TI #018: Before xT. Before VAEP. Before possession value became mainstream.</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>How do you value the passes that move the ball into a dangerous zone — when none of them result in a shot? </p><p class='mb-2'> </p><p class='mb-2'>In 2017, Tom Decroos and colleagues proposed one of the first structured answers. </p><p class='mb-2'> </p><p class='mb-2'>It wasn’t flashy. It wasn’t deep learning. </p><p class='mb-2'> </p><p class='mb-2'>But it introduced an idea that would shape football analytics for the next decade.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-9.html Mon, 02 Mar 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-8.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-014-the-four-signals-recruiters" target="_blank">TI #014: The Four Signals Recruiters Look For in Young Footballers</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The real question is not how talent should be identified, but how it is identified when a recruiter has to make a decision. </p><p class='mb-2'> </p><p class='mb-2'>This paper approaches that question by focusing on the people making those decisions. Instead of analysing players, the authors analyse experienced youth recruiters and ask them to explain, in detail, what they prioritise when selecting Under-13 players. </p><p class='mb-2'> </p><p class='mb-2'>These judgements are not reconstructed later or framed through general development ideas. They are made at the moment selection actually happens.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.cannonstats.com/p/introducing-garbage-time?hide_intro_popup=true" target="_blank">Introducing "Garbage Time"</a> </h4> <p class="source_author ml-1 mr-1"> cannonstats.com - Scott Willis</p> <blockquote> <p class='mb-2'>What is Garbage Time? This is a concept that originate with NBA basketball, “garbage time” is the term that describes late game situations where the stats and performances aren’t really reflective of a normal competitive game. It is those final minutes of a blowout when the outcome is certain, coaches rest their stars, and the game is just played out to the final buzzer. </p><p class='mb-2'> </p><p class='mb-2'>This has been adapted to American Football and baseball, but for the most part it has not been adopted very widely into soccer. </p><p class='mb-2'> </p><p class='mb-2'>For most matches in the Premier League this is not something that will occur a lot, but it does happen often enough that it I believe that it is valuable to identify these situations because they alter the way that the match is played and resulting stats would reflect those same distortions.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-016-explainability-might-be-the" target="_blank">TI #016: Explainability Might Be the Real Competitive Edge</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>What if the biggest problem with expected goals isn’t how accurate it is… but how little people trust it? </p><p class='mb-2'> </p><p class='mb-2'>For years, we’ve built better and better models. More data. More features. More advanced models. </p><p class='mb-2'> </p><p class='mb-2'>And yet, walk into many football departments, walk into X/Twitter conversations or football talk shows and you’ll still hear the same question: </p><p class='mb-2'> </p><p class='mb-2'>“Why would I trust your numbers?” </p><p class='mb-2'> </p><p class='mb-2'>This week’s Timeless Insights dives into a paper by Jan Van Haaren that tackles exactly that tension — not by chasing higher predictive performance, but by rethinking how we design models in the first place.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">👁️ Computer Vision </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://learnopencv.com/yolov26-real-time-deployment/" target="_blank">YOLOv26: An Object Detector Built for Real-Time Deployment</a> </h4> <p class="source_author ml-1 mr-1"> learnopencv.com - Bhomik Sharma</p> <blockquote> <p class='mb-2'>Every few months, the computer vision community prepares for a new YOLO release, typically faster, marginally lighter, and incrementally more accurate than the last. YOLOv26 (object detector), released by Ultralytics in January 2026, breaks that pattern. </p><p class='mb-2'> </p><p class='mb-2'>Rather than increasing architectural complexity, YOLOv26 adopts an edge-first engineering approach. The focus shifts to latency, export paths, and hardware-friendly design. The result is a detector specifically designed for applications in robotics, drones, mobile devices, and embedded systems.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🌩 Forecasting </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://valeman.medium.com/stop-trusting-point-forecasts-and-stop-trusting-naive-conformal-prediction-too-737f5485deb2" target="_blank">Stop Trusting Point Forecasts. (And Stop Trusting Naive Conformal Prediction, Too)</a> </h4> <p class="source_author ml-1 mr-1"> medium.com - Valeriy Manokhin, PhD, MBA, CQF</p> <blockquote> <p class='mb-2'>Why standard uncertainty methods break on time series data — and the production-grade fix you need to implement today.