The record
Every forecast published between July 24, 2026 and August 26, 2026, how each one resolved, and how much the model held back alongside them. Losing days and days with no output are included — a record that drops them is not a record.
Summary
Accuracy
60.6%
60–39 across 99 published forecasts
Held back
77%
324 of 423 modeled games never published
Forecasts published
99
Across 34 days
Days with no output
4
Nothing cleared threshold
By grade
Bar length is the size of each grade’s sample; the solid portion is the share that resolved correct. Every rate is shown with the count it came from.
In each bar the solid colored segment is the correct share and the faded segment is the incorrect share. This record covers 99 published forecasts across 34 days. That is still a small sample and should be read as one.
Day by day
| Date | Published | Record | Outcomes |
|---|---|---|---|
| Jul 24 | 4 | 3–175% | CorrectCorrectCorrectIncorrect |
| Jul 25 | 3 | 2–167% | CorrectCorrectIncorrect |
| Jul 26 | 2 | 2–0100% | CorrectCorrect |
| Jul 27 | 2 | 1–150% | CorrectIncorrect |
| Jul 28 | 5 | 2–340% | CorrectCorrectIncorrectIncorrectIncorrect |
| Jul 29 | 0 | — | Nothing cleared threshold |
| Jul 30 | 0 | — | Nothing cleared threshold |
| Jul 31 | 3 | 2–167% | CorrectCorrectIncorrect |
| Aug 1 | 4 | 4–0100% | CorrectCorrectCorrectCorrect |
| Aug 2 | 5 | 5–0100% | CorrectCorrectCorrectCorrectCorrect |
| Aug 3 | 1 | 1–0100% | Correct |
| Aug 4 | 2 | 2–0100% | CorrectCorrect |
| Aug 5 | 4 | 2–250% | CorrectCorrectIncorrectIncorrect |
| Aug 6 | 3 | 2–167% | CorrectCorrectIncorrect |
| Aug 7 | 3 | 1–233% | CorrectIncorrectIncorrect |
| Aug 8 | 5 | 2–340% | CorrectCorrectIncorrectIncorrectIncorrect |
| Aug 9 | 5 | 4–180% | CorrectCorrectCorrectCorrectIncorrect |
| Aug 10 | 2 | 1–150% | CorrectIncorrect |
| Aug 11 | 4 | 3–175% | CorrectCorrectCorrectIncorrect |
| Aug 12 | 5 | 3–260% | CorrectCorrectCorrectIncorrectIncorrect |
| Aug 13 | 3 | 0–30% | IncorrectIncorrectIncorrect |
| Aug 14 | 3 | 2–167% | CorrectCorrectIncorrect |
| Aug 15 | 3 | 0–30% | IncorrectIncorrectIncorrect |
| Aug 16 | 1 | 0–10% | Incorrect |
| Aug 17 | 1 | 0–10% | Incorrect |
| Aug 18 | 0 | — | Nothing cleared threshold |
| Aug 19 | 3 | 1–233% | CorrectIncorrectIncorrect |
| Aug 20 | 3 | 3–0100% | CorrectCorrectCorrect |
| Aug 21 | 3 | 2–167% | CorrectCorrectIncorrect |
| Aug 22 | 5 | 5–0100% | CorrectCorrectCorrectCorrectCorrect |
| Aug 23 | 0 | — | Nothing cleared threshold |
| Aug 24 | 2 | 0–20% | IncorrectIncorrect |
| Aug 25 | 4 | 2–250% | CorrectCorrectIncorrectIncorrect |
| Aug 26 | 6 | 3–350% | CorrectCorrectCorrectIncorrectIncorrectIncorrect |
What this record does not tell you
- Ninety-nine forecasts over five weeks is a short window. Runs inside it have ranged from 12-0 across four days to 3-11 across seven — neither streak is the model's true rate, and neither is the headline number, exactly.
- The grade ranking has not been stable: the densest-pattern grade was the worst tier over one stretch of this window and the best over another. At these sample sizes the ordering between grades is mostly noise — read every rate with the count next to it.
- Games the model logged as informational rather than publishing ran 173-151 over the same window, close to a coin flip. The published forecasts are the product; the rest is the model declining to say anything useful.
- The model forecasts game outcomes. It does not price markets and does not account for line movement or availability.
- Late scratches, weather, and lineup changes after publication are not re-modeled.
Past performance does not predict future results. Model accuracy varies by sample and by period, and any published record reflects a limited window. Accuracy is reported rather than financial return, because accuracy is exactly reproducible from the model’s own database and a return figure would not be.