Vandel Gokart - Live Tracking and history

Stat Windows

hour/day/week/month/year/all time are independent windows. A kart can rank differently across windows because recency, sample size, and the current field differ.

Math Model

heat_score = min(1.2, 0.7 * position_score + 0.3 * lap_quality + special_bonus)

posterior_mean = (weighted_heat_score_sum + prior_strength * global_mean) / (weighted_heats + prior_strength)

risk_adjusted = max(0, posterior_mean - uncertainty_lambda * standard_error)

confidence_blend = 0.65 * heat_confidence + 0.35 * lap_confidence

prediction = clamp(0, 1.35, risk_adjusted + trend_adjustment + exploration_bonus)

All 3 lap types are included in the model: normal-layout laps, Alternative layout laps, and jokerlaps. Normal laps drive the base score, while Alternative layout and jokerlaps add weighted signal.

Recent weighted form shifts the score with a bounded trend term, so a kart can rise or fall after a tire swap or a change in the driver field.

Alternative-layout and jokerlap-heavy heats add a mild field-strength bonus because pro formats often include these race elements.

Joker-rule compliance also factors in: heats where drivers tend to take exactly one joker and avoid lap 1 joker receive a positive pro-format signal.

Mixed layout signal: heats containing both normal and alternative-layout laps get a small additional strength weight.