> ## 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. # int_compound_cliff_predicted > Tyre compound degradation prediction using fitted survival-analysis parameters. Part of the [Pace Baselines](/transform/families/pace-baselines) family. > For the conceptual foundation, read [Int Compound Cliff Predicted](/decomposition/methodology). Tyre compound degradation prediction using fitted survival-analysis parameters. One row per lap. Predicts degradation rate and cliff onset based on fitted compound parameters and current thermal conditions. ## Lineage ```mermaid theme={null} flowchart LR dim_compounds_season --> int_compound_cliff_predicted int_stint_geometry --> int_compound_cliff_predicted stg_weather --> int_compound_cliff_predicted int_compound_cliff_predicted --> fct_cliff_prediction_features int_compound_cliff_predicted --> fct_ghost_car_pace int_compound_cliff_predicted --> int_constructor_deg_sensitivity int_compound_cliff_predicted --> int_lap_residual_decomposed int_compound_cliff_predicted --> int_pit_strategy_value int_compound_cliff_predicted --> int_synthetic_teammate style int_compound_cliff_predicted fill:#e40404,stroke:#e40404,color:#fff ``` ## Overview | Property | Value | | ----------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | | Layer | `intermediate` | | Family | [Pace Baselines](/transform/families/pace-baselines) | | Contract enforced | No | | Upstream | [`int_stint_geometry`](./int_stint_geometry), [`dim_compounds_season`](../dim/dim_compounds_season), [`stg_weather`](../stg/stg_weather) | **7 tests** [guard](/transform/ci/overview) this model - 5 generic, 2 singular. ## Columns | Column | Type | Nullable | Tests | Description | | -------------------------- | ---- | -------- | ---------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `lap_id` | | Yes | [`not_null`](/transform/ci/structural), [`unique`](/transform/ci/structural) | Foreign key to stg\_laps | | `compound_wear_s` | | Yes | [`not_null`](/transform/ci/structural) | The age-dependent portion of the hockey-stick curve, bounded at var('compound\_wear\_max\_s\_per\_lap'). Exposed as its own column so the bound is assertable directly -- see assert\_compound\_wear\_bounded.sql. expected\_compound\_pace\_s is this plus grip\_peak plus the temperature offset, so the pace total legitimately exceeds the bound while this column may not. As of 08m (2026-09-16) the formula is wear\_gradient*age + cliff\_ramp, where cliff\_ramp is the de-double-counted, moment-matched saturating post-onset term defined once in macros/compound\_cliff\_wear.sql. It was previously wear\_gradient*age + 0.002*age^2 + severity*laps\_past\_cliff. Both of the removed terms were unfounded: compound\_cliff\_severity is FITTED as a level shift in seconds over a \~5.5-lap window and was being charged once per lap (laps\_past\_cliff reaches 48), and the 0.002 quadratic was a literal constant identical across all 438 seed cells that was never fitted against anything. 08l measured them at 60.4% and 18.9% of the seed's contribution to the ML target respectively. | | `compound_wear_gradient` | | Yes | [`not_null`](/transform/ci/structural) | The fitted linear wear slope (s/lap) for this lap's circuit x compound x season cell, passed through from dim\_compounds\_season. Exposed by 08m alongside compound\_cliff\_severity because any consumer needing a post-onset RATE needs both: severity is a level shift, and de-double-counting it against this gradient is what turns it into a rate. Use the cliff\_ramp\_slope\_s\_per\_lap() macro. | | `expected_compound_pace_s` | | Yes | [`not_null`](/transform/ci/structural) | Expected pace loss vs compound baseline, in seconds. Bounded as of Phase 8 (2026-08-25): the wear portion is capped, having previously reached 93.5 s/lap on laps that were actually run. This column is subtracted into driver\_skill\_residual\_s, so the unbounded tail was depressing the residual on 12,575 post-onset laps and, through it, corrupting both ML targets -- it hid 6,825 cliff-label crossings (99.5% of them into 'none\_in\_stint') and moved the degradation target on 9.05% of rows. 08m (2026-09-16) changed what this column MEANS, not just its precision: the severity and quadratic terms it carried were removed (see compound\_wear\_s). Any number in the tree measured against next\_5\_lap\_cumulative\_jump\_s before that date is on the old quantity and is not comparable to one measured after it. |