Recommendations
0 to acquire · 0 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 |
|---|---|---|---|---|---|---|---|---|---|
| NVDA | NVIDIA Corporation | 209.66 | NO ACTION | 209.66 | 207.19 | 213.00 (+1.6%) | 220.78 (+5.3%) | 212.27 (+1.2%) | Holding all 253 Nvidia shares and cancelling the trim our exit rules had triggered, because the quarter reported after the close answered the exact question that triggered it. The rule fired on a dead twelve-week outlook - our model saw only +1.2% of upside left over three months - but Nvidia then delivered of revenue against expected, of earnings against, data-centre sales of up 117% on the year, and guided the current quarter to -110.1B, roughly 10-14% higher again. The shares traded after hours, up 4.11%, which is already 2.8% above the very target whose exhaustion set off the trim. We would be selling on a forecast the news has just made obsolete. OPERATIONAL: ADR-0012 armed (T12_Progress 98.77) + tick (c) fired - Target_12Week 212.27 is -0.74% through the 1.02x dead band of 213.85 (was alive at +3.06% last night); ticks (a) bare PROFIT_ARMED, (b) = -7 vs the 20-pt bar, (d) HOLD - all clean. Base tier would be 30% = =. CANCELLED on four independent legs: (1) the tick's basis is counterfactual at the post-print price; (2) ADR-0012's catalyst-aware hold governs - earnings 0 days out, no confirmed lower high, no stop break (close 209.66 is +1.18% over the 207.19 trail); (3) L8 mandates repricing on a >~1.5% gap and the gap is +4.11%; (4) Step 7's news layer explicitly permits an override to cancel with catalyst + URL. Logged optional:true with 30 so the counterfactual grader prices what the override cost - and it is on the wrong side of the optional_trim aggregate (n=13, -5.14%/1wk, -15.33%/4wk, 'cost') and of NVDA's last four post-print sessions (-1.8/-5.5/-3.2/-0.8%). RE-ARM WITHOUT APPEAL if the 08-27 settled close prints below 207.19, or a recomputed Target_12Week stays inside its dead band, or any Tier-1/2/3 Exit_Warning code joins PROFIT_ARMED. TRAIL HELD at 207.19 - no legal raise exists off the settled close: chandelier 198.54 and EMA100 205.79 are below the standing stop, EMA50 210.53 and EMA20 214.15 are above the 209.66 close and illegal, and the 1.0x ATR clearance line computes to 204.10. Against the 218.28 post-print reference 207.19 sits exactly 2.00x ATR below the market, which is ADR-0012's specified standalone-PROFIT_ARMED trail. Risk 253 x = = 0.064% of book. G2 suppresses any amplify (DCF 163.98 = 1.28x price). Finding 10: this position is -7.56% underwater yet arms on T12_Progress - the arming leg is structurally unsound. · news |
| ASML | ASML Holding NV | 1745.64 | HOLD | 1745.64 | 1736.47 | 1771.67 (+1.5%) | 1832.42 (+5.0%) | 1747.22 (+0.1%) | Leaving the ASML position alone tonight and letting the sale we already ordered do its work. Our exit rule triggered again on the same ground as the past two sessions - the model's twelve-week outlook has flattened to essentially nothing, +0.1% - but the trim we decided on Monday has not been executed yet, and stacking a second sale on top of an unfilled one would cut half the position out of two orders each labelled 30%. The business itself is in good shape: full-year sales guided to -51.5B against a consensus, an EUV order book stretching into 2028, and TSMC's -64B capex plan underwriting the tool cycle. We are raising the protective stop instead. OPERATIONAL: armed (T12_Progress 99.91) + tick (c) for a 3rd session - Target_12Week 1747.22 vs the 1.02x dead band of 1780.55, forward view decayed +3.55% (08-21) -> +0.82% (08-25) -> +0.09% now. Ticks (a) bare PROFIT_ARMED, (b) flat, (d) HOLD - clean. REAFFIRM-DON'T-STACK: the 08-25 card (SELL 2 of @1744.16, execution market, execute_on 2026-08-26) is still open and unfilled at the executor's