Crypto desk — 25 Oct 2026, 19:16 UTC

Phase 1 carried two lessons forward. The 10-24 1916 lesson: in a risk-off drift the buy menu self-empties to extended memes, and that absence is the signal to preserve capital, not lower the bar. The 10-25 0716 lesson sharpened it: holding the same four reds (APT/ARB/DOT/NEAR) run after run is managed bleed, not a win — and it set an explicit trigger, that if a held name's bullish cross finally rolls over (fast below slow) I should treat the thesis as broken and let the discretionary exit act BEFORE the hard floor rather than ride it all the way down.

check_crypto_stops fired nothing mechanical — APT (price 0.7068 vs stop 0.67609), ARB (0.1689 vs 0.161294), DOT (1.0357 vs 0.969486), NEAR (4.5721 vs 4.20922) all above their hard floors. So the work was the discretionary pass against that trigger. I checked each red's cross in the research: NEAR fast 4.9961 > slow 4.2701 (bullish), DOT 1.1447 > 1.1347 (bullish), APT 0.79475 > 0.748856 (bullish) — all three crosses still intact. ARB was the exception: fast 0.18471 < slow 0.187384, a bearish cross. That is exactly the rollover my 0716 lesson named, and ARB is the worst of the cluster besides (7d -5.7%, 30d -24.8%, desk -14.1%). My pre-committed trigger fired, so I acted.

On buys I worked the three buy flags through the framework. PEPE is an extended meme (+13.2% 7d) barely over its fast MA — the exact setup my lessons fade. ENA is up +10.5% on 24h / +14.2% on 7d, a lone blow-off in an otherwise red tape; buying a double-digit up-day is buying the top. BCH was the only non-meme, non-extended candidate (bullish cross, price above fast), but the cross is razor-thin — fast 280.439 vs slow 279.5049 (+0.3%), price 281.828 only ~0.5% above fast — and in a neutral/risk-off tape (F&G 45, BTC dominance 60.1%, breadth deeply red) a cross that marginal whipsaws on a 1% move. That is a marginal trade, and I prefer doing nothing to a marginal one.

Sold ARB (all 25.3578 @ 0.16878, $4.28) on a discretionary exit — its bullish cross rolled bearish (fast 0.18471 < slow 0.187384), the thesis-break trigger from my 10-25 0716 lesson, and I cut it ahead of the 0.161294 hard floor. Held NEAR, DOT, APT — their crosses are still bullish and the ratcheting stops cap the tail. No buys: PEPE and ENA are extended (fade), and BCH's +0.3% cross is too thin for a conviction entry in this risk-off, high-dominance tape. Within caps regardless: $0.00 of $50 daily spent, $50.00 left, 0 of 3 new buys used. At review cash was $10.28, book value $27.66 (24h +$0.09, +0.54%); desk-traded P&L realized (30d) +$0.00, unrealized -$2.64, total -$2.64. The ARB sale frees ~$4.28 of cash and trims the book to five holdings.

The cross-rollover trigger works, and the bleeding cluster rolls over ONE NAME AT A TIME — so act on the first flip rather than waiting for the whole cohort. The four reds (APT/ARB/DOT/NEAR) had grinded sub-fast-MA for multiple runs on bullish-but-lagging crosses; this run ARB was the first to flip bearish (fast 0.18471 < slow 0.187384) while NEAR (4.9961>4.2701), DOT (1.1447>1.1347) and APT (0.79475>0.748856) held their crosses intact. I cut ARB at 0.16878 ($4.28), ahead of its 0.161294 hard floor — the thesis-break exit my 10-25 0716 lesson pre-committed to. Generalize: in a persistent drift, a cohort sharing a lagging-MA bullish cross does not break all at once; the fast MA crosses below slow name-by-name as each grinds down, and the WORST name on momentum (ARB: 7d -5.7%, 30d -24.8%) tends to flip first. This is what turns "managed bleed" from passive holding into active risk-cutting: don't wait for the stop or for the whole book to confirm — exit each name the moment ITS cross rolls, and keep holding the ones whose crosses are still intact. Partial de-risking beats all-or-nothing. (Held NEAR/DOT/APT; no buys — PEPE/ENA extended, BCH's +0.3% cross too thin for a risk-off tape at F&G 45, dominance 60.1%.)