Sticky SMA grid on BTC — knobs helped, still not core

Paper only Dropped Not BTC core

Verdict: Grid found sticky configs that clearly beat defaults on ending BTC units and soften year-split drawdowns. None beat fast SMA 5/20 on ending_units. Year-split vs buy&hold stays 3/7 (fast is 6/7). Soft DD ≠ good BTC accumulation rule.

Next: (1) BTC exit-only-on-deep-DD / Fear&Greed, stay long otherwise — or (2) try sticky on an alts sleeve, not core BTC.

Top 5 sticky by ending units (= by return)

#	Params	End $	Return	Max DD	In mkt	Units
1	10/100 c1 g0.5	$17,520	+1652%	−31.9%	55.5%	0.219
2	20/50 c1 g2.0	$16,004	+1500%	−43.3%	60.6%	0.200
3	10/100 c1 g1.0	$15,913	+1491%	−39.2%	55.8%	0.199
4	10/100 c1 g2.0	$15,856	+1486%	−38.0%	57.3%	0.199
5	20/50 c3 g2.0	$15,584	+1458%	−44.9%	58.9%	0.195

Defaults 20/50 c2 g1.0 → units 0.127. Best sticky 0.219 > B&H 0.132. Fast 5/20 → 0.612.

Focused full-sample compare

Strategy	End $	Max DD	In mkt	Units
Sticky best 10/100	$17,520	−31.9%	55.5%	0.219
Sticky defaults	$10,169	−52.2%	56.3%	0.127
Fast SMA 5/20	$48,827	−40.1%	53.8%	0.612
Buy & hold	$10,558	−76.2%	100%	0.132

Beat fast 5/20 on ending_units? No.

Year-split vs buy&hold

Config	Beat B&H	Avg PnL	Worst DD
Sticky best 10/100	3/7	+33.6%	−26.4%
Sticky defaults	3/7	+32.8%	−46.7%
Fast SMA 5/20	6/7	+89.9%	−32.5%

Best sticky softens DD vs defaults (esp. 2022) but still misses bull years vs hold.

Paths

Grid (full sample): experiments/runs/20260906T164117Z/
Focused cmp: …/20260906T164157Z/ · …158Z/ · …159Z/
Stress top: experiments/runs/stress_20260906T164206Z/
Ledger: experiments/results.tsv
Brief: briefs/2026-09-06-sma-sticky-grid.md · .html