This website is for educational purposes only and does not constitute investment advice. Always do your own research and assess your own risk tolerance before making investment decisions.

US End-of-Day Verdict 17 Aug 2026, 16:00:00 GMT-4

US Semiconductors — End-of-Day Verdict

0 ACQUIRE, 0 ANTICIPATE, 2 DISPOSE | Rates-driven sector rout on 08-18 (SOX ~-3.7%, 30y Treasury yield highest since 2007): every board name fell (COHR -12.8%, LITE -9.9%, SNDK -9.0%, ASX -8.8%, MU -7.0%, ASML -4.26%, TSM -4.08%, NVDA -2.34%) and BUY-ANTICIPATE prints went from 3 to ZERO | DISPOSE 1 ASML SELL 30% (5 of) @1802.98 =, retaining 13 - ADR-0012 armed (T12_Progress 91.22) + tick (b) FIRED, conviction, exactly the 20-pt threshold; no other tick (12wk 1976.51 = +9.62% alive, no 2nd code, no repeat NO TRADE) so profit_take_1 = 30%. The amplify to 50% is blocked by G3 NOT G2: MoS -4.92 <= -3 would bump the tier, but the 12wk is solidly positive on a bullish signal above EMA50/100/200 - and unlike TSM the DCF anchor here IS reliable (836.55 vs 1802.98 = 2.16x, cleanest on the board with NVDA 1.34x), so calling it a G2 block would contradict the 08-17 entry. Base trim justified on geometry: 1802.98 is only 1.50% above the 1776.26 stop. Stop STANDS (chandelier 1666.43 and model 1666.43 both below it, 3rd session the mandatory armed raise is unavailable). Realized - | DISPOSE 2 TSM SELL 30% (7 of) @413.41 =, retaining 17 - armed + tick (c) 3rd consecutive session and materially worse: Target_12Week moved from +0.55% ABOVE the close on 08-17 (433.32) to -1.68% BELOW it tonight (406.45 vs 413.41), through the price not merely inside the 421.68 dead band. FRESH origination not a re-issue (the 08-17 card filled this morning). G2 suppresses the MoS -14.46 amplify for a 3rd session (DCF 45.97 vs 413.41 = 8.99x; same class defect on TSLA 9.83x, INTC 7.15x, ARM 5.51x). L2 is fourteen cents away - 413.41 is +0.034% above the 413.27 trail, negative-12wk half met, decisive-close-below half not; if it breaks tomorrow the stop takes the remaining 17. Realized - | FALSE-ARM DEFECT now firing on 100% of held names, 33rd session: ASML armed at T12_Progress 91.22 while -1.31% BELOW cost (Gain_ATR -0.35), TSM at 101.71 while -5.62% below (Gain_ATR -2.01). Both 'profit-takes' realize losses. Advisory (a) unshipped | ACQUIRE NO ACTION - 6th consecutive session, and the blocker CHANGED CHARACTER. The prior five were unheld names failing the quality gate. Tonight the sole bullish-actionable row was NVDA BUY-ADD cv71 Buy_Rank 1 PullbackBuy_VWAP - a HELD name (gate inapplicable), on the exact family L1 favours, with the cleanest DCF anchor on the board - and it still died on arithmetic: val_adj -10 (MoS -2.56) -> adj 61, four points under the 65 floor. Note B/G2 does not rescue it (T12 246.90 vs DCF 163.98 = 1.51x is not a wide margin). Would have been QUEUED not bought anyway (Entry_Distance_ATR -1.43, limit 211.85 = -3.6%). RE-ARM: raw cv >= 75 clears the floor | ANTICIPATE none - ZERO BUY-ANTICIPATE prints anywhere on the board (3 last night); MU/LITE/SNDK all 1-of-3 and all faded to WATCH | THE 08-17 ENTRY'S HEADLINE CALL RESOLVED IN OUR FAVOUR: it named MU 'the cleanest acquire since inception if it re-prints'. MU did NOT re-print and fell -7.0% (1011.75 -> 940.76). ASX, declined on quality-gate grounds at 39.91, fell -8.8%. LITE, the highest-conviction name on the board at cv83, fell -9.9%. Three declines, three avoided drawdowns averaging -8.6% in ONE session - USC-3/USC-4/L6 all paid | NVDA HOLD 253 - not armed (T12_Progress 89.00 < 90, Gain_ATR -1.28) so ADR-0012 does not apply; the two-consecutive-NO-TRADE state flagged on 08-17 resolved UPWARD, vindicating the USC-5/US-3 'still-constructive fade -> trail' call; 6.64% above stop, above EMA200, 12wk +12.4% alive; marks -/-3.11%; FQ2 FY27 08-26 is SIX sessions out so the <=5-session catalyst window opens at the NEXT EOD | QUEUE HYGIENE: five stale GTC ledger lines closed - the Step-7 expiry reconcile had never been applied to older lines, so the counterfactual grader was under-counting non-fills for up to a month | SCORECARD: graded 26/107, calib_1w -0.207 (11th consecutive negative -> f=0.30 mandatory, L5); actionable cells FailedBreakdown_Up_20D|buy -5 at n=5 and NoTrade|sell -6 at n=10; but the loudest line is the counterfactual - optional_trim n=13, median withheld -3.89% 1wk / -15.33% 4wk, verdict 'cost', up from n=4/-0.9% at the 07-18 review. This book's characteristic error is UNDER-trimming, which is the direct warrant for firing both trims as real DISPOSE cards rather than prose (USC-6/L3) | Valuation fresh (FetchedAt 08-17 22:26) but 20/20 GROWTH_PRICED_IN for a 5th run - zero cross-sectional discrimination, val_adj can only subtract | Book, 10.0%; the two trims free against of buys -; -> (91.2%)

