Recommendations
1 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 |
|---|---|---|---|---|---|---|---|---|---|
| AMD | Advanced Micro Devices, Inc | 516.13 | BUY | 516.13 | 501.35 | 528.37 (+2.4%) | 556.92 (+7.9%) | 535.25 (+3.7%) | Advanced Micro Devices is the strongest setup on our board and the first name in nine sessions to clear our buying bar. The shares closed above every moving average we track, our conviction reading rose rather than faded, and the entry price sits right at the closing price - we are not chasing a spike. Behind the chart, data-centre revenue has more than doubled from a year ago on a major OpenAI commitment, the new MI450 accelerator begins shipping this quarter, and the average analyst target sits about 19% above today's price with no sell ratings on the stock. We are deliberately starting at half our normal size: this particular signal pattern has a poor track record in this book, so we want the position to prove itself before we commit the rest. OPERATIONAL: cleared the floor at 76-10=66 with USC-3's pre-committed -10 FailedBreakdown floor applied UNCHANGED from 09-10 - the name grew into the bar, the bar did not move (declined at adj 58 last night). Half-starter 0.5x is over-determined: size modulator (MoS -11.22 <= -3) AND USC-3 half-starter both land on 0.5x independently. val_adj suppressed to 0 under learnings note B (DCF 171.74 / FV_per_share 42.64 vs a price is a broken anchor, not a valuation signal). Sizing: (35.06/100) x book x (20/100) =, x0.50 =, /516.13 = =. Entry range 511.03-521.23 (+-0.25xATR 20.39). R:R 2.76:1. f=0.30 on the 1wk per L5 (calib_1w -0.226, degenerate/negative). 12wk 535.25 EXCEEDS DCF anchor 171.74 - projection assumes growth beyond priced-in; note also the 12wk prints BELOW the 4wk (systemic finding 3). L8 re-validation at the Mon 09-14 open: reprice if it gaps >1.5% above 516.13; WITHDRAW if conviction falls below the 65 floor on the Monday board. · news |
| NVDA | NVIDIA Corporation | 218.29 | SELL 30% | 218.29 | 205.91 | 222.00 (+1.7%) | 230.67 (+5.7%) | 232.04 (+6.3%) | We are reducing Nvidia by 30% and keeping the other 70%. The shares have now closed below our risk line for a second straight session, and - the part that changed our mind - on a day when the rest of the chip sector bounced about 1.6%, Nvidia finished flat, among the five weakest names on our list. Falling with the market is normal; failing to rise with it is the position telling us its weakness is its own. Nothing is wrong with the business: data-centre revenue grew 117% last quarter and management has publicly reaffirmed confidence in roughly 70% growth next year, which is precisely why we are trimming rather than selling out. Money is visibly rotating from Nvidia into AMD and Qualcomm, and we are following it with part of the proceeds. OPERATIONAL: executes the conversion trigger pre-committed in the 09-10 memory entry ('2nd consecutive close below 221.79 -> mandatory 30% de-risk, at the next open'). Origination is CLAUDE.md section 9 signal degradation: conviction -> over four sessions, signal label degraded to [WATCH], Momentum 32, two consecutive sub-trail closes. Explicitly NOT L2 (12wk +6.30%, conjunction fails), NOT ADR-0012 (arm is the phantom PROFIT_ARMED on a -3.75% position; no tick fires anyway - conviction -12 is short of the >=20 bar), NOT ADR-0020 (erosion 0.00% vs peak 221.79), NOT valuation (MoS -1.93 is above the -3 amplify threshold; section 9 is amplify-only regardless). Capped at 30% by G3 (12wk solidly positive, label HOLD not SELL-REDUCE) and USC-5 (close above EMA50 214.95 / EMA100 209.23 / EMA200 201.67, Consolidation_Bullish intact = fade, not break). Proceeds; position ->; retained. STOP on the retained 177 RE-BASED to the engine's live 205.91 and peak_stop reset to 205.91 - the 221.79 trail is retired having triggered the cut (ADR-0020 re-bases the watermark after a size reduction). NEW CONVERSION TRIGGER: any close below EMA50 214.95 -> further mandatory 30% de-risk at the next open. Cost if NVDA simply works: ~5.7% on over 4wk. f=0.30 on the 1wk per L5. · news |
| ASML | ASML Holding NV | 1698.30 | SELL 30% | 1698.30 | 1622.58 | n/a | n/a | 1571.94 (-7.4%) | We are trimming 30% of ASML and keeping the rest. Our model has now returned no tradable signal on it for a second straight session with conviction at zero, the price has slipped below every short and medium-term moving average, and the twelve-week outlook has turned negative at about -7%. Our rule is that a held position which sustainably loses its signal gets reduced, and this is the second confirming print. The business news is actually getting better - 2026 sales guidance of -51bn sits well above expectations, orders for its lithography machines stretch into 2028, and Intel has become the first chipmaker to put its next-generation High-NA tools into production - which is exactly why this is a