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-8.html Mon, 23 Feb 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-7.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.americansocceranalysis.com/home/2026/2/8/check-out-futi" target="_blank">Check Out futi</a> </h4> <p class="source_author ml-1 mr-1"> americansocceranalysis.com</p> <blockquote> <p class='mb-2'>futi is a next-gen live score app built to bring pro-quality analytics to fans in a familiar format. We’re rolling out model explainers and early data releases so you can preview the models behind the app. Follow us to keep up to date and be part of the future of fan-facing stats at the links below:</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://futiapp.substack.com/p/how-futi-sees-football-in-goal-probabilities" target="_blank">How futi sees football in goal probabilities</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - John Muller</p> <blockquote> <p class='mb-2'>What I mean is: how good is this pass, quantitatively? Go ahead, try to put a number on it. </p><p class='mb-2'> </p><p class='mb-2'>As far as most football stats are concerned, the number is 1. Messi attempted one pass. He completed one pass. Maybe you get fancy and label it a throughball or a final third entry or something, but it’s still just 1/1. </p><p class='mb-2'> </p><p class='mb-2'>See the problem here? Counting stats treat different actions the same. They starve them of context. A pass is simply a pass whether it’s a slow roller to the goalkeeper or a heat-seeking missile between the opponent’s defensive lines. A final third entry could travel six inches or 60 yard</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-015-football-is-not-about-passes" target="_blank">TI #015: Football Is Not About Passes — It’s About Connections</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>For years, football analytics tried to understand teams by adding up individual actions. More passes. More touches. More duels. More running. </p><p class='mb-2'> </p><p class='mb-2'>But teams don’t play as collections of players. </p><p class='mb-2'> </p><p class='mb-2'>They play as systems. </p><p class='mb-2'> </p><p class='mb-2'>This paper makes a deceptively simple claim: if you want to understand how a team really plays, you shouldn’t start with players at all — you should start with relationships.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/the-promise-and-limits-of-machine" target="_blank">The Promise and Limits of Machine Learning in Football Attacking Analysis</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>A review of how machine learning has been applied to analyze attacking performance, identify key indicators, and support tactical decision-making in professional football.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://futiapp.substack.com/p/introducing-phases-of-play" target="_blank">Introducing phases of play</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - John Muller</p> <blockquote> <p class='mb-2'>How do you put football data into tactical context? </p><p class='mb-2'> </p><p class='mb-2'>Most of the time, people don’t. A pass is a pass and a tackle is a tackle no matter how they happen. Football stats typically treat each action as an isolated event, ignoring what’s going on in the game around it. </p><p class='mb-2'> </p><p class='mb-2'>That’s because most raw football data — what’s known as “event data” — doesn’t contain information about what’s happening away from the ball. Somebody watches the game and logs every on-ball action, adding details about when and where it happened, who did it, whether it was successful, stuff like that. There’s a lot you can do with that data, but it doesn’t tell you how the teams are set up or how players are moving off the ball. As somebody once put it, event data is like listening to the game on the radio. </p><p class='mb-2'> </p><p class='mb-2'>To help fill in the missing context, futi developed a model based on a common framework that coaches and analysts use to describe the game: phases of play.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/expanding-elo-to-evaluate-individual" target="_blank">Expanding Elo to Evaluate Individual Player Performance</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Traditional player rating systems in football generally fall into two categories: subjective evaluations and objective statistics-based approaches. In subjective systems, experts or journalists assign ratings based on their perception of performance. In objective systems, ratings are derived from recorded match data such as passes, duels, or other events, often using statistical or machine learning techniques. Both approaches have limitations. Subjective ratings can be influenced by bias, while event-based models often require extensive tracking or event data that are not universally available. </p><p class='mb-2'> </p><p class='mb-2'>The paper “A football player rating system” introduces a new approach: an adaptation of the Elo algorithm, originally developed for individual sports such as chess, to evaluate individual players in football. The key idea is to construct an objective and adaptive rating system that relies only on official match reports. These reports include the final score, lineups, substitutions, and minutes played.