Recommendations

0 to acquire · 2 to dispose

TickerNameCloseActionAction 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
ASMLASML Holding NV1802.98SELL 30%1802.981776.261762.02 (-2.3%)1939.53 (+7.6%)1976.51 (+9.6%)Taking a third of ASML off at because the market's conviction in the setup collapsed in a single session and the position is now sitting almost exactly on its stop. The business news is good - memory makers moving to HBM4/DDR5/3D DRAM pull more EUV lithography through ASML, the 2026 outlook was raised, and the interim dividend went up 17% - but the stock fell 4.3% on 2026-08-18 in a market-wide chip sell-off driven by the 30-year Treasury yield hitting its highest level since 2007, and our own signal score dropped from with it. At the shares are just 1.5% above our exit line, so a third off now converts a fragile position into a smaller one we can hold with conviction rather than one the stop is about to decide for us. We keep exposed to the twelve-week view, which is still a healthy +9.6%. || OPERATIONAL: ADR-0012 profit-take, ARMED (PROFIT_ARMED, T12_Progress 91.22%) and FIRED on tick (b) - conviction, a drop of exactly 20 pts, the threshold. One tick = profit_take_1 = 30% (5 of). Ticks (a)/(c)/(d) did NOT fire: no second Exit_Warning code, 12wk 1976.51 vs the 1.02x dead band 1839.04 is alive, no repeated NO TRADE. G1 does not soften anything here (signal is FailedBreakdown_Up_20D cv61, not a fresh Breakout/MomentumContinuation cv>=80) and tick (b) fires regardless of G1 in any case. VALUATION AMPLIFY BLOCKED BY G3 (forward-view gate): MoS_FV -4.92 <= -3 would ordinarily bump %, but Target_12Week 1976.51 is solidly positive (+9.62%) against a bullish-family signal with price above EMA50/EMA100/EMA200 - trimming half here would be cutting a winner on a valuation read. Amplify suppressed, portion stays 30%. Note the DCF anchor IS reliable on this name (836.55 vs 1802.98 = 2.16x, the cleanest read on the board alongside NVDA), so this is a G3 block, not a G2 block - consistent with the 2026-08-17 entry's assessment. KNOWN DEFECT: PROFIT_ARMED fired on a position that is -1.31% BELOW cost (Gain_ATR -0.35) - the T12-progress arm leg is not gated on a positive gain (learnings advisory (a), unshipped, 33rd consecutive session). Realized on vs the 1826.98 basis: -, so this is a de-risk, not a profit-take, whatever the flag says. STOP 1776.26 STANDS on the retained - the mandatory armed trail-raise is unavailable for a 3rd session because the 2.0xATR chandelier (1666.43) and the model Stop_Price (1666.43) both sit below it, and a trail never moves down. 1wk uses f=0.30 (calib_1w -0.207, degenerate - L5), mirrored to the model stop leg. News classified neutral / NON-OVERRIDING: the fundamental tape is supportive but weeks old and already absorbed; the 08-18 move was rates-driven multiple compression across the whole complex, not an ASML event. · news
TSMTaiwan Semiconductor Manufacturing413.41SELL 30%413.41413.27406.07 (-1.8%)437.87 (+5.9%)406.45 (-1.7%)Trimming another 30% of TSMC at because our twelve-week projection has now gone outright NEGATIVE on the stock - 406.45 against a 413.41 close - and the shares are sitting fourteen cents above the level at which we would sell anyway. TSMC's business could hardly be better: July revenue up 44.7% to NT6bn, Arizona sales up 145% year over year, a Sony image-sensor joint venture, and a dividend going ex on 16 September. But the stock fell over 