signal-platform/reports/backtest-20260713-gtl-tuning-matrix.md

GTL tuning matrix

Status: complete
Holdout split: 2024-07-01
Completed arms: 20/20

Arm	Full Sharpe	CAGR	Max DD	Trades	Train Sharpe	Test Sharpe	Ex-top-5% R	Screen
control	2.03	50.0	21.4	321	1.28	2.78	0.049	0/0
lookback_252	1.43	26.0	22.8	241	0.78	2.07	0.152	1/6
lookback_504	1.75	38.8	21.4	289	1.32	2.20	0.055	4/6
lookback_756	1.60	36.4	21.4	318	1.28	1.81	0.013	4/6
candidates_8	1.91	44.5	18.9	320	1.44	2.39	0.024	4/6
candidates_all	1.74	39.0	18.9	322	1.15	2.31	0.014	3/6
max_atr_5_5	1.61	37.0	23.1	348	1.15	2.09	-0.011	1/6
max_atr_8	1.89	46.3	19.7	348	1.51	2.27	0.003	4/6
touch_0	1.88	45.5	22.0	321	1.25	2.48	0.053	2/6
touch_0_25pct	1.94	47.9	20.8	323	1.27	2.57	0.050	3/6
merge_0_25pct	1.82	43.4	20.8	327	1.78	1.71	0.062	4/6
merge_1pct	1.52	34.8	19.4	319	1.49	1.42	0.045	4/6
zones_1pct	1.71	38.3	20.3	291	1.43	1.91	-0.044	3/6
zones_3pct	1.71	41.9	17.8	339	1.34	1.92	-0.006	3/6
grid_12	1.74	42.7	22.0	337	1.25	2.16	0.023	2/6
grid_32	1.63	37.8	17.3	325	1.54	1.57	0.043	4/6
pivots_none	1.89	44.7	17.3	309	1.72	1.96	0.077	4/6
pivots_11bar	1.31	27.9	18.7	334	1.15	1.43	0.011	3/6
strength_250	1.82	45.1	18.9	326	1.29	2.32	0.056	4/6
strength_1000	1.88	46.7	18.5	338	1.42	2.32	0.080	4/6

Interpretation guardrail

The post-2024 interval has already informed prior research. The train/test columns are robustness checks, not a pristine holdout. A passing arm is a candidate for forward paper validation, not automatic production promotion.