Model Performance

Transparent accuracy tracking. Every prediction is measured, every miss is counted.

Forecasts scored
432
of 437 · 926 underlying lines
Skill score
0.107
over 400 forecasts · 0 = no better than the base rate
Log loss
0.6485
penalises confident mistakes

How to read this page

A forecast here is one decision — one match, one race, one stage, one title — not one line of probability.

Raw Brier scores cannot be compared across markets, because some questions are harder than others. A forecaster with no knowledge at all scores about 0.250 on Tennis Match Winner and about 0.188 on Tennis Set Betting — a lower Brier on the second says nothing about the model being better. The skill score removes that: it measures how much better the model does than knowing only how often this kind of call comes in. Zero means it adds nothing, one is perfect, and a negative number means it is worse than guessing.

Some markets are worked out from the same forecast as others. In football, double chance, draw no bet, other goal lines, team goals and correct score come from the same scoreline forecast as the match result and over/under 2.5. In Formula 1, the top 10 and the team-mate battles come from the same simulated races as the winner, podium and top 6. They are scored and listed below, but left out of the overall figures, the sport summary, the monthly trend and the model version comparison: counting several restatements of one forecast per event would let them drown out the forecasts that carry the information. A draw no bet forecast is not scored when the match is drawn.

For football match result and goals markets there is a second, much harder benchmark: the betting market's closing odds — the average across bookmakers just before kick-off — with the margin removed. vs market compares the model with those odds on exactly the same matches. A positive number means the model's probabilities were closer to what happened than the market's; a negative one means the market knew more. Beating the closing line consistently is rare, so a value slightly below zero is the normal result for a good public model.

Calibration

When the model says 70%, does it happen 70% of the time? The closer to the diagonal, the better calibrated. Head-to-head markets (match result, goals, match winner) and field markets (races, rallies, stages) are drawn separately: field markets are mostly low-probability lines and would otherwise hide everything else.

Skill over time

Monthly skill score by sport. Higher is better. Months with fewer than 5 scored forecasts are excluded.

By market

Raw Brier sits next to the baseline it is measured against; compare Brier only within a row, and skill across rows. The highlighted row is where the model adds the most over guessing.

Sport Market Field Forecasts Brier Baseline Skill Market vs market
Tennis Match Winner 2 398 0.2242 0.2500 0.103 0.2222 -0.010
Tennis Set Betting best 4 32 0.1661 0.1875 0.114 — —
Tennis Tournament Winner 32 2 0.6579 0.9385 — — —

Model versions

Each model change is tracked separately so improvements can be verified.

Version Sport Forecasts Brier Skill Log loss Status
tennis-elo-v2 Tennis 57 0.2277 0.089 0.6518 active
tennis-outright-v1 Tennis 2 0.6579 — 2.7122 active
tennis-elo-v1 Tennis 343 0.2233 0.107 0.6352 retired
Note: This is statistical information for educational and entertainment purposes. It does not constitute betting advice. Past model accuracy does not guarantee future performance. Proknoz has no affiliation with any betting operator.