> ## 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.

# Which circuits suit each car?

> A heatmap of how each constructor over- or under-performs at specific circuits relative to their own season average the circuit × constructor interaction term.

Some cars are built for Monza's straights, others for Monaco's corners. This heatmap measures that fit directly: for each constructor and circuit, how much they over- or under-perform relative to **their own** season average the circuit × constructor interaction term. Negative (green) means the car is better than its season average at that circuit; positive (red) means worse.

## What it shows

Each cell is one constructor × circuit interaction:

* **Interaction term**: the deviation of the constructor's race-average pace from their own season mean at a circuit, estimated with exponentially weighted smoothing and Bayesian shrinkage toward zero (prior weight = 3 races).
* **Sign convention**: negative = the constructor outperforms their season average at that circuit (green); positive = underperforms (red). Ferrari at Monza, for instance, typically shows a favourable negative interaction from their low-drag setup.
* **Average shown**: cells display the multi-season average of the race-level interaction, requiring at least 2 races at the circuit to appear.

Constructors are sorted by overall average interaction (lowest first); circuits are sorted alphabetically.

<Note>
  Empty cells indicate fewer than 2 race visits too few to estimate an interaction, not a zero interaction.
</Note>

## How to use it

1. **Scan a constructor's row** for their circuit strengths (green) and weaknesses (red).
2. **Scan a circuit's column** to see which cars the track flatters.
3. **Look for low-drag / high-downforce patterns** a car green at Monza and red at Monaco has a clear aero bias.
4. **Remember the shrinkage** a faint colour on a 2-race cell is mostly prior.

## Reading the signal correctly

The interaction is **within-constructor**: it compares a car to its own season form, so it is not a pace ranking. A midfield car can show a strong positive interaction at a circuit that suits it without ever being quick in absolute terms.

<Warning>
  With a prior weight of 3 races, circuits a team has visited only twice are heavily shrunk and faint by design. Read strong colours on high-visit cells, and treat thin cells as directional only.
</Warning>

## Data source

The chart queries `int_circuit_x_constructor_interaction` via DuckDB-Wasm in the browser.

| Column                          | Meaning                                                               |
| ------------------------------- | --------------------------------------------------------------------- |
| `interaction_s`                 | Constructor's circuit pace vs their own season mean (the cell colour) |
| `n_races`                       | Races at the circuit; gates display at `≥ 2`                          |
| `constructor_id` / `circuit_id` | The heatmap axes                                                      |

The data window covers **2018–2024**.

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