Access: Anonymous access. Content follows.
Blog and learning
Articles
- Glass Box vs Black Box Betting Models: Why Transparency WinsGlass box vs black box betting models: why models you can see, test on past seasons, and copy yourself beat opaque pick services long-run.
- Modeling NFL Like an Analyst: Spread, Total, and Player Props from ScratchBuild NFL prop, spread and total models from scratch: what stats to feed them, keeping confidence honest, removing the vig, and sizing bets in-browser.
- No-Code Sports Betting Studio HandbookNo-code betting model handbook: drag and drop a model, test it on past seasons, check it beats the closing line, build an NFL spread model in 10 minutes.
- MLB Platoon-Split Prop Modeling 2026: When the Lefty-Righty Edge Survives the BullpenHow to weight platoon splits in MLB hitter prop models — when the starter L/R edge survives the bullpen, and the Baseball Savant features that drive it.
- Monday Night Football Totals 2026: Where the Prime-Time Totals Edge HidesWhy Monday night NFL totals systematically trade above their fair number, with 2026 MNF schedule context and a pace-plus-weather under model.
- WNBA Pace-Adjusted Totals 2026: Building a Low-Liquidity Edge in a Growing MarketA pace-adjusted WNBA totals model on stats.wnba.com data: per-possession scoring × possessions, a 2025 backtest, and a Studio brick.
- Brier Score for Bettors (Plain English + Live Plot)A live chart of Brier score against forecast probability, with eight real pick confidence values marked -- no Week 1 pick has graded yet.
- Backtest a Sports Betting Model in Your Browser (Cornerstone)How to backtest a sports betting model in the browser with TensorFlow.js. Train, evaluate, and ship a model on your own device with no upload.
- Calibration Plots: Why Your 60% Picks Should Hit 60%A bettor-friendly guide to calibration plots. Reliability diagrams, isotonic and Platt recalibration, and how to spot overconfident models.
- Client-Side vs Python Betting Models: Latency, Privacy, ReproducibilityClient-side betting models (TensorFlow.js) vs Python pipelines (PyTorch, scikit-learn), compared on latency, privacy, reproducibility, and capability.
- Coin Betting Sims vs Real-Money Sportsbooks: When Each WinsA sports betting simulator with no real money is a serious tool, not a toy. Where coin sims beat real-money books for learning, modeling, and strategy.
- How a Deterministic Live-Betting Sim Works (Replayable Seeds)How a deterministic live betting simulator works: seeded play-by-play, replayable rounds, and coin pools that settle byte-identically every replay.
- NCAA Women's Basketball Totals Model: A Browser RecipeBuild an NCAA WBB totals model in the browser using pace, eFG%, and conference splits — with UConn, Iowa, LSU, and USC worked examples.
- NGS Speed and Route Trees as WR Prop FeaturesUse Next Gen Stats max speed and BDB route trees as WR prop features. 21+ mph thresholds, route-mix encoding, target-shadow math, and prop-board impact.
- Open-Source Betting Model Release Notes (Cornerstone)Monthly release notes for our open-source TF.js betting models. Brier vs Vegas close, model cards, changelog, and remixable workshop links.
- Pre-Snap Formation Features: A Bettable WalkthroughEngineer pre-snap formation features that move WR yardage props. Receiver alignment, motion-at-snap, RB depth, and the Brier-lift math.
- A Reproducible Backtest Checklist Bettors Actually FollowA 12-point checklist to tell a real edge from a hot streak: test past seasons, price after the juice, charge for the line you miss, prove it holds.
- A Significant Strikes Prop Model With Stylistic AdjustmentsBuild a UFC significant-strikes prop model in the browser. Per-15-minute rates, stylistic adjustments, takedown rate, and orthodox vs southpaw effects.