Edge = the difference between the model's estimated probability and the market-implied probability (1/odds). Bars show how many signals fall in each edge range; the line shows the win rate for each range. Note: win rate is not expected to increase linearly with edge — high-edge signals often target low-probability outcomes (draws, clean sheets) where the model sees large value relative to the market but the base win rate remains low. When a bucket contains fewer than ~50 signals, the win rate line may fluctuate significantly due to small sample size.
A signal is generated when our model identifies an edge of at least 2% between its estimated probability and the market-implied probability. Only predictions created before the match starts are included — no backtesting, no hindsight.
For informational and educational purposes only. Past statistical performance does not guarantee future results. This does not constitute financial advice.