Out-of-sample evidence shows the model identifies stronger contenders with meaningful consistency.
Historical evidence
Results with
the caveats intact.
A transparent view of how the Hoofs rankings performed on unseen historical races. Strike rate measures identification accuracy; it is not a claim of profitability.
The report order’s headline record
Across its 2022–2025 historical audit, the runner placed first in the Hoofs Ranking won 26.80% of races and achieved an official paid place in 55.12%.
51,605races
26.80%rank-1 win strike
55.12%rank-1 place strike
“Rank 1” is the runner placed first by the report’s PLACE-led ranking. Average historical odds and strike rates are descriptive; neither establishes positive expected return.
Same races, same years
The cleanest comparison uses the exact 13,041 races in 2024–2025 where every component model and the final Hoofs Ranking produced a valid first-ranked runner. The Hoofs Ranking led both win and place strike rates on that shared set.
| Ranking | Races | Win strike | Place strike | Avg WIN BSP |
|---|---|---|---|---|
| V11 | 13,041 | 24.03% | 52.52% | 6.81 |
| V12.2 | 13,041 | 24.98% | 54.44% | 6.27 |
| V17.2 | 13,041 | 24.39% | 53.59% | 7.18 |
| V25 | 13,041 | 25.24% | 54.50% | 6.25 |
| Hoofs Ranking | 13,041 | 25.76% | 55.43% | 5.95 |
The defensible conclusion
The Hoofs Ranking produced a better hit-rate order on identical races. Its first-ranked runners also had shorter average odds, so this should not be reinterpreted as proof of superior betting value.
How often the winner appeared near the top
Across the 2022–2025 V2 audit, winner containment rose steadily as more of the ranked set was included.
All-race figures shown. Small fields can make wider Top-N containment unavoidable; adjusted Top-3 coverage excluding fields of three or fewer was 59.48%.
Higher ratings produced stronger historical outcomes
The three confidence scales were frozen using 38,106 expanding walk-forward out-of-sample races from 2023–2025. Their useful property is the largely monotonic gradient: stronger ratings corresponded to stronger historical outcomes.
| Top 1 rating | Races | Rank-1 place | Rank-1 win | Winner Top-3 |
|---|---|---|---|---|
| 1 | 3,811 | 37.31% | 15.11% | 38.18% |
| 4 | 3,810 | 48.74% | 21.26% | 52.28% |
| 7 | 3,811 | 59.96% | 28.37% | 63.45% |
| 10 | 3,811 | 82.66% | 47.63% | 83.50% |
Important context
Rating 10 does not mean certainty. The historical Top 1 rating-10 group was also heavily concentrated in non-handicaps and shorter-priced, clearer market leaders.
Probability calibration
The report’s commission-aware minimum-back prices are based on separately calibrated WIN and PLACE chances. On pooled 2023–2025 out-of-sample runners, predicted and realised rates were closely aligned.
| Market | Runners | Mean predicted | Realised | Difference |
|---|---|---|---|---|
| WIN | 360,323 | 10.68% | 10.69% | +0.02 pp |
| PLACE | 359,438 | 29.33% | 29.68% | +0.35 pp |
Pooled calibration is encouraging but does not guarantee accuracy for every race, runner or probability band.
What these results do—and do not—say
Strike rate, average odds and calibration alone do not establish a profitable betting strategy.
V17.2 has clean evidence only for 2024–2025 because older sectional coverage is insufficient.
Higher confidence groups contained shorter-priced runners, so model strength and market obviousness overlap.
The Hoofs position
This page reports model behaviour honestly. Hoofs provides analysis and context; it does not promise outcomes, prescribe stakes or sell a guaranteed system.
See what the figures mean in practice
Learn how to interpret each layer before opening the daily report.