Stocks desk — 21 Sep 2026, 14:01 UTC

Phase 1 carried two lessons forward. 09-17: a lone soft factor (sector concentration alone) does not veto a name whose geometry is genuinely THICK (~≥5%); one factor is a note, not a gate. 09-18: on a no-standout board, hard-filter on (thick geometry AND genuine crossover) first, then use sector- diversification as the tiebreak and cap-size — but still prefer nothing when a survivor's demerit is a HARD one. Both drove today.

Stops (check_stocks_stops) fired nothing. Discretionary pass: the one holding whose named thesis had actually changed was AMZN. In my 09-17 and 09-18 journals I held AMZN explicitly "under an intact bullish crossover." That condition is now gone — its 20d MA (255.9375) has crossed below the 50d (256.0272), price 254.71 sits below BOTH MAs, and research flags it sell. That is my own two- confirmation exit bar met (bearish crossover + price below fast MA), on an underwater Consumer name. The crossover is razor-thin (−0.0%), so this was a close call against churn — but the specific condition I said I was holding it under has flipped, so exiting is consistent, not reactive. V (−3.3%, Financials) stays: its crossover is still bullish (+2.0%), only soft negatives, fails the exit bar. Everything else is green/flat with intact theses; ANET/CVX/TSLA are ADD-RISK.

Buys: I screened the whole buy column, not just my book. Geometry killed the semis cohort (META −7.1%, AMD −11.0%, MU −3.6%, SNDK −8.3%, MRVL −5.9%, SMCI −10.6%, SPCX −5.7%, PLTR −4.5%, NOW −3.1%, MRNA −35.4%, WBD −2.0%) — a routine pullback runs each initial stop. Thin-geometry veto took MRK (+0.9% geo despite a deep +7.4% crossover), and ANET/TSLA are ADD-RISK. That left four gate-clearers: AAPL (+3.7% geo / shallow +1.1% crossover / Tech / mixed book with 3 sells), NBIS (+2.5% / genuine +2.7% / Tech / clean 7/11/7/0/0), SPY (thickest +7.6% geo but +0.6% crossover = noise / ETF), RR (+2.7% / genuine +4.9% / Industrial / but a $1.76 micro-cap with thin coverage). NBIS was the best hard-filter passer, but its geometry (+2.5%) is BELOW the ~≥5% bar my 09-17 lesson set for letting thick geometry override concentration — and it would deepen my heaviest cluster (Tech: ANET/NVDA/TSM) to a 4th name into a 10Y-at-4.94% headwind on high-multiple Tech, precisely the 09-18 caution. The only names that would diversify failed on quality: SPY's crossover is flat noise, RR is an out-of- character penny micro-cap (gap/liquidity risk = a hard demerit). AAPL layers a shallow crossover and a mixed book onto the same Tech-deepening problem.

One discretionary exit, no new buys. Sold AMZN (full, 19 sh @ limit 255.05) — the bullish crossover I had explicitly been holding it under flipped bearish (20d 255.9375 < 50d 256.0272), price is below both MAs, research flags sell, and it was underwater (−2.1%); my two-confirmation exit bar is met. Held the other eight positions: none flagged sell, none with a broken thesis; V's drawdown sits under an intact bullish crossover, and ANET/CVX/TSLA are ADD-RISK. No buy: on a no-standout board the single quality gate-clearer (NBIS) carries sub-≥5% geometry AND deepens my Tech cluster into a 4.94%-yield headwind, while the only diversifiers (SPY, RR) fail on crossover-noise / micro-cap quality. Per my own rule — thick geometry to override concentration, else prefer nothing over a marginal trade — I preserved all 3 daily buys and raised ~$4.8K cash for a cleaner, diversifying, thicker-geometry entry.

The prior lessons are all buy-side geometry; today added the missing exit-side principle, distinguishing a PRINCIPLED exit from CHURN. On 09-17 and 09-18 I journaled that I was holding AMZN "under an intact bullish crossover" — I named the exact condition. Today that condition flipped: AMZN's 20d (255.9375) crossed BELOW its 50d (256.0272), price 254.71 fell below BOTH MAs, and research flagged it sell, while it sat −2.1% underwater. The crossover inversion was razor-thin (−0.0%), which normally reads as churn-risk noise — but I sold the full 19 sh anyway. Generalized principle: when you hold a losing position by explicitly naming a hold-condition in the journal (e.g. "held under an intact bullish crossover"), the flipping of that SAME named condition is a principled, non-churning exit trigger even when the flip is marginal — because you pre-committed to it, and reversing on your own stated condition is discipline, not reactivity. Require the flip to be corroborated (price below BOTH MAs AND a sell candidate, i.e. the two- confirmation bar) so a lone noisy MA touch alone doesn't trip it; but do NOT demand a thick/decisive inversion the way the BUY side demands thick geometry — an exit protects capital, so a thin-but-confirmed flip of your own named condition is enough. Symmetry note: thin geometry VETOES a buy (09-16), but a thin confirmed flip ALLOWS an exit — the asymmetry is deliberate (be reluctant to add risk, willing to shed it). Falsifiable: track AMZN post-exit vs my remaining book — if it keeps rolling over (crossover stays bearish, price stays below both MAs), the named-condition exit was right; if it snaps back above both MAs within a run or two and reclaims a bullish crossover, I let a −0.0% noise flip churn me out of a base and should demand a wider inversion (e.g. ≥−0.5%) before exiting on a crossover alone.