23 836
Predictions checked after the result.
Period: last 120 days · Glicko-2 and math models (no live signals)AI scoreboard
Intelligent AI scoreboard for sports analytics
The AI scoreboard shows automated match analytics, outcome probabilities, live signals and post-result model checks.
First we look at strong outcomes and the market's agreement with the model; exact score remains a reference market.
Public metric
PropickAI prediction accuracy
Available to everyone. Informational. The model is calibrated to the market; this is prediction accuracy, not betting profitability.
59.30%
Overall hit-rate on the main outcome.
14 134 hits out of 23 836
Model confidence
Average prediction
Actual (hit-rate)
Predictions
50–60%
53.0%
58.7%
12 510
60–70%
64.6%
65.3%
4 241
70–80%
74.4%
72.3%
2 239
80%+
84.3%
88.4%
1 246
Calibration: the higher the model confidence on the left, the higher the real hit-rate on the right. This means the model's probabilities match reality.
Performance by sport
Hit-rate is the share of outcomes correctly predicted by the model. ROI is the historical return on the model's value picks (flat 1 unit, over the period). Informational, backtest, not a profit guarantee.
Sport
Predictions
Hit-rate
Value picks ROI
Tennis
4 926
65.6%
n/a
Football
4 269
47.0%
n/a
Basketball
1 131
66.3%
n/a
Hockey
357
56.0%
n/a
Volleyball
267
66.7%
n/a
A hit-rate above the market does not guarantee positive ROI: at low odds even high accuracy can go negative (e.g. tennis). Profit holds where the model agrees with the market (basketball).
Method
How our AI works
PropickAI compares sports statistics, odds, Glicko-2, live signals and the results of completed matches. The AI scoreboard shows which models give a stable signal, where the outcome probability is confirmed by the market and how predictions pass the post-game check.