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/how-tabular-foundation-models-are" target="_blank">How Tabular Foundation Models learn without real data</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>In this post, we take a deep dive into how TFMs like TabPFN and TabICL are pre-trained to enable in-context learning (= single forward pass without weight updates). We’ll have a look into the pre-training procedure and how the pre-training data are generated (also called the prior). This post is a bit more general about pre-training TFMs, not a particular one, but I’ll reference TabPFN and TabICL mostly.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-7.html Mon, 16 Feb 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-6.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.expectinggoals.com/p/predicting-set-piece-goals-and-assists" target="_blank">Predicting Set Piece Goals and Assists, a Mini-Study</a> </h4> <p class="source_author ml-1 mr-1"> expectinggoals.com - Michael Caley</p> <blockquote> <p class='mb-2'>In the midst of working on last week’s set piece study, I pulled some data together on player shot and goal production from set pieces, and on teams’ primary set piece takers and set piece targets. I found that player set piece goals and assists are best predicted not by open play statistics, but by a combination of that player’s set piece production and by their role on the team’s set plays. These results suggest that set piece production should be understood as a separate aspect of player statistical production, and players should be evaluated and projected in distinct ways on set piece and open play skills. That is an interesting result even if it ended up being extraneous to the argument of last week’s newsletter.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-6.html Mon, 09 Feb 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-5.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.americansocceranalysis.com/home/2026/1/25/introducing-glass-onion-by-us-soccer-an-identifier-synchronization-tool" target="_blank">Introducing Glass Onion by US Soccer, An Identifier Synchronization Tool</a> </h4> <p class="source_author ml-1 mr-1"> americansocceranalysis.com - Akshay Easwaran</p> <blockquote> <p class='mb-2'>U.S. Soccer released an open-source tool called Glass Onion to solve the common data engineering problem of matching teams, matches, and players across multiple soccer data sources, since there’s no universal identifier in the sport. </p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/can-data-tell-if-two-players-will" target="_blank">Can Data Tell If Two Players Will Click on the Pitch?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>The following summary critically reviews the research paper titled “Player Chemistry: Striving for a Perfectly Balanced Soccer Team” by Lotte Bransen and Jan Van Haaren. All data, figures, and analysis presented here are drawn from their original work; I do not claim any authorship or ownership of the content. This summary has been written to provide a concise and technically informed synthesis of the paper’s findings, methodologies, and implications, while maintaining fidelity to the authors’ intellectual contributions.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.expectinggoals.com/p/the-set-piece-revolution" target="_blank">The Set Piece Revolution</a> </h4> <p class="source_author ml-1 mr-1"> expectinggoals.com - Michael Caley</p> <blockquote> <p class='mb-2'>The piece argues that set pieces in the Premier League have shifted from a quirky part of the game to a central tactical element, with teams increasingly optimizing them for goals. </p><p class='mb-2'>It highlights a significant rise in scoring from corners and long throw-ins, driven by analytical insights and strategic exploitation of rules like the lack of offside on these plays. </p><p class='mb-2'>This tactical shift coincides with a decline in open-play chances, suggesting that clubs are trading risky open attacks for higher-efficiency dead-ball opportunities.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🌩 Forecasting </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://x.com/RohOnChain/status/2017314080395296995" target="_blank">The Math Needed for Trading on Polymarket (Complete Roadmap)</a> </h4> <p class="source_author ml-1 mr-1"> x.com - @RohOnChain</p> <blockquote> <p class='mb-2'>I'm going to break down the essential math you need for trading on Polymarket. I'll also share the exact roadmap and resources that helped me personally.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-5.html Mon, 02 Feb 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-3.