Datasets
4
Transparency dashboard
Performance
Historical backtesting, live lifecycle tracking, and settled public predictions are deliberately separate layers.
Historical fixtures
841
Eligible predictions
721
Draw recall
Known limitation
Mean 1X2
50.645%
Balanced accuracy
41.3925%
Brier score
0.6113
Log loss
1.021
Historical backtesting
Four complete datasets. Historical evidence only, not live performance.
1X2 accuracy range
47% - 57%
Mean marker shown within the observed dataset range.
Mean balanced accuracy41%
Draw selection
Not promoted
V1 draw handling is weak, so draw picks are treated as a known limitation.
Dataset comparison
Detailed responsive table plus visual cards for all registered datasets.
Scottish Premiership 2024/2025
228 fixtures · 198 eligible
1X250%
Balanced38%
Over 1.579%
Scottish Premiership 2025/2026
228 fixtures · 198 eligible
1X257%
Balanced46%
Over 1.577%
Danish Superliga 2024/2025
193 fixtures · 163 eligible
1X247%
Balanced39%
Over 1.582%
Danish Superliga 2025/2026
192 fixtures · 162 eligible
1X249%
Balanced42%
Over 1.581%
| Dataset | Fixtures | Eligible | 1X2 | Balanced | Brier | Log loss | O1.5 | O2.5 | BTTS | Draw recall |
|---|---|---|---|---|---|---|---|---|---|---|
| Scottish Premiership 2024/2025 | 228 | 198 | 50% | 38.25% | 0.595771 | 1.004711 | 78.79% | 54.55% | 55.05% | 0% |
| Scottish Premiership 2025/2026 | 228 | 198 | 56.57% | 46.04% | 0.590954 | 0.990493 | 77.27% | 51.01% | 51.52% | 0% |
| Danish Superliga 2024/2025 | 193 | 163 | 47.24% | 39.38% | 0.62019 | 1.028811 | 82.21% | 62.58% | 65.03% | 0% |
| Danish Superliga 2025/2026 | 192 | 162 | 48.77% | 41.9% | 0.638384 | 1.060016 | 81.48% | 61.73% | 59.26% | 0% |
Live lifecycle tracking
Current fixture workflow only. These counts stay separate from historical backtesting and settled accuracy.
Lifecycle counts
Imported: 253AI review: 172Approved: 0Public: 21
Imported fixtures
253
Stored locally from supported leagues.
AI review queue
172
Reports waiting for validation.
Public analyses
21
Visible only after quality checks pass.
Settled records
Pending results
Measured after public matches finish.
Settled public predictions
This section starts only after a public prediction reaches full-time and can be checked against final results.
Awaiting completed public matches
No public prediction has reached settlement yet. Accuracy will appear only after real completed records exist.
Model experiment history
Analysis Engine V1
Baseline
Used for historical demonstration and internal tracking.
Analysis Engine V2
Rejected experiment
Not promoted after holdout and calibration review.
ML Baseline V1
Inconclusive
Useful research signal, not product-ready.
Public readiness
Research/development
Statistical estimates only, no profit claims.