Recommendations
0 to acquire · 2 to dispose
| Ticker | Name | Close | Action | Action Price Where to act — a buy-limit on a pullback or a buy-stop above resistance, not the current price. Sanity-check vs Close: on a held name, an action price stranded above the close means the queued level was set when the stock was higher — stale, awaiting reprice. | Stop The risk-defined exit, set below the entry for a long (~1.2–1.8×ATR). A held long whose Stop sits at/above the Close is already below its stop — a stale box awaiting reprice. | 1wk Model estimate of the likely price at this horizon — a forward projection, not a guarantee, and separate from the trade's Action Price or Stop. Where the ML estimator is enabled these are per-ticker calibrated forecasts (~80% coverage band) that are refreshed periodically and can drift between refreshes; otherwise they're interpolated from the signal's swing target. Confirm against the settled close. | 4wk Model estimate of the likely price at this horizon — a forward projection, not a guarantee, and separate from the trade's Action Price or Stop. Where the ML estimator is enabled these are per-ticker calibrated forecasts (~80% coverage band) that are refreshed periodically and can drift between refreshes; otherwise they're interpolated from the signal's swing target. Confirm against the settled close. | 12wk Model estimate of the likely price at this horizon — a forward projection, not a guarantee, and separate from the trade's Action Price or Stop. Where the ML estimator is enabled these are per-ticker calibrated forecasts (~80% coverage band) that are refreshed periodically and can drift between refreshes; otherwise they're interpolated from the signal's swing target. Confirm against the settled close. | Rationale |
|---|---|---|---|---|---|---|---|---|---|
| TSM | Taiwan Semiconductor Manufacturing | 392.31 | SELL ALL | — | 406.37⚠ | n/a | n/a | 346.32 | Sell the last 103 TSMC shares in full at the open. The stock has now closed below our 406.37 risk level for a fourth straight session and the gap is widening, not healing - 392.31 is 3.5% through the stop, versus 1.8% a day ago and 0.7% before that. The model's own 12-week view has fallen again to 346.32, roughly 12% BELOW today's price, and the signal engine has printed no tradeable setup four sessions running. TSMC's business is emphatically not the problem: second-quarter revenue grew 33.7% to 2bn with gross margin of 67.7% and operating margin of 60.3%, both above guidance; full-year growth guidance was raised above 40%; the September quarter is guided to -45.8bn, up 37% year on year; and Citi lifted its target price to NTWhat the market is repricing is the cost of getting there - a -64bn capital spending guide and the depreciation and margin drag that follows - and it has taken the shares roughly 15% lower in a month. That is a valuation debate we can re-enter at a better price. It is not a swing-trade setup, and our stop has already done its job. Booking the loss at about -10.4% (roughly -) frees of capital. OPERATIONAL: this is the THIRD arming of the same exit, not a new decision - the 2026-07-25 verdict armed it for the 07-27 open and the 2026-07-28 verdict re-armed it for the 07-28 open; neither filled because the auto-executor has been idle since 07-23, which has cost a further ~2.65% of slippage. UT-11/UT-3/L3: an unfilled armed exit is an incomplete exit - re-fire at market-on-open. Both L2 legs are met (decisive stop-break AND negative 12wk), so this is a full completion. PROFIT_ARMED on the row is the UT-9 arm-on-loss mislabel (Gain_ATR -2.81, position underwater) and is inert - loss-prevention, not a profit-take. Valuation amplify suppressed by G2 (DCF anchor vs price, ~8.7x broken). 4wk/1wk n/a - NO TRADE prints a blank Target_Price; model band 371.56-412.76. · news |
| DELL | Dell Technologies Inc | 392.10 | SELL 50% | — | 358.88 | 414.96 | 468.29 | 540.61 | Sell half the Dell position - 174 of - at the open, and keep the other half working. Our 423.45 stop did not just break, it broke hard: Dell closed at 392.10, some 7.4% below the level we had written down as our risk line, after trading as low as 358.88 (-14%) intraday. A decisive break of our own stop is a signal-based reason to reduce, full stop. The trigger was not only sector beta: an analyst note singled out Dell's dependence on a small group of hyperscale customers and revived doubts about whether booming AI-server volume can carry acceptable margins - the same bear case behind the 07-16 UBS downgrade. Peers confirm it is high-beta AI hardware being repriced rather than the market as a whole, with HPE down about 8% and Super Micro about 5%, while software and Apple actually rose. The signal degraded with the price: conviction slipped from a stable, the internal read fell to WATCH, and volatility turned bearish as the average daily range blew out from 30..1 points. Why half and not all: the model's 12-week view is still 540.61, some 38% above the close, the four-week target 468.29 is 19% higher, the trend reading is still bullish with the price above its 50-day and far above its 200-day average, and the stock closed 9.3% off its low on a genuine intraday reversal. The underlying business news has not changed - 4bn of AI server orders, a 3bn backlog, an FY27 guide of about $60bn in AI-optimised server revenue, and the stock is still up around 200% this year even after the crash. What changed is the market's willingness to pay for it. Selling everything into a panic close on a name whose forward view is intact would be the mirror image of the mistake we made cutting NBIS. So we halve the risk and let the rest work. Booking roughly -10.8% on the half sold (about -) frees OPERATIONAL: L2's full completion needs stop-break AND a negative 12wk; only the stop-break leg is met (12wk +37.9%), so this is a staged de-risk, not a