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">🎲 Betting </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.economist.com/christmas-specials/2025/12/18/the-battle-to-stop-clever-people-betting" target="_blank">The battle to stop clever people betting</a> </h4> <p class="source_author ml-1 mr-1"> economist.com</p> <blockquote> <p class='mb-2'>The tools bookmakers use to block data-savvy gamblers, and how to get round them</p> </blockquote> </div> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-011-what-actually-wins-football" target="_blank">What Actually Wins Football Matches?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Every season, clubs produce thousands of pages of match reports. </p><p class='mb-2'>Shots, possession, passes, duels, sprints, xG, field tilt — all carefully tracked, ranked, compared. </p><p class='mb-2'> </p><p class='mb-2'>Yet the main question remains: which of these actually matter for winning football matches? </p><p class='mb-2'> </p><p class='mb-2'>This paper tackles that question head-on by doing something most analyses avoid: </p><p class='mb-2'>throwing almost everything into the model and letting the data decide.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/what-if-transfers-could-be-simulated" target="_blank">What If Transfers Could Be Simulated Before They Happen?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Using player-conditioned GPT models to evaluate transfer fit through counterfactual match sequences.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.americansocceranalysis.com/home/2026/1/14/measuring-goalkeeper-impact-in-mls-a-data-driven-look-at-value-volatility-and-efficiency" target="_blank">Measuring Goalkeeper Impact in MLS: A Data-Driven Look at Value, Volatility, and Efficiency</a> </h4> <p class="source_author ml-1 mr-1"> americansocceranalysis.com - Lucas Morefield</p> <blockquote> <p class='mb-2'>Goalkeeper evaluation has always been one of the thorniest challenges in soccer analytics. The position is defined by small sample sizes, high variance, and context-driven outcomes. In Major League Soccer, where roster rules magnify the impact of every marginal dollar, understanding goalkeeper value is especially important. With clubs often operating near budget ceilings, a single overperformance or underperformance in goal can shift playoff probability, alter roster-building timelines, or change the financial implications of a season.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/tabular-ml-is-about-to-get-weird" target="_blank">The rise of tabular foundation models</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>However, PFNs are not just one more algorithm in scikit-learn, but they turn the way we model tabular data upside down. </p><p class='mb-2'> </p><p class='mb-2'>It’s time we talk about why tabular ML is (maybe) getting weird.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🧮 Statistics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.argmin.net/p/hermeneutics-of-crapshoots" target="_blank">Hermeneutics of crapshoots</a> </h4> <p class="source_author ml-1 mr-1"> argmin.net - Ben Recht</p> <blockquote> <p class='mb-2'>How many times do I need to see something to believe it?</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-3.html Mon, 19 Jan 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2026-2.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.cannonstats.com/p/why-xg-doesnt-need-fixing-but-our?hide_intro_popup=true" target="_blank">Why xG Doesn't Need Fixing (But Our Communication Might)</a> </h4> <p class="source_author ml-1 mr-1"> cannonstats.com - Scott Willis</p> <blockquote> <p class='mb-2'>On the Cannon Stats discord (it’s a pretty fun place to hang out and talk Arsenal/Soccer and free to join) a subscriber shared this video titled “The 4 Flaws Of Expected Goals”.This seems specifically designed to needle me (and probably Adam too), but it’s a good video despite the bait-y title/thumbnail, and it really got me thinking.The first thing is that this is not really a video about the flaws or xG but rather how more casual fans use xG and some of the more common misunderstandings of what the intentions are.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-010-why-some-players-just-cant" target="_blank">Why Some Players Just Can’t Be Replaced</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Most attempts to describe players start and end with outcomes: goals, assists, chances created. But football is mostly something else — long stretches of circulation, structure, and coordination far from the box.This 2015 paper asked a deceptively simple question: can a player’s passing style be captured as a stable, quantifiable pattern?