US Semiconductors — End-of-Day Verdict
0 ACQUIRE, 0 ANTICIPATE, 1 DISPOSE | RE-ISSUE the TSM 50% trim - 23 of @ 430.97 =, retaining a 24-share core, stop 413.27. The 08-14 card decided this trade and it NEVER EXECUTED: the auto-executor ran this morning (0 filled / 0 unfilled / 9 skipped) and the card was not among the nine - it was written execution='market_next_open', outside _EXEC_KINDS, and was silently dropped. Across this book's TSM history 13 sell cards were written and exactly 2 filled - the 2 written execution='market'. Re-derived fresh tonight TSM fires only tick (c) (12wk 433.32 vs the 439.59 dead band, 2nd consecutive session; tick (b) did NOT re-fire, cv which is profit_take_1 = 30%/ - but this is a RECONCILE of a standing decision, not a fresh origination, so it stays at 50%/ per L3/USC-6; repriced 426.35 -> 430.97 = +1.08%, inside L8's 1.5% band. G2 again suppresses the MoS -14.46 amplify (DCF 45.97 vs 429.66 = 9.3x scale mismatch - and tonight the same break shows on TSLA 9.9x, INTC 7.7x, ARM 6.0x, so it is a class of names, not one ADR slip) | ACQUIRE NO ACTION - ASX was the sole bullish-actionable row (BUY cv69, Buy_Rank 1) and failed four ways: hard quality gate (Quality_Pass False), MoS -10.48 -> val_adj -10, FailedBreakdown_Up_20D|buy bias -5 now ACTIONABLE at n=5 -> adj 54 (USC-3's -10 floor reading takes it to 49), and this book's own ASX FailedBreakdown buy on 07-10 realized -8.65% | ANTICIPATE none eligible - three names printed AnticipationUp_VCP [STRONG BUY] above the 55 floor and NOT ONE satisfies >=2-of-last-3. LITE (cv83, highest on the board) printed ANTICIPATE on 08-13 AND 08-18 with WATCH/BUY between = 2-of-4, never 2-of-3, and is quality-gate blocked anyway (MoS -37.68, worst on board) despite a genuine inflection (Q4 rev +109.3% YoY, +108% YTD, PT). SNDK (cv76) is 1-of-3, gate-blocked, and ALREADY EXTENDED at Entry_Distance_ATR -0.48 - the buy-stop sits half an ATR BELOW price so it would fill at market, destroying the fills-only-on-strength property. MU (cv81) is 1-of-3 and is THE NAME TO WATCH: the ONLY Quality_Pass True candidate on the board, MoS -1.35 -> val_adj -7 -> adj 74, entry 1017.79 = +0.10 ATR clean, with the memory tape confirming (Samsung supply warning, Musk memory-bottleneck, BofA Buy 'structurally stronger'). If MU re-prints BUY - ANTICIPATE at the 08-18 EOD it is 2-of-3 and becomes this book's cleanest acquire since inception | NVDA 2nd consecutive NO TRADE cv0 and still a HOLD - NOT armed (T12_Progress 87.93 < 90, Gain_ATR -0.34) so ADR-0012 tick (d) does not apply; a still-constructive fade not a broken technical (9.2% above stop, above EMA200/EMA50/VWAP, 12wk 255.91 = +13.7%), the USC-5/US-3 distinction; and NoTrade|sell is the book's worst cell at n=10, bias -6 ACTIONABLE, pricing a NoTrade-basis sell as a mistake. FQ2 FY27 08-26 = 7 sessions out, catalyst window opens at the 08-19 EOD | ASML armed-but-NOT-fired (cv ROSE, 12wk 2088.75 alive at +10.9%) - the mandatory trail-raise is UNAVAILABLE for a 2nd session because the 2.0xATR chandelier (1766.16) and model stop (1762.58) both sit below the standing 1776.26, which STANDS; marks + / +3.07% | Valuation layer still blind cross-sectionally: GROWTH_PRICED_IN, val_adj can only subtract, and the quality gate fails 3 of the 4 bullish candidates - 5 consecutive sessions of ACQUIRE NO ACTION at 88.7% cash | Book, 11.3%; the trim frees against of buys, ->
Recommendations
