Analysis paused. Showing saved records. Automatic predictions, grading, and settlement are paused during maintenance.
Global model laboratory · customer records excluded
Model Leaderboard
Choose one game and jurisdiction. Every row comes from a candidate locked before that exact official occurrence. Historical backtests, customer plays, and shared-game results from the other jurisdiction are not mixed into this forward record.
New York
Pick 10 model standings
50 completed pre-draw fields graded per model.
| Model | Fields graded | Exact hits | Pattern hits | Average pattern match | Average main matches | Bonus hit rate | Score vs random control | Prize-bearing fields |
|---|---|---|---|---|---|---|---|---|
| All-draw frequency | 50 | 0 · 0.0% | — | — | 4.56 | — | -0.015 | 0 · 0.0% |
| Balanced draw model | 50 | 0 · 0.0% | — | — | 4.68 | — | -0.009 | 0 · 0.0% |
| Draw-shape transition research Comparison model | 49 | 0 · 0.0% | 0 | 37.1% | 4.714 | — | -0.0073 | 0 · 0.0% |
| Recent momentum | 50 | 0 · 0.0% | — | — | 4.96 | — | +0.005 | 0 · 0.0% |
| Same-draw frequency | 50 | 0 · 0.0% | — | — | 4.56 | — | -0.015 | 0 · 0.0% |
| Uniform control Control | 50 | 0 · 0.0% | — | — | 4.86 | — | 0.0 | 0 · 0.0% |
How to read this: average matches measures partial field quality; exact hits remain the strict full-result measure. Pattern hits require the complete pre-declared shape to match; average pattern match gives partial structural credit without counting it as a number win. “Score vs random control” compares each strategy with the uniform-control model on the same locked sample. A positive value is evidence to investigate, not proof of an edge, and it never changes the official odds.