No season stats entered โ falling back to the last-5 sample as the season baseline too. Calibration will read low, same as a genuinely thin sample.
Fill in the Inputs tab and hit Compute Prediction to see markets here.
Compute a prediction first to see the verdict.
โ
โ
A season-long Bayesian baseline (8-game prior) anchors each team's strength, and last-5 form becomes a capped ยฑ30% recency nudge on top of it โ not the whole signal. Rho and tau both adapt to league scoring rate and blend toward a full bivariate-Poisson model as your sample size grows, with the resulting confidence shown as a calibration score. Defensive form is benchmarked against the pure opposing league average, not a self-referential blend. Elo uses PitchPulse's own logistic curve.
Calibration reflects how much real season+recency data you've fed the model relative to its 15-match reference point โ low calibration means treat the output as a rough prior, not a sharp forecast.
Best Bet in the top bar is simply the single highest-probability market reading across 1X2 / Over-Under / BTTS โ it is not an edge-vs-market-odds calculation, since this tool doesn't take bookmaker prices as input.