Model Insights
Performance breakdowns from 611 backtested picks. Understand when the model fires and when to exercise caution.
Team Style Impact
TRANS teams generate more open play disposals — their players tend to run more and accumulate more kicks. STOP teams favour contested ball — fewer disposals but more tackles. When a TRANS player faces a STOP team's defence, the model gets a double signal.
Condition Performance
Position × Condition Heatmap
| Position | Dry | Wet | Roof |
|---|---|---|---|
| MID | 57.1% 175 picks | 50% 10 picks low n | 56.1% 132 picks |
| DEF | 52.4% 82 picks | 40% 5 picks low n | 56.9% 72 picks |
| FWD | 37.5% 56 picks EXCL | 50% 2 picks EXCL low n | 35% 40 picks EXCL |
| RUCK | 47.4% 19 picks low n | 0% 1 picks low n | 47.1% 17 picks low n |
FWD picks are excluded from filtered recommendations due to higher variance. RUCK/Wet and RUCK/Roof have insufficient sample sizes to draw conclusions.
Model Bias
The model historically over-predicts by +0.78 disposals on average. This creates a systematic UNDER edge — predictions are systematically set slightly high.
The optimised configuration reduces this to +0.09, near-perfect calibration. Use the Simulator to test optimised weights against historical picks.
When to Act
Key Data Findings
The five most actionable findings from 611 picks across Rounds 3–6.
All figures based on 611 verified picks across Rounds 3–6 · 2026 season. Past performance does not guarantee future results. Analytics are updated weekly. For methodology, see How It Works.