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 |
|---|---|---|---|---|---|---|---|---|---|
| ASML | ASML Holding NV | 1646.19 | SELL ALL | 1646.19 | 1736.47⚠ | n/a | n/a | 1476.83 (-10.3%) | Exit the position in full. After four sessions in which the model simply could not find a trade here, it has now issued an outright instruction to reduce - the strongest sell reading on the board - and the shares closed 5.2% below their protective stop, the widest that gap has been in eleven sessions. The price has also given up its 20-, 50- and 100-day averages, and the model's own twelve-week view now sits more than 10% BELOW today's level. ASML's business case is genuinely strong - July guidance was raised to 43-45bn euros for 2026 with a 30% capacity expansion planned - but that thesis is nearly two months old and has not stopped the shares falling on a day when three-fifths of the sector rose. Holding a position against both its risk line and its forward case is not patience, it is drift. Operational: STRONG SELL - REDUCE cv61, Sell_Rank 1, 77.89, Position_Status CLOSING. L2 basis (decisive sub-stop close 1.98xATR through the 1,736.47 line + 12wk -10.28%) - a trend-break exit, so the ADR-0012 never-100%-from-profit-take cap does not bind and G1/G3 do not apply (they protect uptrending names). Valuation amplify (MoS_FV -4.28 <= -3) bumps the 77.89% engine tier one step to 100%; it does not originate. G2 not engaged (DCF 836.25 is a low anchor, not an ADR artifact). PROFIT_ARMED on this row is a false arm (position -9.90%, Gain_ATR -3.96; systemic finding 2). RallySell_BearTrend|sell is n=4 report-only - no bias applied. Row prints no Target_Price so 1wk/4wk are n/a, not fabricated; advisory linear read of the 12wk path ~1,632 / ~1,590. Seventh re-fire, repriced from 1,682.30. Realized - (-9.90%) on a 1,826.98 cost; proceeds. · news |
| TSM | Taiwan Semiconductor Manufacturing C | 417.01 | SELL 33% | 417.01 | 405.91 | n/a | n/a | 395.99 (-5.0%) | Take one share off a holding the model has gone quiet on, and keep two. TSMC's results could hardly be better - second-quarter revenue up 33.7% on the year, full-year growth guidance lifted to above 40%, record capital spending of -64bn and the next-generation 2-nanometre line ramping - yet the shares have gone nowhere and sit nearly 5% below what this book paid. For a third session running the signal engine can find no tradeable setup here at all, and its own twelve-week view is now 5% BELOW today's price. When the forward case is dead and the signal has fallen silent, trimming part of the position is more disciplined than waiting for the risk line to be tested. The trend itself is not broken, which is why this is a partial step down rather than an exit. Operational: basis CORRECTED from the prior two cards - this is a CLAUDE.md section 9 signal-degradation trim (3 consecutive NO TRADE cv0 on a held name + Target_12Week 395.99 < close), NOT an ADR-0012 profit-take. Hand-derived the position is NOT armed (Gain_ATR -2.34, -4.80% below cost); the printed PROFIT_ARMED comes only from the defective Close/Target_12Week disjunct (systemic finding 2). Not escalated to a full exit: no stop breach (+2.66% headroom over 405.91), close sitting on EMA20 417.34 / EMA50 417.03 and well above EMA100 406.33 and EMA200 376.68. Tier held at, consistent with four prior unfilled cards (50% of 3 = 1.5, rounded DOWN - never round a trim up past its tier). G2 SUPPRESSES the valuation amplify: DCF 46.22 is an ADR-ratio artifact (ADR = 5 ordinary shares), so MoS_FV -13.82 must not bump the tier. G3 satisfied (forward view negative, not cutting a winner). Scorecard NoTrade|sell -5 (n=16, hit_1w.250) argues AGAINST this card and is why the tier was not escalated; counterfactual optional_trim n=15 median -5.06%/-15.33% [cost] argues for firing something. Ex-div on 2026-09-16 is 8 sessions out, outside the <=5-day catalyst hold. Retain, stop 405.91. Fourth re-fire, repriced from 415.50. Realized - (-4.80%); proceeds. · news |
