> ## Documentation Index
> Fetch the complete documentation index at: https://offthepace.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Same car, same race who would be faster?

> A pace comparison that weight-corrects each driver against a synthetic teammate, isolating pure skill from strategy and track-position effects.

The fairest comparison in motorsport is teammates in identical cars but even teammates diverge on strategy and track position. The synthetic teammate comparison removes that divergence: instead of comparing raw lap times, it weight-corrects each driver's pace against a teammate's condition-adjusted lap time, isolating pure skill from strategy and position. The result answers a clean question: given the same car in the same race conditions, which driver would be faster?

## What it shows

Each driver carries a skill proxy built from like-for-like lap pairs:

* **Skill proxy**: the driver's pace minus the synthetic teammate's world-coordinate-adjusted pace, averaged over non-divergent laps. Negative means the driver is faster than their synthetic benchmark.
* **Quality weight**: confidence in each comparison lap. A higher weight means the lap pair is in similar conditions (tyre age, position, fuel load); low-weight laps are down-weighted in the average.
* **Strategic divergence**: laps where the driver and teammate are on different stint strategies are excluded, so non-pace factors don't contaminate the comparison.

<Note>
  A minimum of 3 non-divergent races is required, and the comparison is aggregated to season level.
</Note>

## How to use it

1. **Select a season** from the filter bar; the comparison reloads.
2. **Read the skill proxy sign** negative means the driver beat their synthetic benchmark.
3. **Weight your confidence by the quality weight** a small proxy on low-weight laps is weaker evidence than the same proxy on well-matched laps.
4. **Export the data** to compare teammates head-to-head on the condition-corrected scale.

## Reading the signal correctly

The benchmark is a **synthetic** teammate constructed from condition-matched laps, not a literal head-to-head average it is designed to strip out the strategy and position noise that makes raw teammate gaps misleading. A near-zero proxy means two drivers are genuinely matched once conditions are equalised.

<Warning>
  Drivers on persistently divergent strategies from their teammate yield few non-divergent laps, so their proxy rests on a thin, possibly unrepresentative sample. Read the quality weight before trusting an extreme value.
</Warning>

## Data source

The chart queries `int_synthetic_teammate` via DuckDB-Wasm in the browser, aggregated to season level.

| Column           | Meaning                                                                   |
| ---------------- | ------------------------------------------------------------------------- |
| `skill_proxy_s`  | Ego pace minus the synthetic teammate's adjusted pace (negative = faster) |
| `quality_weight` | Condition-similarity confidence for the comparison lap                    |
| `is_divergent`   | Whether the lap pair is on different strategies (excluded if true)        |

The data window covers **2018–2024**.

For the full model definition, see the [`int_synthetic_teammate` reference](/reference/models/int/int_synthetic_teammate).