Out-of-sample accuracy
How often the model’s predicted side (1X2) actually wins on the test set.
–%
Model hit rate
Baseline: –%
Backtested on seasons 2018–2025
Model results & live metrics
Track how WinMatic performs against the closing odds: hit rates, log loss, Brier score and calibration by outcome. All computed on a proper train/test split.
Click “Load metrics” to fetch the latest summary from
/model-info.
Progress snapshot
Comparison vs a simple betting baseline and the market 1X2 probabilities.
Edge vs market
Hit rate lift
– p.p.
Extra percentage points over market or naive strategy.
Samples
Train / Test
– /
–
Total: – matches
Log loss
Log loss (1X2)
–
Lower is better. Measures how much probability mass the model wastes.
Brier score
Brier (1X2)
–
Squared error of 1X2 probabilities. Lower is better.
Results vs baseline
Quick comparison between WinMatic and a simple “bet on favourite / market” strategy.
- Load metrics to see a summary of how much the model outperforms baseline.
Technical metrics
Raw numbers from the training run behind
/metrics/model-info.
| Samples (train / test) | – / – (total –) |
|---|---|
| Hit rate – model | –% (1X2) |
| Hit rate – expected from probs | –% |
| Hit rate – baseline / market | –% |
| Edge vs baseline | – p.p. |
| Edge vs market | – p.p. |
| Log loss (1X2) | – |
| Brier score (1X2) | – |
Bankroll performance (PnL)
Flat 1-unit stakes on all bets with edge ≥ your threshold. Odds are estimated from logged edges.
5%
Total bets
–
Settled selections
Total profit
–
Units (flat staking)
ROI per bet
–
Average return
Cumulative profit over time
PnL history will appear here once bets are settled.
League ROI leaderboard
Flat-stake ROI per league based on settled bets.