How BanksPitch works
No jargon required. This page explains exactly what the numbers on the site mean.
Evidence score
A summary of how strongly the current statistical evidence supports the selected market. Higher means more supporting evidence under the model. It is not a bookmaker price and not a guarantee.
Data confidence
HIGH means both teams have a strong recent sample and enough venue-specific data. MEDIUM means useful but less complete evidence. LOW means the sample is limited and needs extra caution.
BanksPitch selection levels
The evidence score has a different job at each level.
Interest
Tracked for calibration and research, but not presented as a BanksPitch Pick.
BanksPitch Pick
The current minimum score for a selection to count as a BanksPitch Pick.
Strong Signal candidate
Must still pass every additional data-quality rule below.
⭐ Strong Signal rules
A Strong Signal is a quality gate, not a quota.
1. Evidence score of at least 80/100
The selected market must clear the published evidence threshold.
2. HIGH data confidence
The model must have a strong enough recent and venue-specific sample.
3. At least 8 recent matches for each team
Neither side can qualify on a tiny recent sample.
4. At least 3 supporting indicators
The score needs several separate pieces of evidence behind it, not one isolated stat.
5. No major opposing indicator
If the model identifies a meaningful conflict with the selection, the fixture does not qualify.
What the match statistics mean
The wording on match pages is designed to tell you both the stat and the sample.
Over 2.5 goals in recent matches: 60%
60% of that team’s recent sample finished with at least 3 total goals. If the sample is 10 matches, that means 6 of 10.
Both teams scored in recent matches: 70%
Both sides scored in 70% of that team’s recent sampled matches. This is a historical rate, not a prediction by itself.
Team scored in recent matches: 80%
The team scored at least one goal in 80% of its recent sampled matches.
Average goals scored: 1.8
The team scored an average of 1.8 goals per match across the displayed recent sample.
Home / away sample
These numbers only use matches played at the same venue type as the upcoming fixture — home form for the home team and away form for the away team.
Supporting and opposing evidence
BanksPitch deliberately shows both sides.
✓ Supporting evidence
Statistics that push the model toward the selected market, such as strong recent Over 2.5 rates or scoring trends.
⚠ Opposing evidence
Statistics that conflict with the selection. They are shown instead of being hidden so you can judge the trade-off yourself.
Why there is a separate Model Health page
Everyday match decisions should not be buried under technical measurements.
Prediction History records what was actually stored before kickoff. Model Health then measures those settled predictions using hit rate, calibration gap and Brier score. These are useful for improving the model, but they stay on their own page so the main match experience remains understandable.