0 to acquire · 1 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 | 430.97 | SELL 50% | 430.97 | 413.27 | 424.83 (-1.4%) | 451.45 (+4.8%) | 433.32 (+0.5%) | Taking half the TSMC position off the table at while it is still comfortably above its stop. The business is doing well - Q2 revenue (+33.7% YoY) with margins above guidance, the 2026 growth outlook raised to over 40%, July revenue +45%, and a new 1-trillion-yen image-sensor joint venture with Sony - but the shares have already run to where our own twelve-week projection sits (433.32 against a 430.97 close, just +0.55%), so there is no modelled upside left to hold the full size for. We keep a 24-share core to stay exposed to the AI-capacity story and bank the rest as. The valuation screen reads expensive but its underlying fair-value anchor is unreliable for this stock, so it is not driving the decision. OPERATIONAL: ADR-0012 profit-take, armed (PROFIT_ARMED, T12_Progress 99.46%) with tick (c) firing for a 2nd consecutive session; tick (b) did NOT re-fire (cv. This RE-ISSUES the 2026-08-14 50% card (ticks b+c) which never executed - it was written execution='market_next_open', outside paper_executor._EXEC_KINDS, and _eligibility dropped it silently (absent even from the run's skip list). Decision is a reconcile, not a fresh origination, so it stays at profit_take_2 = 50%/ rather than re-deriving down to 30%/ (L3, USC-6). Repriced 426.35 -> 430.97 = +1.08%, inside L8's 1.5% band. G2 SUPPRESSES the MoS -14.46 amplify: DCF_per_share 45.97 vs price 429.66 is a 9.3x scale mismatch (same defect on TSLA 9.9x, INTC 7.7x, ARM 6.0x) - directional read only, val_adj 0, no bump; 'never 100% from a profit-take alone' also stands. Not a stop break (430.97 is 4.3% above the 413.27 trail, above EMA50 417.42 / EMA200 370.33 / VWAP 409.97); not ADR-0020 (trail erosion 0%, six ratchets up). Stop 413.27 STANDS on the retained - model Stop_Price 410.49 is looser and a trail never moves down. 1wk uses f=0.30 (calib_1w -0.207 degenerate, L5) mirrored to the model stop leg. News neutral/non-overriding per USC-5 (TSM fell 3-4% on the record beat). Realized on vs 438.02 basis: -. · news |
| NVDA | NVIDIA Corporation | 225.01 | HOLD | 225.01 | 206.05 | n/a | n/a | 255.91 (+13.7%) | Holding all 253 NVIDIA shares into the 26 August quarterly results. The signal engine has gone quiet on the stock - its trend and momentum readings disagree, so it prints no tradeable setup - but a quiet signal is not a broken one: the shares sit 9.2% above our stop, above every major moving average, and our own twelve-week projection of is nearly 14% higher than today's price. The position is barely underwater at -0.8%, three sessions old. We would rather let the earnings print resolve the direction than sell into indecision. OPERATIONAL: 2nd consecutive NO TRADE cv0 ('Consensus conflict + low conviction', Conflicted, Momentum 16) - the count flagged on 08-14 as '1 print, not 2'. Does NOT convert to a sell: (1) ADR-0012 tick (d) requires an ARMED position and NVDA is not armed (Exit_Warning blank, T12_Progress 87.93 < 90, Gain_ATR -0.34); (2) USC-5 counterpoint / US-3 - this is a still-constructive fade (above stop, EMA200 196.81, EMA50 209.40, VWAP 210.27; Trend 71; 12wk +13.7%), not a broken technical, and AMD's identical 07-09 cv76->49 fade was correctly held and rebounded to +8.1%; (3) NoTrade|sell is the book's worst graded cell at n=10, hit_1w 0.20, bias -6 ACTIONABLE - the scorecard prices a NoTrade-basis sell as a mistake. Stop 206.05 held, NOT tightened: risk is only = 0.49% of book, and a 2.0xATR chandelier (214.43) would sit above recent lows and invite a whipsaw exit into the catalyst. FQ2 FY27 prints 08-26 = 7 sessions out; the <=5-session catalyst-aware-hold window opens at the 08-19 EOD - re-check then. News neutral: 5T cap, SpaceX stake, HIVE cloud deal, but +2% YTD vs S&P +7% and the 'soars after Aug 26' pieces are hype, not confirmation. · news |
