dbt Models

Auto-generated reference for all 94 dbt models.

Auto-generated reference for all 94 dbt models.

​

Staging

Model	Description
stg_circuit_info	Corner geometry from FastF1 circuit_info.corners.
stg_events	Race-level events (damage, retirements, penalties) sourced from the raw_dim_events seed.
stg_lap_tyre_qa	Bronze tyre QA (WI-05: F24, F25), one row per race lap; int_stint_geometry reads stint_number / tyre_life / compound from here instead of raw bronze.
stg_laps	Cleaned lap-level data from FastF1 race sessions.
stg_laps_qualifying	Cleaned qualifying lap data (Q1/Q2/Q3), mirroring stg_laps for qualifying sessions.
stg_pits	One row per pit stop: a lap carrying PitInTime.
stg_race_control	The FIA race-control message feed.
stg_results	Official classified race results from FastF1 session.results.
stg_results_qualifying	Official classified results for the qualifying session.
stg_sector_times	Per-sector timing unpivoted from stg_laps.
stg_session_status	Session lifecycle events from FastF1 session.session_status (Inactive / Started / Aborted / Finished / Finalised).
stg_session_status_qualifying	Session lifecycle events for the qualifying session, mirroring stg_session_status.
stg_telemetry	High-frequency car telemetry (~10 Hz) from FastF1, one row per sample.
stg_telemetry_position	The channel split.
stg_track_status	SC/VSC/yellow/red timeline from FastF1 session.track_status.
stg_tyre_allocations	Pirelli weekend compound nominations for 2019-2025: which C-code each of hard/medium/soft actually was, per race.
stg_weather	Weather snapshot nearest each lap start.

​

Intermediate

Model	Description
int_cell_power_law_fit	WI-17: the pooled fit and training target, one per race x compound hardness rank.
int_circuit_x_constructor_interaction	One row per (race_year, race_id, constructor_id).
int_compound_cliff_predicted	Tyre compound degradation prediction using fitted survival-analysis parameters.
int_constructor_car_fe	De-biased constructor car pace (constructor×race fixed effect), one row per (race_year, race_id, constructor_id).
int_constructor_car_fe_isolation	Driver-isolation car term: a thin reader over data/fits/constructor_car_fe_isolation.parquet, fitted by fit_constructor_car_fe.py —panel isolation (mean y_s ~ 1 | driver_id + constructor_race on driver-race means of the pre-cliff Ω laps, separately in each era; fit_method constructor_car_fe_isolation_hdfe_v2), re-centred to a lap-weighted mean of zero inside each race.
int_constructor_deg_sensitivity	One row per (race_year, constructor_id, compound).
int_constructor_structural_pace	One row per (race_year, race_id, constructor_id).
int_constructor_structural_pace_qualifying	One row per (race_year, race_id, constructor_id).
int_corner_drift_from_early_stint	Item 02h.
int_corner_metrics	Per-corner telemetry metrics.
int_corner_skill_residuals	One row per lap per named corner.
int_dirty_air_tax_component	One row per lap.
int_driver_circuit_affinity	One row per (driver_id, circuit_id).
int_driver_circuit_era_affinity	One row per (driver_id, circuit_id, era_key).
int_driver_isolation_lap_pace	Driver isolation with the car removed: pace_isolated_gain_s = -(y_s - car_iso_s) per Ω lap, positive = faster than the field on an equal car, tyre state, traffic and fuel.
int_driver_isolation_lap_panel	Driver isolation step 0: the clean-lap panel Ω and the field-centred isolated time.
int_driver_isolation_lap_values	Driver isolation tier 1 per Ω lap: pure_skill_gain_s = p (pace_isolated_gain_s) plus the declared season offset pure_season_offset_gain_s (W33; var isolation_pure_season_offset_gain_s, currently empty so 0 in every season).
int_driver_isolation_stint_tyre	Driver isolation, tier 2’s stint layer: each stint’s pace line (slope of p on tyre age over its clean early/mid, non-recovery laps; NULL below var(‘isolation_min_stint_laps’) or with Sxx = 0, never NaN), the leave-driver-out peer slope (same race and compound, else the season pool), the car’s degradation character (int_constructor_deg_sensitivity), the tactical slope m_s and its SE, the cliff reference kappa_ref, and the per-season noise (sigma_w, rho1, f) every downstream SE uses.
int_driver_race_skill_loro	Equal-car driver skill, race grain (race_year, race_id, driver_id).
int_driver_season_ratings	One row per (driver_id, season).
int_era_normalized_driver_rating	One row per (driver_id, season).
int_event_corrections	Safety car, red flag, and other race event corrections.
int_field_pace_curve	Field-level baseline pace curve.
int_lap_air_state	Air state classification and thermal load from dirty air.
int_lap_anomaly_flags	Anomaly detection flags for data quality control.
int_lap_corner_drift	Item 02h.
int_lap_corner_inputs	02c (Tier 2).
int_lap_fuel_state	Fuel burnoff and weight correction.
int_lap_fuel_state_qualifying	One row per qualifying lap.
int_lap_normalized_pace	Phase C: lap pace with the dirty-air cost removed, so laps run in traffic stay usable for wear/cliff fitting instead of being discarded.
int_lap_proximity	True pairwise track gaps from the POSITION channel, reduced to per-lap scalars.
int_lap_residual_decomposed	One row per lap.
int_lap_residual_decomposed_qualifying	One row per qualifying lap.
int_lap_residual_stint_detrend	Per-stint linear drift of driver_skill_residual_s against lap_in_stint, fit on pre-cliff laps only (cliff_onset_passed = FALSE) via OLS REGR_SLOPE.
int_lap_telemetry_aggregates	Per-lap telemetry features for the tyre-cliff model.
int_lap_thermal_proxy	Cumulative push load from tyre thermal management.
int_pit_loss_circuit	Empirical circuit-specific pit-lane loss prior, read by int_pit_strategy_value in place of the largely-imputed circuit_reference constant.
int_pit_strategy_cost_curve	One row per (stint_id, horizon_scope, candidate_pit_lap_offset): the Total_Cost(L) surface int_pit_strategy_value’s header described for two years and never computed.
int_pit_strategy_value	One row per stint.
int_power_law_lap	WI-17: the one lap set every power-law fit, the Step-0 identification check (ml/src/powerlaw_ceiling.py) and the held-out grading in ml/src/powerlaw.py read.
int_qualifying_decomposed	One row per qualifying lap.
int_qualifying_driver_summary	02b (Tier 1).
int_qualifying_push_laps	One row per qualifying lap (gated or not, so consumers filter rather than lose the row set).
int_qualifying_segments	One row per (race_year, race_id, quali_segment).
int_sc_hazard_history	Safety-Car / Virtual-Safety-Car base rate per circuit (venue slug) as a hazard per racing lap, POINT-IN-TIME AS OF THE START OF EACH SEASON.
int_sector_residual_decomposed	One row per lap per sector (3 rows/lap).
int_stint_end_regime	Why did this stint end?
int_stint_geometry	Foundation for all downstream window functions.
int_stint_power_law_fit	WI-17: y = intercept + alpha * age^beta per stint, beta profiled on 0.20-3.00 (step 0.05) with REGR_SLOPE / REGR_INTERCEPT / REGR_R2.
int_synthetic_teammate	Synthetic teammate for driver skill extraction.
int_track_evolution	Track temperature and grip evolution over race.
int_track_geometry	Circuit geometry (gradient + curvature) from raw X/Y/Z telemetry, binned to 25m distance intervals.
int_tyre_surface_vs_bulk_decoupling	One row per lap (post-cliff only).