</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://karpathy.ai/zero-to-hero.html" target="_blank">Neural Networks: Zero To Hero</a> </h4> <p class="source_author ml-1 mr-1"> karpathy.ai - Andrej Karpathy</p> <blockquote> <p class='mb-2'>A course by Andrej Karpathy on building neural networks, from scratch, in code.We start with the basics of backpropagation and build up to modern deep neural networks, like GPT. In my opinion language models are an excellent place to learn deep learning, even if your intention is to eventually go to other areas like computer vision because most of what you learn will be immediately transferable. This is why we dive into and focus on languade models.Prerequisites: solid programming (Python), intro-level math (e.g. derivative, gaussian)</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://zekcrates.quarto.pub/deep-learning-library/" target="_blank">Build a Simple Deep Learning Library</a> </h4> <p class="source_author ml-1 mr-1"> quarto.pub - zekcrates</p> <blockquote> <p class='mb-2'>Instead of just learning how to use a deep learning library, we are going to learn how to create one.We start with a blank file and NumPy, and we don’t stop until we have a functional autograd engine and a collection of layer modules. By the end, we will use it to train MNIST, simple CNN and simple ResNet.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/how-can-language-models-inspire-football" target="_blank">How Can Language Models Inspire Football Analytics?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Adapting transformer architectures to learn foundational player features from match sequences.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.youtube.com/watch?v=d95J8yzvjbQ" target="_blank">The Thinking Game</a> </h4> <p class="source_author ml-1 mr-1"> youtube.com - DeepMind</p> <blockquote> <p class='mb-2'>The Thinking Game takes you on a journey into the heart of DeepMind, capturing a team striving to unravel the mysteries of intelligence and life itself. </p><p class='mb-2'> </p><p class='mb-2'>Filmed over five years by the award winning team behind AlphaGo, the documentary examines how Demis Hassabis’s extraordinary beginnings shaped his lifelong pursuit of artificial general intelligence. It chronicles the rigorous process of scientific discovery, documenting how the team moved from mastering complex strategy games to the ups and downs of solving a 50-year-old "protein folding problem" with AlphaFold. </p><p class='mb-2'> </p><p class='mb-2'>Following its world premiere at the Tribeca Festival and a successful international tour, the film is now available here for all to watch for free.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🧮 Statistics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.argmin.net/p/statistical-fatalism" target="_blank">Statistical Fatalism</a> </h4> <p class="source_author ml-1 mr-1"> argmin.net - Ben Recht</p> <blockquote> <p class='mb-2'>People will behave the same before you make a policy and after you make a policy. Most people who work on causal inference know none of this is true, of course. And any seasoned machine learning engineer knows this as well when maintaining systems to continually retrain their stable of prediction models. </p><p class='mb-2'> </p><p class='mb-2'>Dawid is of course not the first person to identify this problem. Fifty years ago, economist Robert Lucas pointed out that you can’t use historical data to predict the impact of economic policy because of feedback effects</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2026-2.html Mon, 12 Jan 2026 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2025-52.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/ti-009-when-teams-get-tired-football" target="_blank">When Teams get Tired, Football gets Predictable</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Football looks chaotic. Players roam, press, stretch, collapse. But beneath the noise, teams quietly settle into patterns; and this paper shows how and when that happens. </p><p class='mb-2'> </p><p class='mb-2'>Using full‑match tracking data from a Premier League game, Duarte et al. (2012) treated teams as complex systems, not collections of individuals. Instead of focusing on passes or shots, they tracked how a team’s shape evolved over time: how much space it occupied, how stretched it became, and how regularly those patterns repeated. </p><p class='mb-2'> </p><p class='mb-2'>What they found is subtle, but powerful. </p><p class='mb-2'> </p><p class='mb-2'>Teams didn’t become more rigid as matches progressed, quite the opposite. The magnitude of variation increased: teams covered more space, stretched more, and deviated further from their average shape. Yet at the same time, the structure of that variation became simpler. Using approximate entropy, the authors showed that team behaviour grew increasingly regular and predictable within each half. </p><p class='mb-2'> </p><p class='mb-2'>In plain terms: teams move more, but in fewer ways.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/which-machine-learning-models-perform" target="_blank">Which Machine Learning Models Perform Best for Football Match Prediction?