Trading lab
Thu Sep 10, 2026 · paper only · ALGO neighborhood

Published 2026-09-11 08:26:08 CEST

Lab pick ETH transfer no

ALGO breakout KEEP moves to a slightly stickier exit

After yesterday’s honest-engine reset, the only KEEP was ALGO “buy a fresh 20-day high, sell a 10-day low.” Today we checked nearby settings. A 12-day exit (still buying the 20-day high) is clearly better on paper — not a knife-edge fluke.

$1,000 paper 2019 → 2026-09-08 15 bps costs next-open fills

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

What $1,000 would have become

New KEEP 20/12
~$19.0k
Prior KEEP 20/10
~$10.7k
Just hold ALGO
~$80
Worst dip (new)
~−65%

Rough ride still — alts are violent. The point is the rule sat in cash a lot (~36% of days) and avoided most of ALGO’s long bleed while catching big upswings.

Honesty checks

Year by year
20/12 · 8/8 beat hold
20/10 · 6/8
Rolling forward
20/12 · 17/25 · +20%/fold
20/10 · 17/25 · +14%/fold

Same win-rate vs hold on the rolling test, but the stickier exit made more money per stretch. Trade-off: slightly deeper worst dips on some windows.

Verdict

Lab pick
donchian_breakout 20/12 on ALGO — new alts paper favorite. Replaces 20/10.
Demote
20/10 — still fine, just second place in the same family.
Don't
Copy 20/12 to ETH — loses to holding ETH ($12k vs ~$17k; only 2/8 years). ALGO-specific for now.
Don't
Treat this as a live order. Paper-live Day 0 queued a paper buy for the next daily open. You own every real trade.

Also today (quiet path step)

Old Donchian 10/5 paper books (demoted yesterday) marked through Sep 8: BTC filled its queued paper buy (~$990), ETH still in (~$1,000), ALGO 10/5 queued a paper buy. Those books are historical trackers only — not the KEEP.

New KEEP paper book: experiments/paper_live/donchian_breakout_ALGOUSD_entry20_exit12/

Next