​

Marts

Model	Description
dim_events	Race-level event dimension.
fct_cliff_prediction_features	Lap-grain feature table for the tyre cliff XGBoost model.
fct_driver_isolation_lap	Driver isolation per Ω lap with the rolling rating: the lap’s three ratings, its tier-3 peer aggregates and identity split, and a trailing window per rating (the Ω laps of the same stint with lap_number in [l - 4, l]; never forward, T44), each NULL below var(‘isolation_window_min_laps’) contributing laps, with SE, lambda, a shrunk value, confidence_pct and trust_label.
fct_driver_isolation_pair_lap	Driver isolation tier 3 per pair of cars: for every Ω lap, each peer on the same race, lap number and compound with tyre age within var(‘isolation_peer_age_tolerance’), teammates included.
fct_driver_isolation_race	Driver isolation per driver-race: the three ratings rolled up with lap weighting (pair-lap weighting for relative), SEs that add each stint’s slope error in quadrature, lambda, shrunk value, confidence_pct and trust_label, the tier-3 identity split, and context.
fct_driver_isolation_stint	Driver isolation per stint and phase: one row per (stint_id, stint_phase) with stint_phase in early / mid / cliff / recovery, plus an ‘all’ row per stint (built with UNION ALL, never GROUPING SETS).
fct_driver_skill_features	Race-grain driver skill feature table for the driver skill extraction model.
fct_ghost_car_pace	One row per (ego_driver, host_constructor, race, lap).
fct_ghost_race_finish	One row per (host_constructor, ego_driver, race).
fct_lap_residuals	Lap-grain analytics table.
fct_power_law_training	WI-17: the training frame for ml/src/powerlaw.py, one row per fit_eligible race x hardness rank cell.
fct_stint_features	One row per stint.
fct_telemetry_deltas	Driver-vs-teammate corner-metric deltas.
mart_corner_skill_driver	Driver corner skill vs a LORO (leave-one-race-out) same-car baseline.
mart_degradation_history_envelope	Pre-aggregated historical stint envelope for the Degradation Simulator.
mart_driver_season_summary	One row per driver-season, shaped for plain-language explanations (“22nd of 22”).
mart_metric_benchmark	Per-metric distribution by season over clean (ml_eligible) laps of fct_lap_residuals, so an explanation can say what is typical for a value.
mart_race_driver_summary	One row per driver-race, shaped for plain-language explanations (“why so slow at Monza”).

​

Reference

Model	Description
dim_circuits	One row per event slug (circuit_key).
dim_compounds_season	One row per circuit × compound × season.
dim_constructors	One row per constructor (team) name, from a fixed list in the model rather than from the loaded laps (CO-29), so every build, the CI fixture included, has every team.
dim_corners	One row per race × circuit turn: the telemetry distance window each corner owns, derived from FastF1’s own corner table (stg_circuit_info) rather than maintained by hand.
dim_driver_seasons	One row per season × driver code.
dim_drivers	One row per driver_id.