Lab pick No upgrade

ALGO 20/12 still wins the nearby settings

Yesterday’s paper KEEP was “buy a 20-day high, sell a 12-day low” on ALGO. Today we tried nearby buy windows (18–22 days) and nearby sell windows (10–14 days). 20/12 is still first. Nothing next door made more money.

$1,000 paper 2019 → 2026-09-09 15 bps costs next-open fills

Path: Screen ✓ · Stress ✓ · Paper-live marked · Real $ needs your OK

What $1,000 would have become

KEEP 20/12

$18,662

Nearest 19/12

$17,069

Just hold ALGO

$79

Worst dip (KEEP)

−65%

All 25 nearby settings beat holding ALGO. The ridge peaks at exit=12. Dollars are a bit lower than yesterday’s $19.0k because ALGO’s latest daily close (Sep 9) was a down day — not because the rule changed.

Honesty checks

Year by year

20/12 · 8/8 beat hold

19/12 · 8/8

Rolling forward

20/12 · 17/25 · +19.6%

19/12 · 17/25 · +19.3%

19/12 is almost as honest, but it does not beat 20/12 on dollars or on average fold return. Rule for an upgrade: it has to win (or clearly match) both the history and the honesty tests. It didn’t. KEEP stays 20/12.

Paper book (not a real order)

The ALGO 20/12 paper book filled its queued buy on the Sep 9 open at 0.1008811. It is now long. Marked value: $983.22 vs $1,000 start (price slipped after the fill). Signal still says stay in. No next-open order pending.

Yahoo’s latest daily bar is still Sep 9 — Sep 10 has not printed yet. Demoted 10/5 trackers were marked too; they are not the KEEP.

Verdict

Lab pick

donchian_breakout 20/12 on ALGO — confirmed. Not a knife-edge cell.

Still testing

19/12 — closest neighbor. Same honesty, less money. Leave it.

Don't

Change BTC. Core sleeve stays buy&hold unless a new winner is obvious. None today.