> ## 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.
# fct_driver_skill_features
> Race-grain driver skill feature table for the driver skill extraction model.
Part of the [Marts](/transform/families/marts) family.
Race-grain driver skill feature table for the driver skill extraction model. One row per (race\_year, race\_id, driver\_id). Aggregates clean-lap residuals, synthetic-teammate deltas, constructor pace indices, and circuit context. Clean-lap filter: correction\_weight = 1.0 AND anomaly\_class NOT IN ('mistake','conditions') AND is\_rain\_lap = FALSE.
Breaking a column type on this model fails the dbt build - this is the enforced handoff contract with `ml/`.
## Lineage
```mermaid theme={null}
flowchart LR
dim_circuits --> fct_driver_skill_features
dim_constructors --> fct_driver_skill_features
dim_drivers --> fct_driver_skill_features
int_constructor_structural_pace --> fct_driver_skill_features
int_lap_anomaly_flags --> fct_driver_skill_features
int_lap_residual_decomposed --> fct_driver_skill_features
int_synthetic_teammate --> fct_driver_skill_features
style fct_driver_skill_features fill:#e40404,stroke:#e40404,color:#fff
```
## Overview
| Property | Value |
| ----------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Layer | `marts` |
| Family | [Marts](/transform/families/marts) |
| Contract enforced | ✅ Yes |
| Upstream | [`int_lap_residual_decomposed`](../int/int_lap_residual_decomposed), [`int_lap_anomaly_flags`](../int/int_lap_anomaly_flags), [`int_synthetic_teammate`](../int/int_synthetic_teammate), [`int_constructor_structural_pace`](../int/int_constructor_structural_pace), [`dim_circuits`](../dim/dim_circuits), [`dim_drivers`](../dim/dim_drivers), [`dim_constructors`](../dim/dim_constructors) |
| Model-level tests | `expect_column_values_to_be_between` |
**1 tests** [guard](/transform/ci/overview) this model - 1 generic, 0 singular.
## Columns
| Column | Type | Nullable | Tests | Description |
| ------------------------------------ | --------- | -------- | ----- | ---------------------------------------------------------------------------------------------------------------- |
| `driver_race_id` | `varchar` | No | | Surrogate PK hash of (race\_year, race\_id, driver\_id) |
| `driver_id` | `varchar` | No | | FastF1 three-letter driver code (e.g. VER, HAM) |
| `race_year` | `integer` | No | | Season year |
| `race_id` | `varchar` | No | | Race identifier (e.g. 2021\_19) |
| `constructor_id` | `varchar` | Yes | | Team name as used in FastF1 |
| `circuit_key` | `varchar` | Yes | | Stable circuit identifier matching dim\_circuits |
| `pu_family` | `varchar` | Yes | | Power unit family (e.g. mercedes\_pu, honda\_pu) |
| `driver_debut_year` | `integer` | Yes | | First season in the dataset |
| `driver_residual_mean_s` | `double` | Yes | | Mean driver skill residual over clean laps. Negative = faster than field. |
| `driver_residual_stddev_s` | `double` | Yes | | Stddev of driver skill residual over clean laps |
| `driver_residual_median_s` | `double` | Yes | | Median driver skill residual over clean laps |
| `clean_lap_count` | `bigint` | No | | Number of laps passing the clean-lap filter |
| `driver_skill_proxy_mean_s` | `double` | Yes | | Weighted mean of synthetic-teammate delta over clean laps. Positive = ego driver faster than synthetic teammate. |
| `pair_quality_weight_mean` | `double` | Yes | | Mean pair quality weight (higher = more consistent comparison) |
| `teammate_available_flag` | `boolean` | No | | TRUE if a teammate comparison was available for at least one lap |
| `strategic_divergence_share` | `double` | No | | Fraction of clean laps where strategic\_divergence\_flag = TRUE |
| `constructor_power_pace_index_final` | `double` | Yes | | EW-smoothed constructor power-sector pace index at race end |
| `constructor_aero_pace_index_final` | `double` | Yes | | EW-smoothed constructor aero-sector pace index at race end |
| `constructor_index_confidence_final` | `double` | Yes | | Confidence score (0–1) for constructor pace index at race end |
| `race_wet_flag` | `boolean` | No | | TRUE if any rainfall was detected during the race |
| `circuit_energy_index` | `double` | Yes | | Track energy index from dim\_circuits (proxy for power sensitivity) |
| `circuit_abrasiveness_index` | `integer` | Yes | | Track abrasiveness index from dim\_circuits (proxy for tyre wear) |
| `total_race_laps` | `bigint` | No | | Total laps recorded for this driver in this race |
| `ml_eligible_lap_share` | `double` | No | | Fraction of driver's laps with usable\_for\_modelling = TRUE |