int_driver_circuit_affinity

One row per (driver_id, circuit_id).

Part of the Skill family.

One row per (driver_id, circuit_id). How much better or worse does this driver do against his teammate at this circuit than he does on average? Built on int_driver_race_skill_loro.driver_skill_loro_s (median lap-by-lap gap to the teammate). circuit_id is the physical-circuit identifier, so renamed events and double-headers at one venue (e.g. mexican_grand_prix / mexico_city_grand_prix) pool into a single track record. shrunk_affinity_s is a LEVEL (the shrunk circuit mean gap to the teammate), shrunk toward the driver’s own all-circuit mean (Bayesian, prior = 5 virtual races). The affinity itself is affinity_vs_driver_mean_s = shrunk_affinity_s - global_driver_mean_s: negative = better against the teammate at this circuit than the driver’s own average (WI-14b, F44: the page used to draw the level, so every cell was green). affinity_confidence [0,1]: < 0.17 = prior-dominated.

​

Lineage

​

Overview

Property	Value
Layer	intermediate
Family	Skill
Contract enforced	No
Tags	driver_rating, driver_affinity
Upstream	int_driver_race_skill_loro

21 tests guard this model - 19 generic, 2 singular.

​

Columns

Column	Type	Nullable	Tests	Description
driver_circuit_id		Yes	not_null, unique	Surrogate PK = hash(driver_id, circuit_id)
driver_id		Yes	not_null, relationships	Driver identifier
circuit_id		Yes	not_null, relationships	Physical-circuit identifier (slug of circuit_name). Collapses event-slug aliases (renamed events, double-headers) onto one venue.
circuit_name		Yes	not_null	Canonical circuit display name for this circuit_id.
n_obs		Yes	not_null	Number of race-residual observations backing this estimate
seasons_observed_n		Yes	not_null	Count of distinct seasons the driver raced at this circuit
raw_affinity_s		Yes	not_null	Unshrunk mean gap to the teammate at this circuit (a level). Negative = faster than the teammate there.
shrunk_affinity_s		Yes	not_null	Bayesian posterior mean of the circuit gap to the teammate (a level), pulled toward global_driver_mean_s. Negative = faster than the teammate at this circuit. NOT a deviation from the driver’s own average; that is affinity_vs_driver_mean_s.
global_driver_mean_s		Yes	not_null	The driver’s mean gap to his teammate over every race at every circuit: the prior mean each of his cells is shrunk toward.
affinity_vs_driver_mean_s		Yes	not_null	The circuit affinity: shrunk_affinity_s - global_driver_mean_s, seconds. Negative = the driver does better against his teammate at this circuit than he does on average; positive = worse. What the Driver Circuit Affinity page draws (F44).
shrunk_affinity_se_s		Yes	not_null, expect_column_values_to_be_between	Posterior standard error in seconds. Non-negative.
shrunk_affinity_ci_low_s		Yes	not_null	Lower bound of 95% credible interval
shrunk_affinity_ci_high_s		Yes	not_null	Upper bound of 95% credible interval
affinity_confidence		Yes	not_null, expect_column_values_to_be_between	Data fraction of posterior mass. n_obs / (n_obs + 5). [0, 1].