Was the pit stop on the optimal lap?

A Gantt of every stint in a race, marking where each tyre hit the cliff, where the model says the optimal stop was, and how much time a late stop cost.

The pit strategy Gantt lays every driver’s race out as a row of coloured stint bars, then overlays two things a lap chart can’t show you: the lap each tyre actually crossed the degradation cliff, and the lap the model says they should have stopped. The gap between those two measured in seconds of opportunity cost is the verdict on each strategy call.

What it shows

Each stint is a horizontal bar spanning its actual race laps:

• Bar fill: the tyre compound for that stint.
• Border colour the verdict: optimal, overran, early, undercut forced, or unknown (no fitted degradation curve for that compound; shown explicitly rather than as a silent null).
• Cliff onset (dashed line inside the bar): fct_stint_features.cliff_lap_in_stint, the lap within the stint where the compound crossed the degradation threshold.
• Optimal pit window (triangle below the bar): int_pit_strategy_value.optimal_pit_lap_in_stint, the argmin of Total_Cost(L) over every candidate lap in the horizon wear on the set you are on, wear on the next set, and the pit-lane loss discounted by the chance a safety car has already appeared. Not stopping at all is one of the candidates.
• Opportunity cost: Total_Cost(actual) minus Total_Cost(optimal), in seconds so stopping too early carries a cost too, not only overrunning.

unknown means the stint’s compound has no fitted degradation curve at that circuit and season, so there is nothing to minimise over 266 stints of 7,129. It is shown explicitly rather than being forced into a verdict.

How to use it

1. Select a race from the filter bar; every driver’s stint sequence loads as Gantt rows.
2. Find the dashed cliff line inside each bar if it sits well before the bar’s end, the driver ran deep into degradation.
3. Compare the cliff line to the optimal-window triangle a triangle far left of the actual stop end signals a late stop; far right signals a stop taken earlier than degradation alone justified.
4. Read the opportunity cost on overran stints to size the loss; sort drivers to see who left the most time on track.

Reading the signal correctly

The optimal lap is a model counterfactual computed from the fitted degradation curve, the circuit’s pit-lane loss and its safety-car rate it assumes the rest of the race plays out as observed, and that exactly one stop happens inside the horizon. The horizon is this stint plus the next, so the verdict grades the lap the stop landed on, not the number of stops the team chose. Across seven seasons the median stop comes one lap earlier than the model’s optimum, which is roughly what track position is worth: a stop that looks early may have been the correct response to something a degradation-only model does not price.

Neither overran nor early is automatically a mistake: teams trade lap-time loss for track position, safety-car optionality, or tyre-offset at the end. Read the opportunity cost as the pace price paid, not as proof of an error.

Data source

The chart joins fct_stint_features (stint windows and cliff lap), fct_lap_residuals (lap windows), and int_pit_strategy_value (the optimal-lap counterfactual), via DuckDB-Wasm in the browser.

Column	Meaning
cliff_lap_in_stint	Lap within the stint where the compound crossed the cliff (fct_stint_features)
optimal_pit_lap_in_stint	Model-optimal stop lap, in valid laps on the set (int_pit_strategy_value)
opportunity_cost_s	Total_Cost(actual) - Total_Cost(optimal); median 5.5 s, p95 83 s
strategy_verdict	optimal / overran / early / undercut_forced / unknown

This is a single-race view across 2018–2024 data.

For the full model definitions, see the int_pit_strategy_value reference and the fct_stint_features reference.