completion - the same tranche pattern used on TSM. G1/G2/G3 do not apply: those gates protect an uptrending WINNER from a divergence/valuation trim, and this is a -10.77% position on a fired stop. STOP RE-BASE (deliberate exception, flagged): the 423.45 ratchet fired and the half-exit is the response to it; the residual are re-based to the model's current chandelier level of 358.88 (today's low), capping further residual risk at This is a re-base after a fired stop, NOT a ratchet-down of a live stop on an unbroken position. UT-10 vindicated - both prior pullback ADDs (07-10 @411.73, 07-15 @419.03) graded stop_hit and a third would have compounded today. Escalation on the residual: a decisive close below 358.88, a 12wk flip negative, or 2 consecutive NO TRADE completes the exit. Estimates are the forward view partly foregone, f=0.30 per L5; 12wk 540.61 exceeds the DCF anchor - projection assumes growth beyond priced-in; model band 341.76-460.48. Executes into the 07-29 FOMC (horizon band +/-3%, stance tighten-trail) and ahead of Microsoft's Azure capex guide that evening, which is the key read-through for this complex. · news |
| AAPL | Apple | 340.08 | NO ACTION | 341.23 | 317.33 | 355.57 (+4.2%) | 389.02 (+14.0%) | 414.24 (+21.4%) | Apple was the strongest buy on the board and we stood down deliberately - worth recording why, because for the first time the entry was actually clean. It prints a STRONG BUY at conviction 93 and the #1 buy rank, and the model entry of 341.23 sits right at the 340.08 close rather than 6-8% above it as it did on the two previous attempts, so there was no chasing involved. The problem is the calendar. Apple reports FY26 Q3 tomorrow after the close, with a two-sided expected move of about 5%, and it is Tim Cook's final earnings call as chief executive - an unusually binary event for a stock at an all-time high and a roughly trillion market value, with the gross-margin guide under pressure from rising memory costs. Buying at tomorrow's open means owning a full position through that print one session later. Our own decision a day earlier explicitly set the order to stand down before this event; honouring that plan rather than talking ourselves out of it is the discipline. The bull case is real and unchanged - Apple is the direct beneficiary of the rotation that wrecked the rest of this book, rewarded for restrained AI capital spending while Dell, TSMC and Micron are punished for spending heavily - but that case will still be there on Friday, with the uncertainty resolved. RE-ARM TRIGGER: the 07-31 open, post-print. Take the then-live Entry_Price from the 07-31 snapshot, do NOT chase a positive gap (L6), and require adjusted conviction >=65 after val_adj. OPERATIONAL: two further discounts stack against taking it today - the family flipped out of the L1-favoured PullbackBuy_VWAP into MomentumContinuation_Up, this book's worst graded buy cell (n=2, 0/2 at 1wk, median error -4.19%, report-only per UT-12), and the entry is a buy-stop at the 52-week high (High_52W 342.89 = today's high). Quality gate PASS; MoS -3.16 gives val_adj -10 and 0.5x size, so adjusted conviction 83 would have cleared the 65 floor. Sizing had it been taken: @ 341.23 ~= Estimates at f=0.30 per L5; 12wk 414.24 exceeds the DCF anchor - projection assumes growth beyond priced-in. Ledgered direction:none/optional:true so the counterfactual grader prices exactly what standing down through the print cost or saved. · news |
| SNOW | Snowflake Inc. | 270.36 | NO ACTION | 265.63 | 251.20 | 272.07 (+2.4%) | 287.08 (+8.1%) | 384.40 (+44.7%) | Not taken for a fifth consecutive session, and it remains the best-formed setup in the book. Snowflake prints a buy at conviction 71 and the #2 buy rank on the pullback-to-VWAP pattern that is historically our most reliable entry, with a model entry of 265.63 just below the 270.36 close, a 12-week view 44% higher, and - notably - it barely moved on a day when the AI hardware complex fell 8-14%, closing down less than 1%. It is blocked twice over. On valuation, a margin of safety of -1.39 trims adjusted conviction to 64, one point below our 65 floor. Independently and more decisively, it fails the quality screen outright: the F-Score reads 5 and earnings-power value is actually NEGATIVE at -a share, meaning the entire intrinsic case rests on a projection of future cash flows rather than anything the business earns today - and our hard gate blocks new entries on that basis regardless of signal strength. The news tape does not argue the screen is wrong: product revenue reaccelerated and BofA has it as a Q3 top idea, but the shares trade near 18x sales and insiders have been persistent sellers. We stand aside. OPERATIONAL: two independent blocks (adj 64 < 65 floor; Quality_Pass FALSE hard gate on NEW entries only - it would not block an ADD to a held name). Ledgered direction:none/optional:true so the counterfactual grader prices what the gate costs - the fifth consecutive session this exact pattern recurs, which is now a standing systemic finding rather than a one-off. Estimates at f=0.30 per L5; 12wk 384.40 exceeds the DCF anchor. · news |
Outcome & track record
Accuracy at the time of this record.
9 Graded decisions Past recommendations old enough to score against what actually happened (target hit / stopped out / closed). This is the evidence base — the bigger the number, the more history behind the calls. Click to inspect every one. See the decisions →
0.134 1-week calibration How much of the eventual 4-week move typically shows up in the first week (0 = almost none, 1 = most of it). Use it to read the 1-week estimates: a low number means prices drift toward the target slowly, so don't over-read a soft week-1.
24 Published records How many dated end-of-day verdicts exist for this book — the length of the public track record you can browse in the Archive below.