Models / gradient-boosted-ensemble

Tire Compound

The next compound a pitting driver takes, as probabilities.

gradient-boosted-ensembleanalytics.tire-strategy

How it works

A five-class classifier over soft, medium, hard, intermediate and wet predicts the compound fitted at the driver's next stop. It is only scored for drivers approaching a pit window, so the feed stays sparse. Intermediate and wet are published as one wet-weather class tuned to catch wet stops; on a dry track that class is held near zero so a dry compound stays on top.

When to use it

Strategy screens: show which tyre a car approaching its window is likely to take.
Markets on the next compound choice and related strategy props.
Combine with pit-stop timing and stop count for a full picture of a driver's remaining strategy.

Inputs

current compound & age
circuit degradation profile
current weather
race phase

Output

drivers[].nextCompoundProbability.soft / medium / hard / wetWeatherPer-compound probabilities summing to 1.

drivers[].nextCompoundProbability.dryCompoundProbabilitiessoft / medium / hard renormalised to sum to 1 among dry tyres only.

drivers[].pitProbabilityThe per-lap pit probability that qualified the driver for this message.

Example payload

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

json

{
  "feed": "analytics.tire-strategy",
  "lap": 22,
  "drivers": [
    {
      "number": "4",
      "tla": "NOR",
      "pitProbability": 0.29,
      "nextCompoundProbability": {
        "soft": 0.08,
        "medium": 0.21,
        "hard": 0.7,
        "wetWeather": 0.01,
        "dryCompoundProbabilities": {
          "soft": 0.08,
          "medium": 0.21,
          "hard": 0.71
        }
      }
    }
  ]
}

Use it

Subscribe to analytics.tire-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.tire-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

Read the four main probabilities as one distribution over the next tyre. In dry running, use dryCompoundProbabilities to compare soft, medium and hard without the wet class. If no driver is approaching a window on a lap, nothing is sent — silence means no one qualified, not a feed fault.

Limitations

Sparse by design: a driver appears only when approaching a pit window.
Intermediate and wet are merged into one wet-weather class.
There is no rain-forecast input; weather comes from current conditions.
Race and Sprint sessions only. Values are rounded to 5 decimals.

FAQ

Why are some laps missing from the tire-strategy feed?

The feed is sparse. A driver is included only when approaching a pit window, and nothing is sent on a lap when no driver qualifies.

How are intermediates and wets reported?

As one wetWeather class. On a dry track it is held near zero and dryCompoundProbabilities gives the soft, medium and hard split on its own.

Related models

Pit Stop Prediction — Probability of a green-flag pit stop on the next lap; stops under safety car, VSC or red flag are not counted.
Pit Stop Count — P(one / two / three+ total stops) for the race, per driver — re-estimated every lap.
Tyre Cliff — P(tyre performance cliff within the next few laps), per driver per lap, plus live tyre health.

Get the data

This model is delivered on analytics.tire-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