Six from nine. Season sits at 70.8%.
Results
The headline is the draw. Collingwood and Hawthorn finished 15.3 (93) to 13.15 (93) at the MCG — the model had Hawthorn at 61% and marked it wrong. You can’t really argue with that scoring, it’s a genuine miss but not one the model could have seen coming.
West Coast at 65% was the more frustrating result — Richmond won by 11 at Optus Stadium, a ground that should have added to the home side’s advantage. Carlton vs St Kilda at 52% was basically a coin flip and St Kilda won convincingly by 39.
The high-confidence calls held: Brisbane Lions (90%) demolished Essendon by 64, Sydney (91%) beat Melbourne by 17, Geelong (88%) beat North Melbourne by 49.
Technical
| Match | Predicted | Prob | Result | Margin |
|---|---|---|---|---|
| Collingwood vs Hawthorn | Hawthorn | 61% | ✗ | Draw |
| Western Bulldogs vs Fremantle | Fremantle | 61% | ✓ | +12 |
| Adelaide vs Port Adelaide | Adelaide | 61% | ✓ | +1 |
| Essendon vs Brisbane Lions | Brisbane Lions | 90% | ✓ | +64 |
| West Coast vs Richmond | West Coast | 65% | ✗ | −11 |
| Geelong vs North Melbourne | Geelong | 88% | ✓ | +49 |
| Carlton vs St Kilda | Carlton | 52% | ✗ | −39 |
| Sydney vs Melbourne | Sydney | 91% | ✓ | +17 |
| Gold Coast vs GWS | Gold Coast | 58% | ✓ | +20 |
Season: 51/72 (71%)
| Round | Correct | Total | Accuracy | Model |
|---|---|---|---|---|
| Opening Round | 3 | 5 | 60% | Logistic |
| Round 1 | 7 | 9 | 78% | Logistic |
| Round 2 | 6 | 7 | 86% | Logistic |
| Round 3 | 4 | 7 | 57% | Logistic |
| Round 4 | 5 | 8 | 63% | Logistic |
| Round 5 | 6 | 9 | 67% | Logistic |
| Round 6 | 8 | 9 | 89% | Phase NN |
| Round 7 | 7 | 9 | 78% | Phase NN |
| Round 8 | 6 | 9 | 67% | Phase NN |
| Season | 51 | 72 | 71% |
Model Notes
The draw is an unavoidable blind spot. No binary classifier picks draws — Collingwood vs Hawthorn finishing level on 93 is the kind of result that lands in the wrong column regardless of the probability assigned. The model had Hawthorn at 61%, which is honest given their form. The draw doesn’t reflect a bad read of the game.
Adelaide winning by 1 at Adelaide Oval is worth noting. The model had them at 61% against Port Adelaide — a derby at their home ground. They won, but by a single point. Last round Port Adelaide (as the away team at Adelaide Oval) beat a high-confidence Geelong by 30. The model keeps getting the Adelaide Oval direction right but the margins are erratic. Something about that ground produces volatile scorelines.
Carlton at 52% losing by 39 is the kind of outcome that skews perception. At that probability the model had no conviction, yet the result looked lopsided. St Kilda were the better team on the day and the low confidence was deserved — the model just happened to pick the wrong side of a near-50/50.
West Coast (65%) losing to Richmond is a concern. Richmond are one of the weaker sides in the competition on rolling stats, and the model still had them as clear underdogs at Optus Stadium. That Richmond won by 11 suggests either the Richmond numbers are improving faster than the rolling window captures, or West Coast’s home-ground advantage is weaker than the venue factor implies.
Leave a Reply