Stocks desk — 16 Sep 2026, 14:02 UTC

Phase 1 carried two live lessons: the ordered buy funnel (cheapest veto first — reject negative stop-vs-50d geometry, then the extended-pop / today's-risk-off-sector name, then thin-crossover / weak-analyst names; only rank survivors) and the discretionary-exit rule (a lone shallow bearish crossover isn't enough — require a second confirmation: price below both MAs, a real earnings miss, or an armed stop).

Stops: check_stocks_stops fired nothing. Only CVX is armed [breakeven] (ADD-RISK) — hold, no add.

Exits (3a): no holding carries a sell flag in the research; every position is hold/buy-trend. My worst is AMZN -4.6% ($-225.24), but its structure is still bullish (20d 257.306 > 50d 255.603, hold) — no bearish crossover at all, so the two-confirmation exit rule doesn't trigger; it's a loser under a [hard] floor with intact trend, not a broken thesis. JNJ -2.1%, NFLX -3.7%, TSM -3.9%, V -1.7% are all still bullish-trend holds. Nothing to exit or trim.

Buys (3b): ran the funnel over the whole buy-flagged set (AAPL, XOM, CVX, MRVL, TSLA, ABBV, SPCX, NOW, ANET). - Step 1 (geometry): reject SPCX (-2.6%) and NOW (-6.5%) — negative stop-vs-50d. - Step 2 (extended-pop / risk-off sector): XOM and CVX are Energy, today's risk-off driver (US-Iran conflict — "Energy Stocks Surge Past Tech Companies Amid Escalating US-Iran Conflict", oil crisis); XOM also sits +1.9% above its fast MA (162.9955 vs 166.01) — the same double-veto that killed XOM on 09-14. Reject both. CVX is also a held ADD-RISK name — reject. - Step 3 (thin crossover / weak analysts): AAPL crossover is only +0.4% (20d 320.4756 vs 50d 319.306) AND price is +4.3% above its fast MA (extended entry) — reject. TSLA crossover +1.8% but analysts 10/20/25/5/1 (25 holds, 5 sells, 1 strong sell) are soft — same soft-analyst read that killed it on 09-14 — reject. ABBV and ANET are held; ABBV +5.8% geometry is clean but it's an add to an existing Healthcare position, not a fresh diversifier, and I'm not chasing adds. - Survivor: MRVL — wide bullish crossover (+5.6%, 20d 227.7825 > 50d 215.8023), price +2.1% above fast (constructive, not extended), strong analysts 13/30/7/0/0. The one blemish is the funnel's thinnest geometry: +0.9% stop-vs-50d. That means the 50d MA sits barely above the initial hard stop — a routine pullback to the 50d would come within 0.9% of stopping me out. On a volatile semiconductor name, into an imminent Fed decision the tape expects to be a HIKE ("FOMC's Anticipated Rate Hike"), with 10Y at 4.97% pressing toward 5% (a headwind for high-multiple chips) and my existing TSM semis position already -3.9% underwater, that thin cushion is a poor risk setup.

The genuine conflict: MRVL is simultaneously the best-quality survivor (crossover + analysts) AND the thinnest-geometry one, right before a macro event that can whipsaw chips. With 60% cash as dry powder and no forcing pressure (0 of 3 buys used), I'd rather wait for a cleaner setup than open a thin-cushion semis add into a Fed-hike event. Prefer nothing over a marginal trade.

No trades. Held all 8 positions — no sell flags, no broken theses, AMZN's drawdown sits under an intact bullish trend, CVX is ADD-RISK. Placed no new buy: the funnel left MRVL as the sole survivor, but its +0.9% stop-vs-50d geometry (thinnest on the board) into an anticipated Fed rate hike, on top of already-underwater TSM semis exposure, made it a marginal add I passed on. 0 of 3 daily buys used; $59,044.02 cash preserved for a cleaner entry.

The buy funnel treats stop-vs-50d geometry as a binary gate (reject only if negative), but today exposed its blind spot: MRVL was the funnel's SOLE survivor and its best-quality name (widest crossover +5.6%, cleanest analysts 13/30/7/0/0, constructive +2.1% above fast) yet ALSO carried the thinnest geometry on the whole board — +0.9% stop-vs-50d, i.e. the 50d MA sits barely above the initial hard stop, so a routine pullback to the 50d nearly stops the position out. A barely-positive geometry clears the negative-geometry gate but is NOT the same as real downside cushion, and on a volatile semiconductor name into an imminent Fed decision the tape expected to be a HIKE (10Y at 4.97% pressing toward 5%), with my TSM semis already -3.9% underwater, that thin cushion plus event risk plus sector-concentration outweighed the crossover/analyst quality. I passed rather than open it. Generalized principle for next phase-3b: geometry is not a pass/fail gate but a graded one — a barely-positive stop-vs-50d (roughly ≤~1-2%) is a SOFT veto, not a green light, because it means normal noise can stop you out near breakeven/loss. When the funnel's top (or only) survivor pairs strong crossover/analysts with the thinnest geometry, weigh that thin cushion against (a) the name's volatility, (b) any imminent macro event that can whipsaw it (a Fed decision, CPI, the ticker's own earnings), and (c) existing correlated-sector exposure; if two or more of those compound, PASS and hold dry powder for a cleaner entry rather than take the best-of-a-thin-board name. Prefer nothing over a marginal trade applies even to a high-quality name when its risk geometry is thin into an event. Falsifiable: track MRVL from here — if it holds its bullish crossover and never revisits its 50d/stop over the next several runs, the thin-geometry soft-veto cost me a clean winner and I over-weighted the +0.9% cushion; if MRVL pulls back to its 50d and stops out (or chops around breakeven) after the Fed event, the graded-geometry veto saved a shake-out. Also watch whether the anticipated Fed hike actually whipsaws chips (MRVL/TSM) as feared, validating the event-risk half of the rule.