4% on 2026-08-18 in a broad chip sell-off, and we do not hold names our own model expects to be lower in three months. This follows the 50% cut that filled this morning at; we keep as a defined core and let the stop decide the rest. || OPERATIONAL: ADR-0012 profit-take, ARMED (PROFIT_ARMED, T12_Progress 101.71%) and FIRED on tick (c) for a THIRD consecutive session - and materially worse than the prior two: Target_12Week has gone from +0.55% above the close (08-17) to -1.68% BELOW it, well inside the 1.02x dead band of 421.68. Tick (b) did not re-fire (cv. One tick = profit_take_1 = 30% (7 of). This is NOT a re-issue of the 08-17 card - that one filled this morning (txn ussemicon-20260819-0001, @417.00; this is a fresh fire on newly deteriorated conditions (close -4.08% in one session, 12wk flipped negative). G2 SUPPRESSES the MoS -14.46 amplify for a 3rd session: DCF_per_share 45.97 vs a 413.41 price is an 8.99x scale mismatch (the same class defect prints tonight on TSLA 9.8x, INTC 7.2x, ARM 5.5x) - MoS read directionally only, val_adj 0, no bump. 'Never 100% from a profit-take alone' also stands. L2 is one tick away, not yet met: 413.41 is +0.034% above the 413.27 trail - a negative 12wk is in place but there is no decisive close below stop yet. STOP 413.27 STANDS on the retained; both the model Stop_Price (388.95) and the 2.0xATR chandelier (388.95) sit below it and a trail never moves down. If it breaks tomorrow the stop takes the remaining 17. KNOWN DEFECT (advisory (a)): armed at T12_Progress 101.71% while -5.62% below cost with Gain_ATR -2.01 - the arm leg measures progress toward a 12wk target that has fallen through the price. Realized on vs the 438.02 basis: -. SYSTEMIC: Target_12Week (406.45) prints BELOW Close (413.41) on a bullish-family row - the KLAC/LMT pattern; sanity check wanted in core/predictions.py. 1wk uses f=0.30 (L5). News neutral / non-overriding per USC-5 - this book has already paid once for treating a TSMC record beat as licence to hold an oversized technical laggard. Position size after: =, 0.72% of book. · news
NVDANVIDIA Corporation219.74HOLD211.85206.05n/a222.70 (+5.1%)246.90 (+16.5%)Holding all 253 Nvidia shares and NOT adding, even though the signal turned constructive again. After two sessions of no signal at all, Nvidia printed a buy-the-pullback setup on 2026-08-18 and ranked first on the board - but the entry it wants is, some 3.6% below the close, and after the valuation haircut the setup scores 61 against our 65 minimum. Four points short is not a trade. The stock fell about 2.2% with the whole chip complex as thirty-year Treasury yields hit their highest since 2007 and investors questioned whether AI capital spending can earn its return; the business itself is not in question. We stay fully held with the stop at and wait for the 26 August results. || OPERATIONAL: BUY - ADD cv71, Buy_Rank 1, PullbackBuy_VWAP - the exact family L1 says to favour - but val_adj -10 (MoS_FV -2.56 -> clamp(round(-2.56x5), -10, 0)) takes adj to 61, below the 65 floor. The -10 haircut is NOT suppressed here: NVDA's DCF anchor is the cleanest on the board (163.98 vs 219.74 = 1.34x; Target_12Week 246.90 vs DCF 163.98 = 1.51x is not a wide margin), so note B / G2 does not apply. scorecard_bias 0 (PullbackBuy_VWAP|buy is n=1, report-only). Independently it would have been QUEUED not bought: Entry_Distance_ATR -1.43 means price sits 1.43 ATR above the limit. RE-ARM TRIGGER: with