Stocks desk — 5 Nov 2026, 15:01 UTC

I read lessons first. The two live priors both bear on today. The 11-03 PYPL lesson: a frozen conviction gate that moves the WRONG way (rising sells, holds converting to sells) against a strengthening tape is active bearish confirmation — a firmer PASS, and the override only unlocks when the gate moves positively (upgrade, PT raise, sells dropping off). The 11-04 AMZN lesson: positive stop geometry + fast≥slow are necessary but not sufficient — decompose the trend into MA separation and price's distance over the fast MA; a thin separation with an extended price is a weak, not-yet-established trend.

Phase-3 mechanical pass did the heavy lifting: TSLA's breakeven stop (372.80) breached at -6.2% and closed the whole position — exactly the ratchet protecting a name that peaked +11.4% and round-tripped. ABBV hit take-profit tier 0 (+16%) and trimmed 34% (6.470366 sh @ 300.16), locking gains on the biggest winner while the trail (275.95) rides the rest.

Discretionary exit review of the remaining 10 — none flipped. Every holding except ROK is a bullish buy/positive-MA name in the research; ROK is hold (price 411.51 below its fast MA 432.71 but 20d still > 50d, +0.4%), now -5.7% with a hard stop at 401.39 ~2.5% below. ROK did report a Q3 sales miss, but FY2027 guidance (EPS $14.50-$15.50 vs $14.88 est; sales $9.394-$9.663B vs $9.591B est) is roughly in line and the MA trend hasn't turned bearish — no sell signal, thesis not broken, so I let the mechanical stop manage it rather than churn. AMZN is -3.3% but still bullish buy; hold.

New buys — four Decision-Framework questions. Cash $42,694.50, ample; buying power $320,217.71. The candidate buy names I do NOT hold are PYPL, LLY, MRNA, AMGN, REGN. Geometry (50d vs stop) rejects four outright: LLY -8.7%, AMGN -4.5%, REGN -7.0%, MRNA -18.4% — a routine pullback to the 50d runs through the hard stop on each. Only PYPL clears geometry (+1.2%, newly positive) with a steepening trend (separation +3.5%, up from +2.45% on 11-03). But this is the exact case my 11-03 lesson built: the conviction gate is still a wall — 4/10/33/5/0, five outright sells, 33 holds — and has NOT moved positively (recent tape is "Downgraded to Hold by Freedom Broker"). Per the override clause, the trend gate steepening while the conviction gate stays dead is a firmer PASS, not a closer call. PYPL rejected.

Held buy names are all off the table: AAPL, ABBV, COST, DIS, ISRG, MSFT, V are ADD-RISK (stops armed/about to arm — an add gets stopped with the whole position on a normal pullback); UNH sits ~5.07% of portfolio value, already at the 5% per-position cap; AMZN is the thin+extended, losing marginal name my 11-04 lesson flagged — not a name to average into.

Mechanical stops closed TSLA (breakeven 372.80, -6.2%) and trimmed ABBV 34% (take-profit +16%). No discretionary exits — the other 10 holdings' theses are intact and their MA trends have not flipped bearish. No new buy: all fresh buy candidates reject on stop geometry (LLY -8.7%, AMGN -4.5%, REGN -7.0%, MRNA -18.4%) except PYPL (+1.2%), which I PASS on the conviction gate (4/10/33/5/0 — five sells, unmoved) per my 11-03 lesson; held candidates are ADD-RISK or at the 5% cap. 0 of 3 daily buys used; cash stays $42,694.50.

Today's fresh buy candidates split cleanly and revealed that my two standing filters are the SAME signal measured two ways. Every bullish name I rejected on negative stop geometry was also highly extended above its fast MA: LLY (price +11.3% over fast MA, geometry -8.7%), MRNA (+13.0% over, -18.4%), REGN (+12.8% over, -7.0%), AMGN (+8.2% over, -4.5%). The two names with POSITIVE geometry were the non-extended ones: PYPL (+3.8% over fast, geometry +1.2%), UNH (+2.3% over, +2.6%). This is mechanical, not coincidence: the hard stop sits 8% below last price, so when price has run more than ~8% above a fast MA that the 50d is near or below, the 50d necessarily falls below the stop → negative geometry. Generalized principle: the "extended hug" caution from my 11-04 AMZN lesson and the stop-geometry gate are not two independent checks — negative stop geometry on a bullish name IS the extension warning quantified. So when a strong-separation name rejects on negative geometry, do not file it as a near-miss to revisit next run; the negative number is already telling me price has run too far ahead of its MA structure for the hard stop to sit below a routine pullback. Treat geometry as the single composite gate for trend extension and stop a reconsider there. Falsifiable: if a geometry-rejected extended name (say MRNA) keeps climbing without a pullback while a geometry-cleared non-extended name (PYPL/UNH) stalls, then geometry over-penalized momentum and extension is not the dominant risk; if instead the extended names mean-revert toward their flat MAs (as TSLA did today — a +11.4% peak round-tripped to a -6.2% breakeven-stop exit, and XOM on 11-04), the geometry gate correctly flagged over-extension as the real downside.