| ASML | ASML Holding NV | 1883.12 | HOLD | 1883.12 | 1776.26 | n/a | 2230.42 (+18.4%) | 2088.75 (+10.9%) | Holding all 18 ASML shares and leaving the stop where it is. The position is up 3.1% since we bought it a week ago and the setup has strengthened rather than faded - conviction rose from, the shares have pushed above their 50-day average, and our twelve-week projection of is still nearly 11% above today's price. The stock has reached 90% of that projection, which puts it on the watch list for taking something off, but nothing has actually deteriorated yet, so the right move is to keep the risk line where it is and let the position run. ASML is also one of the few names here whose valuation model looks trustworthy. OPERATIONAL: PROFIT_ARMED (T12_Progress 90.16%) but ARMED-NOT-FIRED - zero ticks: (a) no 2nd Exit_Warning code; (b) conviction ROSE; (c) 12wk 2088.75 vs 1.02 x 1883.12 = 1920.78, alive at +10.9%; (d) no repeated NO TRADE. Rule mandates a trail-RAISE, and the raise is UNAVAILABLE: the tightened 2.0xATR chandelier computes to 1766.16 and the model Stop_Price prints 1762.58, both BELOW the standing 1776.26 - a trail never moves down, so 1776.26 STANDS (2nd consecutive session this raise has been unavailable). Marks +. Above EMA50 1730.20 and VWAP 1710.47. Valuation clean: Quality_Pass True, MoS -4.92, DCF 836.55 vs price 1868.35 = 2.2x, NOT scale-broken (contrast TSM 9.3x) - so unlike TSM there is no G2 flag here, it simply has no sell to amplify. |
| ASX | ASE Technology Holding | 39.91 | WATCH | 39.91 | 38.70 | n/a | 43.40 (+8.7%) | 42.91 (+7.5%) | Passing on ASE Technology despite it ranking first on the buy list. It fails our quality screen on profitability and cash generation, it looks expensive against our estimate of what the business is worth, and the specific setup being flagged here is the single worst-performing pattern in this portfolio's history - every one of these entries we have taken stopped out at a loss within a week, including ASE itself in July. Ranking first among a weak field is not a reason to buy. OPERATIONAL: sole bullish-actionable row (BUY cv69, Buy_Rank 1) and blocked four independent ways. (1) HARD QUALITY GATE - Quality_Pass False (F-Score/FCF screen) blocks new entries on a -0 name; USC-3 is explicit that a gate override is never a licence to chase a FailedBreakdown-family buy. (2) MoS_FV -10.48 -> val_adj -10 (clamped floor). (3) FailedBreakdown_Up_20D|buy bias -5, ACTIONABLE at n=5 -> adj 54, eleven under the 65 floor; USC-3's standing instruction to read this cell at the -10 floor takes it to 49. (4) Direct precedent - this book's own ASX FailedBreakdown buy on 07-10 realized -8.65% and is one of the three entries behind USC-3's 3/3 stop-first record (cell now n=5, hit_1w 0.25, hit_4w 0.00). Re-arm only on a PullbackBuy_VWAP print or a passing quality gate, and then only as a half-starter per USC-3. |
| MU | Micron Technology | 1011.75 | WATCH | 1017.79 | 837.13 | n/a | 1379.12 (+35.5%) | 1125.03 (+10.5%) | Micron is the most attractive thing on this board and we are still not buying it tonight - the setup needs one more day to confirm. It is the only name in the whole portfolio that passes our quality screen, it is only mildly expensive against our fair-value estimate, and the memory market backdrop is as strong as it has been: Samsung warning of tightening supply, analysts calling the cycle structurally stronger with a target, and management saying demand exceeds supply into 2028. Our discipline requires the entry setup to print two days out of three before we commit, and today is only the first. If it prints again tomorrow this becomes the cleanest buy this book has had since inception. OPERATIONAL: BUY - ANTICIPATE cv81, AnticipationUp_VCP [STRONG BUY], Trend 91. Anticipate slot requires >=2 of the last 3 snapshots; MU is 1-of-3 (NO TRADE 08-14 -> BUY 08-15 -> ANTICIPATE 08-18) so no card. Fusion arithmetic if it were