SIMULATED BACKTEST
How the engine did on past seasons (simulated)
Replays on historical prices. Not live results, and never mixed into the public record.Simulated These figures come from replaying the rules on seasons that had already been played. Live results are on the public record, recorded before kick-off. Past results do not predict future ones.
Tips (the decision rules replayed) · Simulated · seasons 2019/20 to 2024/25
The published-tip rules replayed walk-forward on each season, with the prices available before kick-off.
| Season | Tips | Won / lost | Hit rate | Avg odds | Flat-stake return | Max drawdown |
|---|---|---|---|---|---|---|
| 2020/21 | 163 | 95 / 68 | 58.3% | 1.66 | −3.3% | −10.4 units |
| 2021/22 | 57 | 39 / 18 | 68.4% | 1.68 | +14.4% | −5.7 units |
| 2022/23 | 105 | 66 / 39 | 62.9% | 1.68 | +5.4% | −7.5 units |
| 2023/24 | 43 | 29 / 14 | 67.4% | 1.66 | +12.1% | −5.0 units |
| 2024/25 | 156 | 99 / 57 | 63.5% | 1.63 | +3.8% | −10.9 units |
Value signals (experimental) · Simulated · seasons 2019/20 to 2024/25
The value-signal rule replayed on named bookmakers only, so there are few of them. Signals are never shown as tips.
| Season | Tips | Won / lost | Hit rate | Avg odds | Flat-stake return | Max drawdown |
|---|---|---|---|---|---|---|
| 2019/20 | 9 | 1 / 8 | 11.1% | 3.48 | −65.6% | −5.9 units |
| 2020/21 | 24 | 7 / 17 | 29.2% | 3.27 | −12.7% | −9.3 units |
| 2021/22 | 9 | 5 / 4 | 55.6% | 3.40 | +87.2% | −2.1 units |
| 2022/23 | 6 | 3 / 3 | 50.0% | 3.81 | +87.5% | −1.0 units |
| 2023/24 | 10 | 3 / 7 | 30.0% | 3.54 | −7.0% | −3.0 units |
| 2024/25 | 11 | 3 / 8 | 27.3% | 3.48 | −8.6% | −5.0 units |
Method: walk-forward (each season replayed with only what was known before it), flat 1-unit stakes, losses and voids counted. Simulated results can flatter a model; that is why only the live record counts.