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Machine learning models that incorporate domain-specific rating features consistently outperform traditional statistical models. Ensemble methods, particularly gradient-boosted trees (e.g., XGBoost, CatBoost), excel when paired with tailored features like Elo, pi-, or Berrar ratings. Feature engineering, notably from limited datasets, remains "a critical component in the predictive power of models". Random Forests and SHAP provide valuable interpretability, which is especially useful in applied contexts like coaching.Nevertheless, standard benchmark datasets are rare, complicating direct model comparison. The 2017 Soccer Prediction Challenge is one of the few shared evaluation platforms. Deep learning methods have yet to consistently outperform ensembles, often due to data constraints or lack of appropriate temporal feature modeling.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🎙️ Podcast </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://smartbettingclub.com/blog/episode-97-anthony-kaminskas-on-akbets-on-modern-bookmaking-restrictions-more/" target="_blank">Episode 97: Anthony Kaminskas on AKBets on Modern Bookmaking, Restrictions</a> </h4> <p class="source_author ml-1 mr-1"> smartbettingclub.com - Peter Ling</p> <blockquote> <p class='mb-2'>In Episode 97 of the Smart Betting Club Podcast, I am joined for the third time by Anthony Kaminskas, founder of AKBets and a former professional punter turned bookmaker.​Nearly three years on from the launch of AKBets, this conversation takes stock of how the business has grown, how the betting industry has shifted, and why life as a modern bookmaker is far tougher than most punters realise – including a healthy discussion on betting restrictions.AK speaks candidly about operating under UK regulation, handling restrictions, managing compliance risk, and absorbing rising costs at every level of the business.​He also lifts the lid on the real economics of running a sportsbook, sharing hard numbers on payroll, data feeds, duties, levies, payment processing and platform fees.​It is an honest look at why betting markets look the way they do, why margins are tightening, and why both punters and bookmakers are being squeezed by the same pressures.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2025-52.html Mon, 29 Dec 2025 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2025-51.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/what-can-data-really-tell-us-about" target="_blank">What Can Data Really Tell Us About Strategy?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>A Markov Decision Framework to Model, Analyze, and Optimize Team Behavior in Professional Football.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/the-unfiltered-reality-of-football" target="_blank">The Unfiltered Reality of Football Analytics in 2025</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>How 200 Practitioners Reveal the Real Analytics Habits Shaping Player Health and Performance.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.americansocceranalysis.com/home/2025/12/15/the-2026-asa-big-dashboard" target="_blank">The 2026 MLS SuperDraft ASA Big (Dash)Board</a> </h4> <p class="source_author ml-1 mr-1"> americansocceranalysis.com - Paul Harvey</p> <blockquote> <p class='mb-2'>The MLS Superdraft in many ways kicks off the offseason, as most teams have yet to make any real signings. Even though many of the draft picks will never play for their respective first teams, it still marks the beginning of a period of excitement and analysis, as hope springs eternal for every team in MLS.As part of that, I have continued my project of analyzing the draft for the best available players. Last year, we divided the best players by data into three tiers. Of the 19 tier one players, 15 were drafted and 6 were taken in the top 10. That cohort saw 5 players play more than 1000 minutes in MLS last season, with players like Manu Duah and Tate Johnson being the highlights. Many went on to dominate in MLS Next Pro or USL, like Michael Adedokun, Nick Fernandez, and Emil Jaaskelainen.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🏛️ Economics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://swissramble.substack.com/p/manchester-city-finances-202425?utm_source=post-email-title" target="_blank">Manchester City Finances 2024/25</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Swiss Ramble</p> <blockquote> <p class='mb-2'>Manchester City’s 2024/25 accounts cover a season when they finished third in the Premier League and reached the FA Cup final, where they were surprisingly beaten by Crystal Palace.On the international stage, they advanced to the knockout round of the Champions League, where they were eliminated by Real Madrid, while they also reached the last 16 in the expanded FIFA Club World Cup, where they lost to Saudi club Al-Hilal.For most clubs, this would be a more than satisfactory season, but City admitted that this represented a “disappointing outcome”, which was perfectly understandable, considering that they had won the league six times in the previous seven years.