Crypto desk — 28 Sep 2026, 19:16 UTC

Phase 1 lessons drive this run: (1) the Candidate = buy flag is a trend screen, not an extension filter — a vertical blow-off reads "buy" exactly like a fresh base breakout; and (2) even a broad red day that pulls blue chips to low extension is NOT a buy trigger unless the risk-off COMPOSITION confirms — deploy only when low extension AND the flush (dominance spike + Greed collapse) both align.

Regime check: Fear & Greed 70 (still Greed), BTC dominance 58.7% (flat — no spike/flight), total mcap $2.87T, book 24h -2.27% ($-12.88), Mood "Soft". This is a soft-red day, not capitulation. Blue chips are mildly pulled back — BTC 83652 sits +3.7% above its fast MA (80657.077), ETH 2684.21 +3.8% above (2584.7915) — modest extension but no sign the low is in. Low extension is arguable-at-best and the flush composition is absent, so no edge — the exact setup my second lesson says to sit out.

The hot names are textbook traps under lesson 1: QNT 234.32 with fast MA 94.2955 (~148% above), +26.2% 24h / +270.7% 7d / +301.2% 30d — the research even carries the "trader sold everything" headline; HBAR +34.3% 24h / +40.2% 7d / +68.9% 30d; NEAR +25.3% 7d / +166.6% 30d. All flagged buy on the trend screen, all blow-off risk — skipped on sight. News (DOGE ETF inflows, XRP/SOL climbing) is attention only, never a reason.

Discretionary exit pass on my [BUY AND SELL] holdings (ETC, AVAX, BTC, AAVE): check_crypto_stops fired nothing — all four sit on the hard floor (ETC -10.7%/peak +0.0%, AVAX -2.6%/peak +5.2%, BTC -0.5%/peak +3.9%, AAVE +1.9%/peak +8.2%), none armed to breakeven/trail, none ADD-RISK. MAs stay bullish (AVAX, BTC, AAVE all above fast MA in research; ETC not in the table but no thesis break), no blow-off, and the backdrop is soft not risk-off — so no discretionary trim. ETC is my worst mark but its stop hasn't fired; let the mechanical floor govern rather than churn a fee.

No trades — held all 18 positions. No stop fired and none is ADD-RISK; no [BUY AND SELL] name (ETC, AVAX, BTC, AAVE) has a broken thesis or blow-off, so no discretionary exit. No buy either: cash $263.86, the full $50 daily cap and all 3 slots open, no pending orders — but on a Greed-70 board with dominance flat at 58.7% and only a soft -2.27% book, the risk-off flush isn't here and blue-chip extension is only modest, so a $5 buy is timing without edge. The hot candidates (QNT, HBAR, NEAR) are momentum traps wearing green labels. Availability isn't edge — preserve the daily cap for a genuine capitulation. Desk-traded P&L at review: realized $-21.40, unrealized $+0.98, total $-20.42.

Extension is measured by distance-above-fast-MA and the multi-day run, NOT by the 24h number — a cooling 24h print does NOT reset a blow-off. When a parabolic name's daily move fades, the screen keeps tagging it Candidate = buy and the softer 24h can read as "calming down," but the stretch that makes it a trap lives in the 7d/30d run and how far price still sits above the fast MA — those persist across runs even as 24h cools. Judge the trap by the gap to the fast MA, not the daily change. Evidence: QNT was +77.2% 24h ~two runs ago; today its 24h cooled to +26.2% yet 7d is +270.7% / 30d +301.2% and price 234.32 still sits ~148% above its fast MA (94.2955) — barely less stretched than before, and the research even carried a "trader sold everything" headline. HBAR (+34.3% 24h/+40.2% 7d/+68.9% 30d) and NEAR (+25.3% 7d/+166.6% 30d) same profile. All flagged buy; all skipped. Regime gave no cover to deploy anyway (Greed 70, dominance flat 58.7%, book -2.27% soft — no flush), so held all 18, $0 of $50 used. Refinement of the candidate buy flag is a trend screen not an extension filter: that lesson said the flag carries no extension info; this pins WHICH metric does — the fast-MA gap and multi-day run, not the headline 24h — so a fading daily print is never the all-clear.