prop/priv/backtest_reports/baseline_td_depression.md
Graham McIntire ab04cb9168 Phase 0: backtest harness
Add Microwaveprop.Backtest: a feature-evaluation framework that runs
a (lat, lon, valid_time) -> float function over the historical QSO
corpus and a matched random-time baseline, reporting distribution
statistics, distance-binned lift, and band-stratified lift.

Adds four baseline feature wrappers around the current scorer inputs
(naive_gradient, td_depression, time_of_day, pressure), a mix backtest
CLI, and the first set of baseline reports under priv/backtest_reports
so downstream phases have a frozen reference point to compare against.
2026-04-09 16:10:54 -05:00

1.2 KiB

Backtest: Microwaveprop.Backtest.Features.td_depression

QSO sample: 5000
Baseline sample: 5000

Matched distribution

Set N Mean Stddev p50 p90 Min Max
QSO times 5000 6.569 5.365 5.573 12.969 -0.693 26.814
Random baseline 161 4.719 4.367 3.534 10.096 -1.258 21.762

Lift by distance (km)

Set N Mean Stddev p50 p90 Min Max
0-100 1249 5.324 4.270 4.151 10.907 -0.693 25.733
100-250 2295 6.941 5.789 5.589 14.564 -0.630 26.814
250-500 1288 7.007 5.510 6.054 14.450 -0.568 25.733
500-1000 167 7.390 3.911 7.422 12.637 -0.142 15.938
1000+ 1 7.722 0.000 7.722 7.722 7.722 7.722

Lift by band

Set N Mean Stddev p50 p90 Min Max
10000 MHz 4444 6.578 5.300 5.573 12.879 -0.693 26.814
24000 MHz 465 6.431 5.895 5.127 14.450 -0.693 26.814
47000 MHz 77 7.232 6.062 5.804 16.799 0.355 25.762
75000 MHz 14 4.824 2.365 3.339 7.317 2.483 7.317