one-session lag; tonight brings no new deterioration a second card would express (conviction flat, close +0.08%, VWAP_Distance -0.01%). This is L3's plain intent and is logged as systemic finding 11. The open card is re-validated: 1744.16 against a 1745.64 close is -0.08% away, far inside the ~1.5% L8 reprice threshold. G2 suppresses the MoS -4.48 amplify (DCF 841.33 = 2.07x price; 12wk = 2.08x the anchor); val_adj 0; adj 49+0+4 = 53 on FailedBreakdown_Up_20D|sell (n=10, hit_1w 0.714, bias +4 - decayed from +7/n=9 and +10/n=8, finding 12). TRAIL RAISED 1736.09 -> 1736.47 (EMA50), +, legal at 0.53% below the close; clearance is 0.21x ATR, which finding 9's proposed 1.0x floor would reject - taken on the stated basis that the floor was adopted for catalyst sessions and ASML has none. Chandelier 1658.86 is looser and a trail never loosens. Retained risk on the post-fill : x 4 =. · news |
| TSM | Taiwan Semiconductor Manufacturing C | 417.69 | HOLD | 417.69 | 404.87 | n/a | n/a | 397.67 (-4.8%) | Holding TSMC where it is and letting the single share we already ordered sold complete the de-risk. The model still projects the stock about 4.8% lower over twelve weeks and still labels a name we own with a bearish signal, so the direction of the decision is unchanged - but the position is down to three shares after nine successive trims, the last one is not yet filled, and the next slice would be one more share worth or four hundredths of one percent of the book. That is bookkeeping, not risk management. The underlying business remains excellent: revenue growth guided above 40% on capex of -64B, 3nm ramping into strong demand, and a dividend going ex on 16 September. OPERATIONAL: armed (T12_Progress 105.03) + tick (c) for a 9TH session - Target_12Week 397.67 vs the 1.02x dead band of 426.04 (-4.79%). Ticks (a) bare PROFIT_ARMED, (b) flat, (d) AVOID so the NO-TRADE run stays broken and the 2-consecutive bar is unmet. REAFFIRM-DON'T-STACK (finding 11): the 08-25 card (SELL 1 of @417.41, market, execute_on 2026-08-26) is open and unfilled; tonight was +0.07% with conviction flat and the 12wk actually improving 396.53 -> 397.67 - no new deterioration. TERMINAL-TRANCHE JUDGEMENT, stated as this run's assumption in the absence of a min_trim_notional rule (finding 7): below a lot-scale remainder the correct actions are exit or hold, never slice. G2 suppresses the MoS -13.76 amplify (DCF 45.96 = 9.09x price). Cell FailedBreakout_Down_20D|sell is n=2, hit_1w 0.000, report-only, bias 0; adj 57+0+0 = 57. 1wk/4wk are n/a - the bearish AVOID row prints a blank Target_Price so there is no risk objective to interpolate against and no f is applied. TRAIL RAISED 404.59 -> 404.87 (EMA100), +, clearance = 1.24x ATR - comfortably clear. EMA50 416.87 is 0.20% below the close, inside a single bar's noise, and is declined on finding 9's reasoning; the printed Stop_Price 438.46 is above the market (a bearish signal's stop) and is not a trail. Retained risk on the post-fill : x 2 =. · news |
| MU | Micron Technology, Inc. | 938.40 | WATCH | 938.40 | 914.71 | 966.83 (+3.0%) | 1033.17 (+10.1%) | 997.77 (+6.3%) | Micron came closer to a buy than anything has in twelve sessions and still falls seven points short. It is the only candidate on the board that passes our quality screen, its chart is the cleanest we have - every moving average stacked in order, no sign of being over-extended - and it is the least expensive name in the mandate on our valuation work. The news is a tailwind too: Micron's entire 2026 high-bandwidth-memory capacity is sold out at locked-in prices, with BMO opening coverage at Outperform with a target, though Citi has trimmed its own target to on the risk that memory prices soften in 2027. What holds us back is the setup type - this is the signal family that has lost this