WinPIQ Weekly
Automatic trust report — last 7 days, straight from the ledger. No cherry-picking.
151
Settled predictions
64.2%
Accuracy
n = 151
+6.7%
ROI
n = 151
1.7 percentage points
Probability gapWinPIQ's probability minus the market's probability, in percentage points — only where a market price exists. · It's the plain-language size of the disagreement between WinPIQ and the market. · 56% versus 49% is a gap of 7 percentage points. Full definition →
0.215
Calibration accuracy — lower is betterHow closely stated probabilities match what actually happened, across many predictions. · A model can pick winners and still be poorly calibrated — this measures the second thing. · A lower score means the stated probabilities were closer to reality on average. Full definition →
0.791
Log loss — lower is betterA second calibration measure that penalises confident, wrong predictions more heavily than calibration accuracy does. · It catches overconfidence that a simpler calibration score can miss. · A confident, wrong call costs far more here than a cautious, wrong one. Full definition →
By market
| 1X2 | 58.7% | n = 63 |
| Both teams to score | 68.2% | n = 44 |
| Over/Under 2.5 | 68.2% | n = 44 |
Calibration this week
| 0.3-0.4 | חזוי 38.3% | נצפה 50.0% | n = 12 |
| 0.4-0.5 | חזוי 44.0% | נצפה 47.6% | n = 21 |
| 0.5-0.6 | חזוי 54.7% | נצפה 73.1% | n = 78 |
| 0.6-0.7 | חזוי 65.1% | נצפה 58.1% | n = 31 |
| 0.7-0.8 | חזוי 72.8% | נצפה 75.0% | n = 8 |
| 0.8-0.9 | חזוי 81.3% | נצפה 0.0% | n = 1 |
Biggest miss
The highest-probability miss — never hidden
Strongest hit
The hit with the largest edge vs the market
AI vs experts · verified wins
Experts: 0 verified predictions. Verified user wins: 0. Zero is zero.