Models / gradient-boosted-ensemble

Pace Normalization (LTOE)

Lap Time Over Expected — the normalized pace delta, per lap.

gradient-boosted-ensembleanalytics.strategy

How it works

A gradient-boosted model predicts the lap time a driver should have run given the context of the lap — fuel load, stint length, compound and tyre age, and track evolution. The residual, actual minus expected, is LTOE: negative means faster than expected. Both lap times are fuel-corrected. When both halves of the prediction interval are available, an 80% band around the expected lap time is published with it.

When to use it

Rank true pace: compare drivers on how much they beat their own expected time, rather than on raw lap time that mixes fuel, tyres and traffic.
Spot a stint going off: a driver drifting from negative to positive LTOE on the same tyres is losing performance.
Feed downstream signals such as the wet-to-slick crossover alert, which uses LTOE as an input.

Inputs

fuel-corrected lap time
stint context
compound & tyre age
track evolution

Output

drivers[].pace.ltoeSecActual minus expected lap time, seconds. Negative = faster than expected.

drivers[].pace.expectedLapTimeSec / actualLapTimeSecThe fuel-corrected expected and actual lap times.

drivers[].pace.categoryFAST | OPTIMAL | SLOW against an era-specific band.

drivers[].pace.expectedLapTimeP10Sec / P90Sec / paceUncertaintySec80% band around the expected time. All three are present together or not at all.

drivers[].pace.trafficAdjustedtrue after two or more consecutive green-flag laps in close traffic.

drivers[].pace.confidenceScale / confidenceFlagEarly-season damper; confidenceFlag is "2026_early" until three races of 2026 data exist, then "stable".

Example payload

An excerpt of one analytics.strategy delivery, trimmed to the fields this model produces. Values are illustrative. The full message shape is on the feed reference.

json

{
  "feed": "analytics.strategy",
  "lap": 35,
  "drivers": [
    {
      "number": "4",
      "tla": "NOR",
      "pace": {
        "ltoeSec": -0.21,
        "expectedLapTimeSec": 74.62,
        "actualLapTimeSec": 74.41,
        "category": "FAST",
        "trafficAdjusted": false,
        "confidenceFlag": "stable"
      }
    }
  ]
}

Use it

Subscribe to analytics.strategy with a webhook and read this model's fields from each delivery:

bash

curl -X POST https://api.racehooks.io/v1/webhooks \
  -H "Authorization: Bearer $RACEHOOKS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"webhookUrl":"https://your-app.com/f1","feedId":"analytics.strategy"}'

To try it without waiting for a race weekend, replay the recorded Canadian Grand Prix (2026-canada_r) from the quickstart.

How to read it

LTOE is measured against the driver's own context-adjusted expectation, so it answers "was this lap better or worse than it should have been?" rather than "who was fastest?". Use it averaged over a stint or a session; single laps are noisy. The 80% band tells you how wide the expected time could reasonably be — if it is absent on a lap, there is no band for that lap, which does not mean the estimate is certain. Treat early-season values with more caution while confidenceFlag reads "2026_early".

Limitations

It is not a direct driver-against-driver comparison: each driver is measured against their own expected time.
Designed for stint and session aggregation, not single-lap precision.
Confidence is damped at the start of a new regulation season until enough current-season races exist.
Published in Race and Sprint sessions only; if scoring fails on a lap, the pace block is omitted.

Calibration

An early-season damper applies until three races of current-season data exist, then lifts automatically. See the calibration reference for the shared pipeline.

FAQ

What does a negative LTOE mean?

The lap was faster than the model expected for that driver's fuel load, tyres and stint. Positive means slower than expected.

Is LTOE fuel-corrected?

Yes. Both the expected and the actual lap times are fuel-corrected before the difference is taken.

Related models

Pace Mode — Which pace mode a driver is running — push, hold, manage, warm-up or delta — per lap.
Tyre Cliff — P(tyre performance cliff within the next few laps), per driver per lap, plus live tyre health.
Qualifying Pace — Per-sector pace deltas vs the session median, per driver.

Get the data

This model is delivered on analytics.strategy. Subscribe by webhook or stream it over SSE. The quickstart walks through registering an endpoint and replaying a past session against it.

← All models