book money five times out of five graded - and a label that has flip-flopped across five sessions rather than settling. We would rather wait for a cleaner trigger than force the trade. OPERATIONAL: adj 66 - 6 - 2 = 58 vs the 65 floor. Quality_Pass TRUE (the only bullish-actionable name that clears the gate). val_adj -6 is degree-scaled from MoS -1.107, not the floor, and note B does NOT suppress it - MU's DCF 772.94 vs a 938.40 price is 1.21x, nothing like the broken anchors on LITE (17.7x), MRVL (3.9x) or TSM (8.7x), so the haircut is legitimate valuation information. Signal FailedBreakdown_Up_20D: cell n=5, hit_1w 0.400, hit_4w 0.000, stop_first 0.600, bias -2 (read at -10 per USC-3). Label sequence BUY -> WATCH -> WATCH -> WATCH -> BUY fails the 'stable or fresh' test. USC-3 RE-ARM UNCHANGED: raw >= 82 on a bullish-actionable print, or a PullbackBuy_VWAP confirmation. Second reason to wait: MRVL prints Q2 tonight (08-27 after the close) and the Event Horizon lists it as a direct HBM read-through to MU. Structure EMA5 939.24 > EMA10 936.81 > EMA20 928.29 > EMA50 909.42 > EMA100 817.94 > EMA200 648.13, RSI 51.3, VWAP_Distance +3.92%. · news |
| LITE | Lumentum Holdings | 939.03 | WATCH | 939.03 | 876.09 | 983.71 (+4.8%) | 1087.97 (+15.9%) | 1076.02 (+14.6%) | Passing on Lumentum for a second night. It prints the strongest buy signal on the board again, but it fails our quality screen outright, and the price action has become the chase our rules exist to refuse - up 6.0% on the day on top of 6.7% the day before, a 13% two-day run that leaves it nearly 10% above its volume-weighted average price. The AI-optical demand story behind it is real: revenue up 109% year-on-year to about last quarter, guidance of -1.275B for the current one, and the stock up roughly 131% so far this year. But the most telling detail is that our conviction reading FELL from while the price rose 6%, which is exactly the shape that precedes a round-trip. OPERATIONAL: Quality gate F - hard block on a new entry. MoS_FV -32.51 -> val_adj -10 (floor); 76 - 10 - 2 = adj 64, BELOW the 65 floor even before the gate, and adj 54 on USC-3's mandated -10 read - so it fails twice over. VWAP_Distance widened +5.69% -> +9.80% against EMA20 856.15 with the close at 939.03. Signal family is FailedBreakdown_Up_20D, this book's 5-graded / 0.400-hit / 0.600-stop-first cell, and USC-3 is explicit that a gate override is never a licence to admit a chase-family buy. 1wk shown at f=0.30 per L5. · news |
| MRVL | Marvell Technology | 245.11 | WATCH | 245.11 | 236.79 | 254.82 (+4.0%) | 277.46 (+13.2%) | 285.14 (+16.3%) | Passing on Marvell again, on the same three independent grounds as last night - and the third one is now immediate. It falls well short of our conviction bar once its stretched valuation is accounted for; it fails the quality screen; and it reports earnings tonight after the close, the very session our order would have been working. It also trades about 11% above its volume-weighted average price, the most extended reading in the whole mandate for a second day, so there is no non-chasing entry available even if the other two blocks did not exist. OPERATIONAL: adj 66 - 10 - 2 = 54 vs the 65 floor - it fails on CONVICTION BEFORE the gate is reached. Quality gate F. MoS_FV -7.07 -> val_adj -10 (floor). Q2 FY27 prints 2026-08-27 after the close = the execution session; L6 forbids queueing into it and USC-5 forbids carrying discretionary size through it. VWAP_Distance +11.41%. Signal family FailedBreakdown_Up_20D (cell n=5, hit_1w 0.400, stop_first 0.600, bias -2, read -10 per USC-3). 1wk shown at f=0.30 per L5. |
Outcome & track record
Accuracy at the time of this record.
34 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 →
-0.226 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.