Stocks desk — 28 Oct 2026, 15:01 UTC

Phase 1: cash $38,295.27, equity $98,441.56, 12 open positions, 0 of 3 new buys used, no pending orders, market open. The 5%/position cap is $4,921.95 (5% of $98,439.01). My two standing lessons: (10-26) on a buy flag the decisive line is where price sits vs its OWN fast MA — hug (~0-2% over) with positive geometry = take, stretched (~5%+ over) = pass; and (10-27) once two names both clear the hug trigger, rank on real stop cushion then orthogonality, but conviction can outweigh diversification.

Phase 3(a) exits: check_stocks_stops fired no mechanical stop. Discretionary review: all 12 holdings are bullish in the research with no bearish crossover and no earnings miss — AAPL +1.3%, ABBV +11.3%, COST -0.7%, CVX +9.1%, DIS +2.5%, ISRG +12.5%, MSFT -0.4%, NOW +9.2%, ROK +1.3%, TSLA +7.9%, V +1.9%, XOM +11.3%. No thesis broken, no trim.

Phase 3(b) new buys — I worked the whole buy table, not just my book. Fresh names measured against their own fast MA: UNH 414.785 vs 395.0158 = +5.0% over (stretched, pass); LLY 1295.9299 vs 1221.4645 = +6.1% over with geometry -0.0% (reversion runs through the stop, pass); AMGN 453.64 vs 425.465 = +6.6% over (stretched, pass); MRNA 216.43 vs 211.983 = +2.1% over but geometry -9.7% (a routine pullback runs well past the -8% stop) plus a fresh William Blair downgrade (pass on geometry). The only non-extended fresh name was PYPL: 55.74 sitting exactly on its fast MA 55.7392, geometry +8.3% — textbook hug with the best cushion on the board. But two real flaws: its fast/slow MAs are only +0.4% apart (55.7392 vs 55.5386) — the weakest trend separation of any candidate, a flat/ranging crossover rather than a constructive uptrend — and its analyst mix is 4/10/34/3/0, hold-dominated with 3 outright sells. That is thin conviction on a barely-bullish trend, and it would be a 2nd Financial (I hold V), so not orthogonal. My 10-26 lesson says take non-extended quality names; PYPL clears the geometry but fails the quality half. Adds to held names are all blocked by ADD-RISK (ABBV/CVX/DIS/ ISRG/NOW/TSLA/V armed) or stretched, and ROK/AAPL/COST are at-trend but I just bought ROK and these are already full positions.

No exits (mechanical or discretionary) — hold all 12. No new buy. The only non-extended fresh candidate, PYPL, pairs a textbook fast-MA hug and +8.3% cushion with a marginal +0.4% MA separation (directionless chop, not a trend) and hold-dominated analysts (34 holds, 3 sells, Morgan Stanley Underweight) — thin conviction that fails the quality half of my hug-the-fast-MA trigger. Every other buy-flagged fresh name (UNH +5.0%, LLY +6.1%, AMGN +6.6% over fast MA) is a stretched chasing entry, and MRNA's -9.7% geometry runs through the stop. Prefer doing nothing over a marginal trade: 0 of 3 buys used, cash stays $38,295.27.

The 10-26 lesson told me to TAKE a non-extended fast-MA hug with a positive geometry cushion; today PYPL exposed a false-positive in that test and forced me to add a trend-separation gate. PYPL sat exactly on its fast MA (55.74 vs 55.7392) with the BEST 50d vs stop cushion on the whole board (+8.3%) and a tiny reversion-to-slow (-0.37%) — a textbook hug by every number the screen computes. But its fast and slow MAs were only +0.4% apart (55.7392 vs 55.5386), the weakest separation of any candidate, meaning the big cushion was an ARTIFACT of a cheap, range-bound stock sitting on top of both of its glued-together MAs, not evidence of a trend. Contrast the hugs I actually bought: MSFT (fast/slow +3.6% apart) and ROK (+1.1%) had real trend slope under the hug. PYPL also carried thin conviction (analysts 4/10/34/3/0 — hold-dominated with 3 outright sells, Morgan Stanley Underweight), so I passed and made no buy with 2 of 3 slots free. Generalized principle: the fast-MA-hug trigger has a hidden pre-condition — the fast MA must be meaningfully ABOVE the slow (a real crossover, roughly ≥1% separation), not glued to it. When fast≈slow (<~0.5% apart), price "hugging the fast MA" is just a flat/ranging stock oscillating around both MAs, and a large 50d vs stop cushion is a cheap-price artifact, NOT the at-trend entry the 10-26 lesson rewards. Gate the hug test on trend separation FIRST; only then does the geometry cushion mean what I think it means. And when a geometrically-perfect hug also carries hold-dominated analysts with outright sells, that thin conviction is sufficient to pass — take the hug only when real trend slope AND conviction back it. Falsifiable: if over coming runs PYPL chops sideways or rolls toward its slow MA while the trend-backed hugs I hold (MSFT/ROK) advance, the separation gate was right; if PYPL instead breaks out and trends up from here, the +0.4% separation was noise I over-weighted and a deep geometry cushion should stand on its own regardless of MA separation.