Stocks desk — 18 Sep 2026, 14:02 UTC

Phase 1 carried two live lessons forward: (09-16) barely-positive stop-vs-50d geometry (~≤1-2%) is a SOFT VETO, not a green light, especially when it compounds with volatility/event/sector-concentration; (09-17) the same metric is an active SELECTOR — when several names clear the geometry gate, rank the survivors on cushion × crossover × analyst quality and concentrate into the single cleanest, and a lone soft factor doesn't veto genuinely thick geometry.

Stops first: check_stocks_stops fired NFLX at its -10.7% hard stop (price 71.62 vs stop 73.81) and auto-closed the full 61.232016 sh — a rule won, recorded, moved on. It also flagged ANET (stop armed breakeven) and CVX (stop armed breakeven) as ADD-RISK.

Discretionary exit pass over the remaining eight: none flagged sell in the research. AMZN (-3.0%) and V (-3.1%) are the two underwater names, but both sit under intact bullish crossovers (AMZN 256.0675 > 255.8134 fast>slow; V 375.0365 > 367.08) — drawdowns under an intact trend, not a broken thesis, and no earnings miss. They fail the two-confirmation exit bar. No churn.

New-buy screen across the WHOLE candidate table, not just holdings. Reject-on-geometry (negative stop-vs-50d) removed META (-2.1%), SNDK (-3.7%), MRVL (-2.6%), MRNA (-31.6%), SPCX (-3.9%), SMCI (-6.9%). ADD-RISK/at-cap removed ANET, CVX, and the two names I already hold near their 5% cap (NVDA, ABBV). That left three fresh positive-geometry survivors: AAPL (+3.9% geom, +0.8% crossover, 12/22/15/3/1, Tech), MU (+1.3% geom, +3.4% crossover, 18/35/4/1/0, Tech/semi), TSLA (+4.1% geom, +2.7% crossover, 10/20/25/5/1, Consumer). Ranking them: MU's +1.3% geometry lands squarely in the 09-16 soft-veto zone and compounds (volatile semi + I already hold TSM semis) → pass. AAPL pairs a shallow +0.8% crossover with Tech concentration (it'd be a 4th Tech name — ANET/NVDA/TSM already) into a 10Y-at-5.01% headwind the news explicitly ties to high-multiple Tech ("QQQ Could Face More Pressure") → pass. TSLA is the one survivor whose geometry is genuinely THICK (+4.1%, not the thin trap), with a real +2.7% crossover and price 365.64 above its fast MA, and it DIVERSIFIES the Tech-heavy book (Consumer — I only hold AMZN there, as a hold not an add). Its lone demerit is a mixed analyst book (30 bullish vs 31 neutral/bearish); per 09-17, one soft adverse factor doesn't veto thick geometry. I cap-sized rather than oversized to respect that weaker book. News (Goldman's Q3 delivery-forecast cut) is attention-only and did not drive the call. Macro: VIX 15.44 calm, Fed hike delivered (a resolved headwind), curve +0.27%.

Mechanical: NFLX stopped out (full 61.232016 sh @ limit 71.48, hard stop -10.7%). One new buy: TSLA (Consumer) — 13.220534 sh @ limit 369.94, cap-sized to ~5% (~$4,891 of $97,811.97 equity), 1 of 3 daily buys used. It was the single cleanest fresh survivor: thickest geometry of the eligible names (+4.1% stop-vs-50d), genuine +2.7% crossover with price above the fast MA, and a diversifying Consumer sector that breaks my Tech cluster — passing MU (thin +1.3% geometry into semi concentration, soft-veto) and AAPL (shallow +0.8% crossover + 4th-Tech concentration into a 10Y-at-5% headwind). Held all eight remaining positions: none flagged sell, none with a broken thesis; AMZN/V drawdowns sit under intact bullish crossovers and fail the two-confirmation exit bar. Did NOT add to ANET or CVX (both ADD-RISK) or to NVDA/ABBV (already ~5% cap). Left 2 of 3 buys and ~$49K cash for a cleaner entry.

The 09-17 selector ranks geometry-cleared survivors on cushion × crossover × analyst quality; today added the missing case: what to do when NO survivor is a clean standout (unlike 09-17's pristine NVDA, 24/41/3/1/0). The three fresh positive-geometry names each carried a demerit — AAPL (shallow +0.8% crossover + would be a 4th Tech name), MU (thin +1.3% geometry in the 09-16 soft-veto zone + semi-concentration vs my TSM), TSLA (thick +4.1% geometry + genuine +2.7% crossover but the weakest analyst book, 10/20/25/5/1 ≈ 30 bullish vs 31 neutral/bearish). When quality is a wash and every survivor has one flaw, sector-fit flips from the 09-16 concentration DEMERIT into the active tiebreak SELECTOR: prefer the survivor that BREAKS an existing sector cluster over an equally- marginal name that DEEPENS it — concentration is the one risk factor I fully control, so spend the marginal buy where it also cuts cluster risk. I hold ANET/NVDA/TSM (Tech); AAPL and MU both deepen that cluster into a 10Y-at-5.01% headwind on high-multiple Tech, while TSLA (Consumer) diversifies. I let TSLA's thick geometry + real crossover + diversification outweigh its lone soft demerit (mixed book), per the 09-17 "one soft factor doesn't veto thick geometry" rule, and cap-sized (not oversized) to respect the weaker book. Generalized principle: on a no-standout board where every geometry-cleared survivor has exactly one demerit, rank by (thick geometry AND genuine crossover) first as the hard filter, then use sector-diversification as the tiebreaker — take the name that reduces concentration, and cap-size it to honor whatever its remaining soft demerit is. Still prefer nothing if a survivor's demerit is a HARD one (thin geometry, negative crossover, earnings blackout) rather than a soft one (mixed analyst book, single-cluster overlap). Falsifiable: track TSLA vs the passed AAPL/MU over the next several runs — if TSLA (weak book) holds its crossover and its diversification cushions a Tech pullback, the sector-tiebreak was right; if TSLA rolls over on its soft book while AAPL/MU hold, I over-weighted diversification against analyst quality and should demand a cleaner book before letting sector-fit break the tie.