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2025-51.html Mon, 22 Dec 2025 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2025-50.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">🎲 Betting </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.itia.tennis/news/sanctions/french-tennis-player-quentin-folliot-suspended-for-20-years/" target="_blank">French tennis player Quentin Folliot suspended for 20 years</a> </h4> <p class="source_author ml-1 mr-1"> itia.tennis</p> <blockquote> <p class='mb-2'>The International Tennis Integrity Agency (ITIA) today confirms that French tennis player Quentin Folliot has been suspended from the sport for a period of 20 years, fined $70,000, and ordered to repay corrupt payments totalling more than $44,000 after committing 27 breaches of the Tennis Anti-Corruption Program (TACP).</p> </blockquote> </div> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.cannonstats.com/p/arsenal-compared-to-expectations-657?hide_intro_popup=true" target="_blank">Arsenal compared to expectations: KPI Update for December 2025/26</a> </h4> <p class="source_author ml-1 mr-1"> cannonstats.com - Scott Willis</p> <blockquote> <p class='mb-2'>Checking in on how Arsenal are performing compared to the Key Performance Indicators for the 2025-26 season</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/why-some-clubs-turn-data-into-wins" target="_blank">Why Some Clubs Turn Data Into Wins — and Others Don’t</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>What research shows about tactical modelling, automation, and the gaps between science and practice.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.pythonfootball.com/p/the-death-of-the-long-range-screamer" target="_blank">The Death of the Long-Range Screamer</a> </h4> <p class="source_author ml-1 mr-1"> pythonfootball.com - MartinOnData</p> <blockquote> <p class='mb-2'>Ever since the xG revolution took place, everyone loves saying that xG killed long-range screamers. Well … that sounds like the perfect topic for this little newsletter.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.pythonfootball.com/p/how-good-really-was-fivethirtyeights" target="_blank">How Good Really Was FiveThirtyEight’s Soccer Power Index?</a> </h4> <p class="source_author ml-1 mr-1"> pythonfootball.com - MartinOnData</p> <blockquote> <p class='mb-2'>FiveThirtyEight has always had a kind of mythic status in the football world. Their Soccer Power Index (SPI) was one of the first “super-model” rating systems I stumbled upon when I began my own analytics journey. And for years — before it was shut down — SPI became a reference point for pre-game probabilities and end-of-season title chances across football Twitter.Last week, I came across a historical dataset that included SPI’s match-by-match outcome probabilities. Naturally, the first question that came to my mind was:“If someone blindly trusted SPI for years… would they have made money?”And that curiosity brings us to today’s issue.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">👁️ Computer Vision </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://blog.roboflow.com/identify-basketball-players/" target="_blank">How to Detect, Track, and Identify Basketball Players with Computer Vision</a> </h4> <p class="source_author ml-1 mr-1"> roboflow.com - Piotr Skalski</p> <blockquote> <p class='mb-2'>Basketball is one of the hardest sports for computer vision. Players move at high speed, causing motion blur. Bodies crash into each other and overlap, making jersey numbers hard to see. Uniforms look almost identical, so appearance alone is not enough to tell players apart. On top of that, the camera keeps zooming and panning, changing perspective every second.In this blogpost, we show how to overcome these challenges and build a computer vision system that can detect, track, and recognize NBA players during a game.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🕰️ Blast From the Past </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://pricetheory.uchicago.edu/levitt/Papers/LevittWhyAreGamblingMarkets2004.pdf" target="_blank">Why are gambling markets organised so differently from financial markets?</a> </h4> <p class="source_author ml-1 mr-1"> uchicago.edu - Steven D. Levitt</p> <blockquote> <p class='mb-2'>The market for sports gambling is structured very differently from the typical financial market. In sports betting, bookmakers announce a price, after which adjustments are small and infrequent. Bookmakers do not play the traditional role of market makers matching buyers and sellers but, rather, take large positions with respect to the outcome of game. Using a unique data set, I demonstrate that this peculiar price-setting mechanism allows bookmakers to achieve substantially higher profits. Bookmakers are more skilled at predicting the outcomes of games than bettors and systematically exploit bettor biases by choosing prices that deviate from the market clearing price.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2025-50.html Mon, 15 Dec 2025 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2025-49.