KB / api
GET /api/v1/signals/base-rates/calibration
Last verified
Auto-generated. This article is rebuilt from FastAPI route introspection by
scripts/build_api_kb.py. Edit the route docstring inapp/and re-run the script — do not edit this file directly.
GET /api/v1/signals/base-rates/calibration
Handler: app.signals.routes.get_base_rates_calibration
Methods: GET
Description
Realised-vs-forecast accuracy for the pattern matcher.
Returns hit rate, MAE, Brier score, and a decile-bucketed calibration breakdown so the dashboard can surface “this prediction class has been X% accurate over Y similar predictions” alongside the live base-rates card.
Args:
horizon: filter to a specific horizon (1, 3, or 5). Omit to
return per-horizon + overall.
days: lookback window for resolved predictions (default 180).
matcher_version: filter by the matcher logic version
that produced each prediction. Default = current
MATCHER_VERSION so a logic change doesn’t pollute the
curve with stale-matcher predictions. Pass
matcher_version=all to include every prediction
including pre-versioning rows (NULL).
Returns:
{by_horizon: {"1": {...}, "3": {...}, "5": {...}}, overall: {sample_size, hit_rate, mean_abs_error, mean_signed_error, brier_score, buckets[]}, matcher_version: str}
Parameters
| In | Name | Type | Default | Description |
|---|---|---|---|---|
| query | horizon | Optional[int] | ||
| query | days | int | 180 | |
| query | matcher_version | Optional[str] | Filter to a single matcher version. Default: current MATCHER_VERSION. Pass ‘all’ to include legacy NULL-version rows. |
Curl
curl -s https://bigclawd.com/api/v1/signals/base-rates/calibration