Pick a fixture, run the simulator, and explore the pitch in detail.
Forward record.
Every moneyline probability on the board is saved before kickoff and never changed after it. When the game is final, it is scored next to the sportsbook consensus captured at the same moment. No backfilling, no picking the good days.
Model vs market, same games
Games scored
Model favourite won
Brier score (lower is better)
Where it holds up, and where it doesn't
Premier League
NBA
NFL
NHL
Weekly performance
| Week of | Sport | Games | Model fav. won | Market fav. won | Model Brier | Market Brier |
|---|---|---|---|---|---|---|
| 2026-10-05 | All sports | 5 | 40% | 60% | 0.489 | 0.453 |
| 2026-10-05 | NFL | 1 | 0% | 0% | 0.514 | 0.545 |
| 2026-10-05 | NHL | 4 | 50% | 75% | 0.482 | 0.430 |
Brier and favourite rates on games where both a model and a market probability were saved before kickoff. One week is a small sample; read the trend, not a single row.
Every scored game
| Game | Kickoff (UTC) | Model pick | Model | Market | Final | Result |
|---|---|---|---|---|---|---|
| NBA · SAC vs LAL (preseason, not counted) | 2026-10-06 02:00 | LAL | 79.1% | 64.7% | 103–127 | Hit |
| NFL · NO vs ATL | 2026-10-06 00:15 | NO | 50.7% | 52.2% | 24–45 | Miss |
| NHL · DAL vs SJ | 2026-10-06 00:00 | DAL | 56.6% | 64.3% | 5–0 | Hit |
| NBA · MIL vs MIN (preseason, not counted) | 2026-10-06 00:00 | MIN | 66.0% | 62.1% | 97–116 | Hit |
| NHL · BOS vs OTT | 2026-10-05 23:30 | BOS | 53.9% | 48.6% | 1–4 | Miss |
| NHL · PIT vs WPG | 2026-10-05 23:30 | PIT | 54.5% | 61.6% | 2–3 | Miss |
| NHL · TB vs PHI | 2026-10-05 23:00 | TB | 56.6% | 65.9% | 4–1 | Hit |
| NBA · ATL vs MEM (preseason, not counted) | 2026-10-05 23:00 | ATL | 75.4% | 59.1% | 123–132 | Miss |
| NBA · DET vs PHX (preseason, not counted) | 2026-10-05 23:00 | DET | 75.5% | 55.2% | 109–107 | Hit |
| NBA · PHI vs NY (preseason, not counted) | 2026-10-05 23:00 | NY | 79.1% | 61.7% | 120–97 | Miss |
Next games on the record
| Game | Kickoff | Model pick | Model | Market |
|---|---|---|---|---|
| NBA · CHA vs BKN | CHA | 86.4% | 63.2% | |
| NHL · BUF vs MIN | BUF | 51.5% | 49.1% | |
| NHL · DET vs OTT | OTT | 53.2% | 48.3% | |
| NHL · MTL vs CAR | CAR | 56.5% | — | |
| NHL · NJ vs UTA | NJ | 50.9% | 52.5% | |
| NHL · TOR vs NSH | NSH | 53.5% | 41.8% | |
| NHL · NYR vs NYI | NYI | 50.6% | 41.2% | |
| NBA · OKC vs NO | OKC | 89.8% | 55.3% | |
| NHL · CHI vs STL | STL | 63.9% | — | |
| NBA · UTAH vs DEN | DEN | 89.9% | 51.0% | |
| NHL · SEA vs VGK | VGK | 64.4% | 60.6% | |
| NBA · GS vs LAL | LAL | 63.4% | 39.9% | |
| NHL · LA vs FLA | FLA | 53.0% | — | |
| NBA · IND vs MIN | MIN | 65.3% | 55.2% | |
| NHL · WPG vs COL | COL | 62.5% | 61.7% | |
| NHL · WSH vs PIT | WSH | 56.2% | 59.3% | |
| NBA · CHI vs PHX | PHX | 65.0% | 56.2% | |
| NBA · MEM vs ORL | ORL | 86.6% | 54.3% | |
| NBA · OKC vs MIL | OKC | 89.5% | — | |
| NBA · POR vs GS | POR | 59.7% | — | |
| NHL · ANA vs EDM | EDM | 53.3% | 53.7% | |
| NBA · CLE vs BOS | BOS | 59.5% | — | |
| NBA · BKN vs PHI | PHI | 82.7% | — | |
| NBA · MIA vs NO | MIA | 69.9% | — | |
| NFL · DAL vs TB | DAL | 59.7% | 78.3% | |
| Premier League · Arsenal vs Leeds United | Arsenal | 73.3% | 68.8% | |
| Premier League · Aston Villa vs Brentford | Aston Villa | 45.0% | 35.9% | |
| Premier League · Chelsea vs AFC Bournemouth | Chelsea | 37.6% | 55.3% | |
| Premier League · Ipswich Town vs Fulham | Fulham | 50.0% | 39.1% | |
| Premier League · Sunderland vs Brighton & Hove Albion | Brighton & Hove Albion | 44.9% | 40.4% |
The latest probability saved before kickoff is the one scored. A market column showing — means no sportsbook price has been captured for that game yet; it is still scored on the model side.
Archived pre-game predictions
These come from the simulator's prediction archive, which saved each published forecast and never edited it. Only predictions saved at least a day before the game are shown, so none can have seen the result. They were made by whichever model the page used at the time (named on each row), so they are listed here and not mixed into the forward record above.
Premier League
| Game | Date | Saved | Model | Pick | Chance | Final | Result |
|---|---|---|---|---|---|---|---|
| Premier League · Leeds vs Newcastle United | 2026-09-14 | 2026-09-13 20:06 | Baseline (fallback) | Newcastle United | 53.1% | 4–1 | Miss |
| Premier League · Aston Villa vs Nottingham Forest | 2026-09-12 | 2026-09-11 21:41 | Baseline (fallback) | Nottingham Forest | 55.4% | 1–2 | Hit |
| Premier League · Arsenal vs Chelsea | 2026-09-06 | 2026-08-12 01:59 | Baseline (fallback) | Arsenal | 50.4% | 2–1 | Hit |
| Premier League · Ipswich Town vs Liverpool | 2026-09-04 | 2026-09-03 21:51 | Baseline (fallback) | Liverpool | 49.1% | 0–2 | Hit |
How this record is kept
Every 10 minutes, and whenever a market board refreshes, the model's win probability for every scheduled game in all four sports is saved, with the sportsbook consensus captured at that moment where one exists (median of complete bookmaker quotes, margin removed). Finished games are scored automatically from the official scoreboards on the same cycle. The last values saved before kickoff are the ones scored; nothing is written after a game starts. Preseason games and two-way games that end tied are shown but left out of the scores. Brier score and log loss measure how good the probabilities are, not betting profit, and a better score than the market over a short run can be luck.