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.expectinggoals.com/p/what-happens-when-teams-have-nothing" target="_blank">What Happens When Teams Have "Nothing to Play For"</a> </h4> <p class="source_author ml-1 mr-1"> expectinggoals.com - Michael Caley</p> <blockquote> <p class='mb-2'>Using analytics to determine who wants it more</p> </blockquote> </div> <h3 class="category mb-3 mt-5">🏛️ Economics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://swissramble.substack.com/p/real-madrid-finances-202425" target="_blank">Real Madrid Finances 2024/25</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Swiss Ramble</p> <blockquote> <p class='mb-2'>Real Madrid’s 2023/24 accounts cover a season which was very much a case of “close, but no cigar”, as they finished runners-up in La Liga and were defeated in the final of the Copa del Rey and the Supercopa de Espana, losing out to their great rivals Barcelona on all three occasions.</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2025-49.html Mon, 08 Dec 2025 04:00:00 -0000 https://kubeia.io/sports_analytics_weekly/2025-48.html <!-- Newsletter content for each week --> <h3 class="category mb-3 mt-5">📝 Sports Analytics </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://thexgfootballclub.substack.com/p/how-can-we-see-a-teams-shape-in-real" target="_blank">How Can We See a Team’s Shape in Real Time?</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Alex Marin Felices</p> <blockquote> <p class='mb-2'>Shape Graphs Offer Instantaneous, Explainable Representations of Tactical Positioning in Football.</p> </blockquote> </div> <div class="newsletter-content mb-3"> <h4> <a href="https://www.mediafire.com/folder/93n2ayin033dg/FTG" target="_blank">Football Analytics with Python The Fast-Track Guide for Beginners</a> </h4> <p class="source_author ml-1 mr-1"> mediafire.com - Martin</p> <blockquote> <p class='mb-2'>Hi friend, </p><p class='mb-2'> </p><p class='mb-2'>About a month ago, I launched my Football Analytics with Python – The Fast-Track Guide for Beginners — a short, no-nonsense way to help you get started quickly with Python for football analytics. </p><p class='mb-2'> </p><p class='mb-2'>The plan was simple: you get a fast on-ramp into Python, and I get to monetise a bit of my writing for The Python Football Review. </p><p class='mb-2'> </p><p class='mb-2'>Then, the very week of the launch, FBref tightened access to their website… and even the lightest scraping through pandas.read_html suddenly stopped working. </p><p class='mb-2'> </p><p class='mb-2'>The reality is that there’s no way around FBref now without proper scraping — and teaching it is an entirely different skillset, far beyond the spirit of a “fast track.” If you need to learn Selenium, BeautifulSoup, HTML structure, session cookies, and browser automation… well, that’s a completely different product. </p><p class='mb-2'> </p><p class='mb-2'>Since the whole promise of the guide was “use advanced football data, fast,” I didn’t feel right keeping it up. So I pulled it and refunded everyone immediately. </p><p class='mb-2'> </p><p class='mb-2'>And honestly, thank you — to the people who emailed me suggestions, workarounds, and incredibly kind words. (I did spend six months building it, and it was pretty discouraging to see it break overnight.) And thank you as well for the genuinely positive reception from the 50 people who bought it. That meant a lot. </p><p class='mb-2'> </p><p class='mb-2'>That said, I rebuilt the guide removing all FBref dependencies (so no FBref data this time, sorry), slightly rewriting certain parts and making sure everything runs smoothly out of the box. </p><p class='mb-2'> </p><p class='mb-2'>And today, I’m giving you the fully updated guide completely free, along with all the code. </p><p class='mb-2'> </p><p class='mb-2'>You don’t need to buy anything. You don’t need to sign up again. Just download it and start learning.</p> </blockquote> </div> <h3 class="category mb-3 mt-5">💰 Quantitative Finance </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://www.youtube.com/watch?v=3V1OBeijRwo" target="_blank">Why we wrote a textbook - YouTube</a> </h4> <p class="source_author ml-1 mr-1"> youtube.com - Ole Peters</p> <blockquote> <p class='mb-2'>In June 2025, the textbook `An introduction to ergodicity economics' was published, with a launch party held at Gresham College in London. The video is a recording of a live lecture there about why we wrote it. </p> </blockquote> </div> <h3 class="category mb-3 mt-5">🤖 Machine Learning </h3> <div class="newsletter-content mb-3"> <h4> <a href="https://mindfulmodeler.substack.com/p/shap-is-not-all-you-need" target="_blank">SHAP Is Not All You Need</a> </h4> <p class="source_author ml-1 mr-1"> substack.com - Christoph Molnar</p> <blockquote> <p class='mb-2'>A most annoying misconception in the world of machine learning interpretability</p> </blockquote> </div> social@kubeia.io https://kubeia.io/sports_analytics_weekly/2025-48.html Mon, 01 Dec 2025 04:00:00 -0000