| NVDA | NVIDIA Corporation | 228.45 | HOLD | 228.45 | 215.33 | 243.64 (+6.6%) | 279.08 (+22.2%) | 250.12 (+9.5%) | Hold the full position and lift the protective stop. Nvidia confirmed a roughly $13bn acquisition of the AI model hub Hugging Face, the shares added another 1.8%, and the model's conviction jumped to 93 - the strongest reading this book has ever recorded on this name - with the twelve-week view intact around 9% above today's price. A momentum indicator flashing a mild divergence in the middle of a strengthening uptrend is not a reason to sell a position that has just been reaffirmed; the right response is to move the stop up and keep the shares. Operational: a MECHANICAL read would have trimmed tonight and it is wrong twice over. FIRST, hand-derived NVDA is NOT armed: Gain_ATR 0.21 vs the 3.0 threshold, and true progress-from-cost toward the 12wk target is (228.45-226.80)/(250.12-226.80) = 7.1%, not the printed 91.34% - that column is Close/Target_12Week, a ratio-to-target (systemic finding 2). Position holds of unrealised gain; there is no profit to take and ADR-0012 is not engaged. SECOND, even at face value only tick (a) RSI_DIV lands - conviction ROSE 17 pts (not a >=20 collapse), 12wk 250.12 > 1.02x228.45 = 233.02 alive, label is HOLD not a second NO TRADE - and a single Tier-1 code on a fresh bullish continuation setup at cv93 is exactly what the G1 softening downgrades to a trail-raise (applied by analogy to AnticipationUp_VCP, a fresh bullish anticipation/continuation family at >=80). TRAIL RAISED 211.56 -> 215.33, a real ratchet: the scan's own printed Stop_Price 215.33 beats the 2.0xATR chandelier at 212.51 and now clears EMA50 213.15, which was the objection that blocked the raise on 09-02. Headroom +5.74% / 1.65xATR on 83.9% of equity. Peak stop re-bases to 215.33; ADR-0020 erosion 0.0%. Estimates f=0.30 (calib_1w -0.226 degenerate, L5); note 4wk 279.08 sits ABOVE the 12wk 250.12 - Target_Price is the VCP setup's measured move, not a 4-week base case. 12wk 250.12 exceeds the 163.98 DCF anchor - projection assumes growth beyond priced-in. Value of the standing non-sale: +vs the that would have been trimmed at 217.44 on 08-31; full line +over three sessions. Seasonality pieces flag a historically weak September for NVDA - a calendar argument, not a signal; the 215.33 trail is the answer to it. · news |
| MU | Micron Technology, Inc. | 958.16 | WATCH | 958.16 | 918.88 | 983.17 (+2.6%) | 1041.53 (+8.7%) | 1024.59 (+6.9%) | Near-miss, no action - but the narrowest miss of the run, and the gap has closed from twelve points to five. Micron is the strongest name in this mandate on every fundamental measure: high-bandwidth memory sold out through 2026 on multi-year agreements with customer deposits, the best quality score of any candidate on the board, a trillion-dollar valuation and Wall Street looking for materially higher prices. The signal has improved too - a second consecutive buy print with conviction rising rather than the buy/watch oscillation that sank last night's card, and an entry that is not extended. What still stands in the way is this book's own realised record on this particular setup family, which is bad enough that a full-size position cannot be justified on it, and two dated hazards sitting squarely in the twelve-week path. Tenth consecutive session on the shortlist without a fill. Operational: 76 raw. val_adj -6 (degree-scaled from MoS_FV -1.160, Quality_Pass True, F=8). FailedBreakdown_Up_20D|buy read at the -10 USC-3 floor - the cell is at n=5, the exact trigger USC-3 named, and the apparent improvement is illusory (hit_1w.400 but hit_4w.000 - not one graded FailedBreakdown buy has ever reached its 4wk target; the three realised ones were 3/3 stop-first). 76 - 6 - 10 = 60 vs the 65 floor. DISCLOSED: at the scorer's mechanical bias of -2 this clears at 68 - the -10 is USC-3, an ACTIVE learning binding under Step 1.4, not the scorecard's own number. USC-3's half-starter route on a passing quality gate was considered and DECLINED - a five-point shortfall is not a rounding error. Hazards: FOMC 2026-09-16 (argued to move MU harder than its own print) and FQ4 2026-09-30 (L6 - do not chase into it). Overlay MU override (add only on signal, not on narrative) honoured - a signal appeared and lost on its merits; overlay is 44 trading days stale regardless. Size mod 0.75x recorded for reference only. 12wk 1,024.59 exceeds the 772.94 DCF anchor. · news |
