// MODEL BACKTEST · 730D

Prediction model accuracy

Our Elo strength model powers the win probabilities on every team head-to-head. Here's how it actually performed — each match graded against the prediction the model would have made from only the games before it, over the last 730 days.

61.8%
favorite won
1,261/2,042
correct predictions
0.2261
Brier score (lower is better)
Calibration

When the model says it's X% confident, how often is it actually right? A well-calibrated model tracks the diagonal.

Confidence Actual win rate Predictions
50–60% 54.6% 1110
60–70% 64.2% 564
70–80% 76.5% 247
80–90% 86.2% 109
90–100% 83.3% 12
Recent predictions
Date Model favorite Confidence Result
Sep 26 magic over GamerLegion 55.9% ✓
Sep 26 Ninjas in Pyjamas over SINNERS 54% ✓
Sep 26 K27 over EYEBALLERS 51.9% ✓
Sep 26 Alliance over 3DMAX 53.5% ✗
Sep 25 GamerLegion over K27 51.5% ✓
Sep 25 magic over EYEBALLERS 61.6% ✓
Sep 25 3DMAX over Ninjas in Pyjamas 57.1% ✗
Sep 25 Alliance over SINNERS 62.6% ✗
Sep 23 Luminosity over 3DMAX 54.2% ✗
Sep 22 Inner Circle over 3DMAX 61.5% ✗
Sep 22 B8 over Luminosity 68% ✗
Sep 20 Vitality over Aurora 66.4% ✓
Sep 20 Vitality over FURIA 64.1% ✓
Sep 19 FURIA over MIBR 71% ✓
Sep 19 Vitality over Aurora 69% ✗
Sep 19 3DMAX over M80 52.5% ✓
Sep 19 MOUZ over FURIA 59.7% ✗
Sep 19 BBL over Luminosity 58.3% ✗
Sep 19 MIBR over NRG 68.3% ✓
Sep 18 Vitality over FURIA 66.3% ✓
Sep 18 3DMAX over INFINITE 51% ✓
Sep 18 Inner Circle over EYEBALLERS 70.7% ✗
Sep 18 INFINITE over EYEBALLERS 57.6% ✗
Sep 18 BBL over Liquid 51.7% ✓
Sep 18 BBL over EYEBALLERS 63.6% ✓

Not betting advice. Every percentage on this page describes what past results imply, not what will happen. The model knows match results and nothing else — it cannot see a roster change until the new lineup has played, and it has no opinion on map vetoes or on whether a match is online or on LAN. We publish its hit rate and calibration so you can judge it for yourself. No bookmaker odds or affiliate links appear on this site.