β Premier League 2008-09 Β· Sun, May 17, 02:00 PM UTC Β· ref: M Atkinson
The betting picturemarket vs our model, market by market
βMarketβ = what bookmaker prices imply after removing their margin β the world's best guess. βModelβ = what our machine-learning model said BEFORE this game, replayed honestly (it never saw the result). A bet is only interesting when the model sees something meaningfully more likely than the market does (the edge) β it's never about picking the winner.
Who wins? (90 minutes)
| Outcome | Market says | Model says | Best price | Verdict |
|---|
| Liverpool win | 68% | 62% | 1.40 | fair |
| Draw | 19% | 24% | 5.00 | value: +21% |
| West Brom win | 13% | 13% | 8.00 | value: +8% |
across 2 bookmakers Β· probabilities de-vigged (bookmaker margin removed)
The match in numberswhat happened + what each side looked like walking in
Match stats
| 0 | Half-time score | 1 |
| 12 | Shots | 16 |
| 7 | On target | 10 |
| 11 | Corners | 5 |
| 13 | Fouls | 4 |
| 2 | Yellow cards | 0 |
| 0 | Red cards | 0 |
Pre-match form
What each side looked like walking in β computed only from games before this one.