quantmodel — Run one bounded strategy-fidelity experiment

Objective

Run one bounded strategy-fidelity experiment.

Notes: Rejected preregistered 63-session downside deviation: locked 2021-04..2026-06 corr 0.88816→0.88169 (−0.00646), failing required +0.005 gate; all other gates passed. Standard estimator remains default; non-default switch retained for reproducibility. Finding: held-out-bbdf-t248.md.

Approach

Reuses existing report research/findings/held-out-bbdf-t248.md — tables/metrics preserved verbatim via source_report pointer. Body below summarizes grounded evidence from that report and adjacent artifacts.

Files / code / data changed

No commit SHA confidently associated with t248 in sampled log [reconstructed — uncertain].
Source report: research/findings/held-out-bbdf-t248.md
Task-history entry: research/findings/task-history.json#t248

Results

Key table from source report (preserved):

interval / series	n	corr to AS	CAGR	abs CAGR error	monthly MAE	max drawdown	avg monthly turnover
calibration baseline	112	0.88686	6.711%	1.186%	0.005724	−6.277%	0.2209
calibration downside	112	0.88503	6.372%	0.847%	0.005835	−6.294%	0.2319
calibration AS	112	1.00000	5.525%	—	—	−8.058%	—
validation baseline	63	0.88816	9.635%	1.604%	0.005456	−6.628%	0.3001
validation downside	63	0.88169	9.624%	1.615%	0.005590	−6.107%	0.3016
validation AS	63	1.00000	11.239%	—	—	−5.034%	—
full baseline	178	0.88872	8.148%	0.239%	0.005566	−6.628%	0.2481
full downside	178	0.88622	7.936%	0.026%	0.005720	−6.294%	0.2552

Conclusions and metrics are in research/findings/held-out-bbdf-t248.md; see detail page for preserved tables/metrics.

Conclusions

Rejected preregistered 63-session downside deviation: locked 2021-04..2026-06 corr 0.88816→0.88169 (−0.00646), failing required +0.005 gate; all other gates passed. Standard estimator remains default; non-default switch retained for reproducibility. Finding: held-out-bbdf-t248.md.

Problems / follow-ups

See source report or PR discussion for follow-ups. No unsupported follow-ups fabricated.