eligible: 81 + val_adj -7 (MoS -1.35, mildest on the board) + bias 0 (AnticipationUp_VCP|buy n=2, report-only, hit_1w 0.0 - warn, do not adjust) = adj 74, well clear of the 55 anticipate floor AND the 65 acquire floor. ONLY Quality_Pass True name among the four bullish candidates (ASX/LITE/SNDK all fail). Entry 1017.79 vs close 1011.75 = +0.10 ATR - clean buy-stop above price, fills only on strength, exactly the USC-4 vehicle that is 3-for-3 in this book. Standing MU overlay override ('ride the trail; add only on signal, not on narrative') is satisfied by waiting for the 2nd print. WATCH THIS FIRST at the 08-18 EOD. · news |
| LITE | Lumentum Holdings | 968.90 | WATCH | 975.06 | 721.18 | n/a | 1482.83 (+52.1%) | 1169.12 (+19.9%) | Lumentum has the strongest signal on the board and the strongest recent results - revenue more than doubled last quarter on AI data-centre optics, and analysts see roughly 15% further upside - but we are not buying it. It has already doubled this year, it trades at a very large premium to any defensible estimate of what the business is worth, and it fails our balance-sheet and cash-generation quality screen, which is a hard block on opening a new position. Buying a stock this extended, on results published six sessions ago, is the exact trade this book's rules were written to prevent. OPERATIONAL: highest conviction on the board (cv83, Trend 93) and blocked twice over. (1) Anticipate slot needs >=2 of the last 3 snapshots - LITE printed ANTICIPATE on 08-13 AND 08-18 but WATCH/BUY between, so it is 2-of-4 and NEVER 2-of-3 (see systemic finding 5: the window may be narrower than the signal's own oscillation; any widening is gated by the champion/challenger shadow test). (2) HARD QUALITY GATE - Quality_Pass False blocks a -0 new entry regardless of the slot. Also MoS_FV -37.68, the worst on the board -> val_adj -10 (adj 73, which would clear the floor, so the gate is doing the work here, not the arithmetic). L6 independently blocks it: Q4 FY26 printed 08-11 (rev +109.3% YoY, non-GAAP GM >50%, OpM 36.6%, EPS) and the stock gapped +13.6% the next session; +108% YTD; consensus PT = +14.7% vs the 968.90 close, Strong Buy. Re-arm only if the quality gate flips. · news |
| SNDK | Sandisk Corporation | 1786.85 | WATCH | 1712.05 | 1241.62 | n/a | 2652.91 (+55.0%) | 1724.87 (+0.7%) | Passing on SanDisk. The memory backdrop is genuinely strong and the stock is riding it, but the price has already run past the level our system wanted to buy at - an order placed here would fill instantly rather than waiting for the stock to prove itself, which turns a disciplined entry into a chase. It also fails our quality screen. We would rather own the memory theme through Micron, which passes quality and is still trading below its trigger. OPERATIONAL: BUY - ANTICIPATE cv76 and blocked three ways. (1) Anticipate slot 1-of-3 (AVOID 08-14 -> AVOID 08-15 -> ANTICIPATE 08-18). (2) HARD QUALITY GATE - Quality_Pass False on a -0 new entry (EPV_per_share is NEGATIVE at -44.64, so the MoS -0.76 / val_adj -4 reading is not a genuine 'nearly fair' signal). (3) ALREADY EXTENDED - Entry_Distance_ATR -0.48: close 1786.85 is half an ATR ABOVE the 1712.05 buy-stop, so the 'fills only on strength' property that makes the USC-4 queue safe is gone and the order would execute at market on placement. Stop 1241.62 is also 3.0 ATR / -30.5% from the close, the same unbounded-stop geometry flagged on AAOI 08-14. Prefer MU for this theme. News confirms the sector (Samsung supply warning, Musk flagging memory as the AI bottleneck, 08-13 investor day, PT hikes) but the sector being right does not fix the entry. · news |
Outcome & track record
Accuracy at the time of this record.
25 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.