| TSLA | Tesla Inc | 376.36 | WATCH | 372.26 | 311.19 | 412.00 (+10.7%) | 494.40 (+32.8%) | 349.92 (-6.0%) | Blocked, no action, and blocked three separate ways. Tesla jumped more than 5% into its Cybercab reveal in Austin, helped by a favourable range revision on the Model Y - which is an event-driven run, not a trading signal, and buying the day before a product launch is precisely the kind of chase that has cost this book money before. The shares also fail the fundamental quality screen outright, carry one of the most extreme overvaluation readings in the mandate, and remain below their own 200-day average. Most tellingly, the model's own twelve-week projection sits 7% BELOW today's price: the engine is calling this a buy and forecasting a lower price at the same time. Operational: (1) Quality_Pass False = hard gate on new entries; USC-3 - a gate override is never a licence to chase. (2) Arithmetic fails independently: 74 - 10 (val_adj clamped, MoS_FV -18.72, DCF 34.25) - 0 = 64 < 65 floor. scorecard_bias 0 by ABSENCE of any graded BreakoutUp_20D|buy cell, not by merit. (3) Already-triggered-extended: Order_Type is a Buy Stop @372.26, BELOW the 376.36 close (Entry_Distance_ATR -0.25) - a chase by construction, and itself a data defect (systemic finding 3). Close 376.36 < EMA200 380.93. Target_12Week 349.92 = -7.03% vs close (systemic finding 4). News classified OVERRIDES: the move is a catalyst run into a scheduled reveal, L1/L6. target_1w shown is an advisory linear read, not a model estimate - the trade was never taken. · news |
| LITE | Lumentum Holdings | 847.37 | WATCH | 847.37 | 818.89 | 881.05 (+4.0%) | 959.65 (+13.3%) | 919.15 (+8.5%) | Blocked, no action - fourth consecutive session of vindication. Lumentum fails the fundamental quality screen outright and carries by far the most extreme overvaluation reading in the mandate. The signal itself has now decayed from a buy print in late August to a watch-list reading at barely half the entry threshold, and the shares are roughly 11% below the levels at which this book repeatedly declined to buy them. Operational: quality gate F blocks new entries (hard). val_adj -10 (clamped from MoS_FV -34.07 x 5). FailedBreakdown_Up_20D|buy at the -10 USC-3 floor. 42 - 10 - 10 = 22. Conviction and price 956.14 -> 847.37 since the 08-26/08-27 refusals; counterfactual grader should price those refusals favourably. |
| SNDK | Sandisk Corporation | 1554.99 | WATCH | 1554.99 | 1507.88 | 1611.52 (+3.6%) | 1743.43 (+12.1%) | 1530.11 (-1.6%) | Near-miss, no action, and a third session of the no-chase rule proving itself. Sandisk printed an outright buy signal on 31 August; it was deliberately not queued, and the signal has since decayed to a watch-list reading barely half the strength of the entry threshold while the shares have gone essentially nowhere. Buying one-day spikes in this mandate has not worked. Operational: val_adj -6 (MoS_FV -1.106, Quality_Pass True). FailedBreakdown_Up_20D|buy at the -10 USC-3 floor. 49 - 6 - 10 = 33 vs the 65 floor. Third consecutive WATCH 49 print after the 08-31 BUY 67. Note the engine's own 12wk 1,530.11 is BELOW the 4wk target 1,743.43 and barely above the close - a further reason the watch label is correct. |
Outcome & track record
Accuracy at the time of this record.
42 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.