Fill Probability Model
Logistic regression fitted on live fill/skip data from ORB limit order entries. Each point is a real trade signal, green if the limit filled, red if it skipped. The curve is the model's predicted fill probability as a function of breakout velocity. For the full context on why this model exists and where it is headed: The Skip Model Needs Work, The Backtester Was Entering One Bar Too Early.
Most backtesting platforms assume 100% fills on limit orders, or use a flat slippage estimate. Neither reflects reality. This model is calibrated on actual trade data and updates daily as new signals come in. The approach generalizes beyond futures. Any limit order entry on any instrument and any timeframe faces the same question: given how fast price moved past your level, what is the probability your order got touched? The features scale, the model structure stays the same.
Collecting live data since Jun 24, 2026
Velocity vs Fill Probability
Velocity = how far the bar close moved past the ORB boundary relative to the bar's range. Higher velocity means price broke out harder and is less likely to retrace to the limit.
Historical Velocity Distribution
Velocity of -- ORB signals from 14 years of NQ data. The fitted model predicts a -- fill rate across this full history — compared to the backtester's default assumption of 100%.
Model Details
Current backtester model: step function. P(fill) = 1 if bar low retraces within 0.5pts of boundary, else 0. On live data this predicts fill for every signal.
Logistic (velocity): P(fill) = sigmoid(...). 50% fill probability at velocity = --.
This model updates as new trades come in. More data will tighten the fit and reveal whether bar range or other features add predictive power beyond velocity alone.