val_adj -10 and bias 0 standing, NVDA needs raw conviction >= 75 on a bullish-actionable print to clear the floor - that is the precise bar to watch at the next EOD. NOT a sell candidate: not armed (Exit_Warning blank, T12_Progress 89.00 < 90, Gain_ATR -1.28), so ADR-0012 does not apply; 219.74 is 6.64% above the 206.05 stop, above EMA200 (197.46), and the 12wk 246.90 = +12.4% is alive. Marks - / -3.11%; risk to stop = 0.35% of book. Stop 206.05 STANDS (model Stop_Price 205.43 is looser). CATALYST: FQ2 FY27 prints 2026-08-26 - SIX sessions out, so the <=5-session catalyst-aware-hold window opens at the NEXT EOD; horizon band +/-6.75%, stance hold-through, with an add-after-print read-through to MU/KLAC. Any future queue must expire before that date or be withdrawn into it (L6). · news
MUMicron Technology940.76WATCH940.76908.16n/a1071.16 (+13.9%)1035.23 (+10.0%)No action on Micron. The 2026-08-17 note called Micron 'the name to watch - this book's cleanest acquire since inception if it prints a second anticipation signal'. It did not: the setup faded to a passive WATCH and the shares fell 7.0% from to in the chip sell-off. Waiting cost us nothing and saved a 7% entry. || OPERATIONAL: anticipate slot NOT eligible - 1-of-3 (BUY -> BUY-ANTICIPATE -> WATCH), needs >=2-of-3. Not bullish-actionable so it never reached the ACQUIRE pool either. Still the only Quality_Pass True unheld candidate on the board; val_adj -7 (MoS -1.35) -> adj 54 even if it had printed. This is the third consecutive session in which the no-chase discipline (USC-4/L6) has been vindicated by the tape. Re-arm: a BUY or BUY-ANTICIPATE print at cv >= 72 clears the 65 floor after the -7 haircut.
LITELumentum Holdings873.31WATCH873.31828.88n/a1051.03 (+20.4%)1026.33 (+17.5%)No action on Lumentum. It was the highest-conviction name on the board last night at 83 and fell 9.9% to today, faded to a passive WATCH, and remains blocked by the quality screen. The optics story is real - Q4 revenue 01bn, +109% year over year, the stock still up ~108% for the year - but it is the most expensive name we track against its own fair value and we have no signal to act on. || OPERATIONAL: Quality_Pass False (hard gate on a new entry), MoS_FV -37.68 (worst on the board) -> val_adj -10 -> adj 41. Anticipate slot 1-of-3. Overlay carries a stale (2026-07-05) 'optics dip-buy supported' line for AAOI/COHR/LITE - noted, not acted on: an overlay override cannot open a hard quality gate.
ASXASE Technology Holding36.41WATCH36.4135.31n/a40.82 (+12.1%)37.20 (+2.2%)No action on ASE Technology - it fell 8.8% to and dropped out of the buy list entirely, one session after we declined it on quality-screen grounds at. Declining it saved 8.8%. || OPERATIONAL: was BUY cv69 Buy_Rank 1 on 08-17, now WATCH cv63 - not bullish-actionable. Quality_Pass False, MoS -10.48 -> val_adj -10, FailedBreakdown_Up_20D|buy bias -5 ACTIONABLE at n=5 (USC-3 reads this cell at the -10 floor) -> adj 43. This book's own realised ASX FailedBreakdown buy on 07-10 lost -8.65%.

Outcome & track record

Accuracy at the time of this record.

26 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 →
-0.207 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.
53 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.

This website is for educational purposes only and does not constitute investment advice. Always do your own research and assess your own risk tolerance before making investment decisions.