A disciplined college-football workflow. Profile the week's slate to your rules, let the GRIDIRON engine project every score, and size each play with quarter-Kelly — live off the real ESPN board.
Set your parameters once — favorite location, points laid, total range. The board pulls the live FBS slate from ESPN and surfaces only the games that fit your system.
GRIDIRON runs each qualifier through shrinkage-stabilized PF/PA, the PF/PA Claw-Back, and 8,000 Monte-Carlo sims to a projected final score — a number that can disagree with the market.
Every edge becomes a definitive side and total, then gets sized with quarter-Kelly under a hard weekly exposure cap so a heavy card can't overcommit the roll.
No touts, no lean-on-vibes — a repeatable process you can run every Saturday.
Three stages, one discipline — exactly what happens between the slate and the stake.
You define the shape of a bet you want — home or away favorite, the range of points laid (default 3 to 4.5), and a total window (default 47 to 57). The tool loads the week's full FBS board straight from ESPN's public scoreboard and filters it down to just the games that match. In-season it also reads each favorite's recent form; in Week 1 those flags sit out until games bank.
Each qualifying game is projected from real seasonal scoring. Team PF and PA are shrunk toward the national mean so small samples don't scream, turned into expected points via offense-versus-defense with home-field, then run through a Markov drive model across 8,000 Monte-Carlo simulations. Out comes a projected final score, a cover probability, and an over/under read.
Ported from the ERA Claw-Back. It fires only when the spread favorite has a worse season net differential (PF−PA) than the underdog — the market backing a team in spite of its fundamentals. In that spot it claws the favorite's rating toward the dog's before projecting, and the strength scales with market conviction: the favorite's implied win probability from the real posted moneyline when the book has one, otherwise derived from the spread. A -2.5/-135 favorite barely gets touched; a -7/-350 favorite gets clawed hard.
The cover probability feeds a quarter-Kelly stake at −110 off your bankroll, flat or compounding. A weekly exposure cap scales the whole card down if a heavy week would overcommit the roll. Every play lands in a tier — SMASH, STRONG, LEAN, or PASS — so you know how hard the model is leaning.
Everything runs in your browser off the live board; nothing here is a guarantee, and probabilities are model estimates. 21+. If gambling stops being fun, call 1-800-GAMBLER.
Browse the board free. Unlock the definitive picks, stakes, and Claw-Back readouts with Pro.
Send payment via Cash App to $TuttlemathWinsGames with your account email in the note. Access is flipped on once payment lands. Prices above are placeholders — set your own.
Prototype note: “Unlock Pro” activates access in this browser only. Real accounts and paid gating need a backend (Supabase + Stripe/Cash App) — that's the next build when you're ready.
Build your game profile once. The board matches the week's slate against it and surfaces the games that fit, GRIDIRON projects each score, and quarter-Kelly sizes the stake off your bankroll — with a hard weekly exposure cap so a heavy card can't overcommit the roll.
| Matchup | Fav | Line | Total | Fav last | PF / PA |
|---|
Live board loads from ESPN's public college-football scoreboard in your browser (no key): it reads each game's spread and total from the posted odds. Team PF/PA isn't in that feed and no team has 2026 scoring yet, so Week 1 priors are neutral — set your preseason ratings inline or via Import, and the projection will move off the market. The Claw-Back fires when the spread favorite has a worse 2025 net (PF−PA) than the dog, and its strength scales with market conviction — the favorite's implied win probability from the real posted moneyline when the book has one, otherwise derived from the spread. GRIDIRON: James-Stein shrinkage on PF/PA → expected points (offense × opponent defense + HFA) → Markov drive model → Monte Carlo, ~8k sims, margin variance calibrated to ~17 pts. Cover probabilities feed quarter-Kelly at −110. If the live load is blocked by CORS on your host, the verified Week 1 snapshot stays put and you can paste a board via Import. Probabilities are model estimates, not guarantees; nothing here is a lock.
Prototype auth — credentials stay in this browser. Don't reuse a real password.