Lab pick Still testing

Alts KEEP transfers to BTC — Donchian 10/5 beats hold and SMA on the core sleeve

Question: does the alts winner donchian_breakout (buy at N-day high, sell at M-day low) work on BTC-USD where the goal is grow wealth and soften crashes — not just chase dollars? Answer: yes for 10/5. Full sample, every calendar year, and walk-forward all beat buy&hold; it also crushes the SMA 5/20 paper baseline on ending dollars, drawdown, and out-of-sample win rate. Not a live trade — paper KEEP for BTC research.

BTC-USD · 2019-01-01 → 2026-09-06 $1,000 start 15 bps round-trip next-bar open · 95% size

Don 10/5 end

$3.41M

vs buy&hold

$20.4k

vs SMA 5/20

$124k

Year beat

8 / 8

What $1,000 becomes · full sample

Same sacred harness on every row. “Ending units” = how many BTC your equity could buy at the end price (purchasing-power score, not a promise you held coins the whole time).

Strategy	End $	Max DD	In market	End units	Trades
Buy & hold	$20,362	−76%	100%	0.25	1
SMA 5/20 (baseline)	$123,586	−40%	55%	1.54	80
Donchian 20/10	$563,027	−25%	47%	7.0	41
Donchian 10/5	$3.41M	−18%	43%	42.4	88

BTC core caveat: Donchian sits in cash ~57% of the time (88 round-trips). That is more selling than a pure hold. For core wealth the strike is “sell too often and lose coins.” Here purchasing power still ends far above hold and SMA — so it clears the KEEP bar — but paper-live should watch whether real path feel matches the backtest before any hand-apply talk.

Year-split stress · Donchian 10/5

Each year restarts with $1,000. 2026 is year-to-date through Sep 6.

Year	Don end	PnL	Max DD	Hold end	vs hold
2019	$4,298	+330%	−18%	$1,869	BEAT
2020	$5,527	+453%	−7%	$3,868	BEAT
2021	$2,971	+197%	−13%	$1,563	BEAT
2022	$1,641	+64%	−11%	$388	BEAT
2023	$3,267	+227%	−7%	$2,469	BEAT
2024	$3,185	+219%	−7%	$2,143	BEAT
2025	$1,472	+47%	−6%	$937	BEAT
2026*	$1,356	+36%	−9%	$920	BEAT

8/8 beat hold · avg PnL +196.5% · worst DD −18.1%. In 2022 it made +64% while buy&hold lost −61% — the crash-softening story that matters for BTC core.

Year-split · Donchian 20/10 (calmer)

7/8 beat hold (lost only 2023) · avg +119% · worst DD −21%. Fewer trades (41), still ahead of SMA — but trails 10/5 on every score.

Walk-forward (out-of-sample)

Train 365 days → test 90 → step 90 · fixed params. Each fold starts with $1,000.

10/5 folds won

21 / 27

10/5 mean / fold

+25.9%

20/10 folds won

17 / 27

SMA WF (prior)

13 / 23

Donchian 10/5 beats buy&hold in 78% of folds (worst fold only −2.8%, worst DD −16%). SMA 5/20’s prior WF was ~57%. Edge looks real out of sample, not just one lucky full-sample path.

Keep · Tweak · Drop

Lab pick

donchian_breakout 10/5 on BTC-USD — beats hold on return and softens DD (−18% vs −76%); ends with far more purchasing power than SMA 5/20; 8/8 years; 21/27 WF. New BTC-core paper KEEP alongside the alts books.

Still testing

donchian_breakout 20/10 on BTC — still beats hold 7/8 and WF 63%, calmer (41 trades). Reserve if 10/5 churn feels too busy for the core sleeve.

Baseline

sma_crossover 5/20 — still the active BTC paper baseline to beat; Donchian just beat it on $ / DD / units / WF. Keep as reference, not as the new primary.

Paper-live side note (alts)

Same morning advance of the ALGO + ETH Donchian 10/5 books (data through Sep 6 PT window):

ALGO — still OUT / cash $1,000; no fill; behind buy&hold window (~$1,019).
ETH — IN after paper BUY at next open (~$2,484.55, 0.38 units); book ~$1,011.47, even with hold this window.

Path step Paper-live only — not a new KEEP by itself.

Path to real

Screen

Stress

Paper-live

Real (needs OK)

Next: start a BTC Donchian 10/5 paper-live book beside ALGO/ETH; keep SMA 5/20 as the foil. No live / broker orders without an explicit per-trade OK.

Suggested next experiment

Bootstrap paper_live for donchian_breakout 10/5 on BTC-USD (same params as alts), then optionally stress a hold-biased hybrid (Donchian entries + slower exits) if the sell-often feel bothers the core sleeve.