partial trim and not an exit. OPERATIONAL: executes the conversion trigger pre-committed in the 09-10 memory entry ('2nd consecutive NO TRADE / Conflicted print -> mandatory 30% de-risk, at the following open'). Origination is CLAUDE.md section 9 'sustained NoTrade/Conflicted on a held position'. Explicitly NOT L2 (12wk leg met at -7.44% but the close is +4.67% ABOVE the 1,622.58 stop - no stop-break, conjunction fails), NOT ADR-0012 (phantom arm on a -7.04% position), NOT ADR-0020 (erosion -6.56%, inside 15%). VALUATION AMPLIFY SUPPRESSED BY G2: MoS -4.46 <= -3 would normally bump the tier %, but the DCF anchor 839.37 vs a 1,698.30 price is the same broken-anchor pattern G2/note B exist to catch - trim stays at the technical's own 30% tier. Conviction trail ->; Pillar_Consensus Indeterminate for a 2nd session; NoTrade_Reason 'Consensus conflict + low conviction'; Exit_Strategy none. 1wk/4wk are n/a because the engine prints no Target_Price on a NO TRADE row - none invented. Proceeds; position ->; retained. STOP on the retained 4 HELD at 1,622.58 (never lowered; engine prints no stop on a NO TRADE row); peak_stop 1,736.47 unchanged. Materiality noted honestly: against a book - the discipline matters more than the dollars. · news |
| TSM | Taiwan Semiconductor Manufacturing C | 433.24 | HOLD | 433.24 | 415.38 | 439.13 (+1.4%) | 452.86 (+4.5%) | 441.16 (+1.8%) | Holding Taiwan Semiconductor - it is the strongest position we own and it got stronger this week. Our conviction reading rose from, the shares trade above every moving average we track, and August revenue grew 53% from a year ago as AI demand keeps its factories fully booked. There is nothing here to sell. The stock goes ex-dividend on 16 September. OPERATIONAL: trail HELD at 415.38 - the engine's live Stop_Price is 413.62, BELOW the existing line, and a trail never ratchets down; close is +4.30% (+1.82xATR) above it. PROFIT_ARMED is phantom (-1.09%, Gain_ATR -0.49) and DISCARDED per learnings advisory (a); had it been trusted, ADR-0012 tick (c) would have fired on a 12wk of 441.16 against the 1.02x threshold of 441.90 and manufactured a phantom profit-take trim on the book's best technical - systemic finding 2. Trend 78 / Momentum 53, Consolidation_Bullish. TSM ex-div on 2026-09-16. · news |
| — | Marvell Technology | — | NO ACTION | — | — | n/a | n/a | n/a | Marvell popped 4% on a new Piper Sandler buy rating with a target, and we are still not buying it. It fails our financial-quality screen outright, which blocks new positions regardless of how good the story sounds, and even setting that aside our adjusted conviction lands 16 points below our threshold. This is the third time in four sessions this stock has jumped on an analyst soundbite in this book; the previous two gave the move straight back. Its investor day on 6 October is the event that would actually settle the question. OPERATIONAL: Quality_Pass FALSE (F-Score 6, Quality_Rank 13/20) = HARD GATE on new entries. 69 -10 (MoS -7.04) -10 (USC-3 FailedBreakdown floor) = 49. Buy_Rank 2, entry 236.10 = close. L6 / the 09-09 precedent applied. |
| — | ASE Technology Holding | — | NO ACTION | — | — | n/a | n/a | n/a | Not buying ASE Technology for a fifth straight session. It fails our financial-quality screen on negative free cash flow - it is spending heavily on a new advanced-packaging build - and that is a hard block on any new position. Our adjusted conviction is 19 points short of the bar in any case, and this is one of three names where this exact signal pattern has already stopped us out for a loss in this book. OPERATIONAL: Quality_Pass FALSE (FCF yield -14.2%, Quality_Rank 16/20) = HARD GATE. 66 -10 (MoS -10.85) -10 (USC-3) = 46. Realized stop-out 07-10 at -8.65%. Anticipate precondition now 1-of-3 (B-ANTIC 09-09, then BUY, BUY) - slot ineligible. |
| — | Micron Technology, Inc. | — | NO ACTION | — | — | n/a | n/a | n/a | Micron stays on the watchlist, not in the portfolio. The buy setup we flagged as imminent two sessions ago never materialised - the signal weakened to a simple watch and has stayed there for a second session. The business passes our quality screen comfortably; there is simply no entry trigger to act on. Its quarterly results land on 30 September. OPERATIONAL: WATCH cv61, 2nd consecutive session; the 09-09 re-arm prediction is confirmed DEAD, not deferred. Anticipate precondition 1-of-3. adj 61-7=54, 11 short. Quality_Pass TRUE. MU FQ4 earnings 2026-09-30. |
| — | Lumentum Holdings | — | NO ACTION | — | — | n/a | n/a | n/a | Lumentum remains blocked. It fails our financial-quality screen, which bars a new position, and separately there is no entry trigger left to queue - the model's buy level has collapsed to the closing price itself, so there is no level above the market to buy strength into. OPERATIONAL: WATCH cv61, 2nd session. Quality_Pass FALSE (Quality_Rank 14/20) = HARD GATE. Entry_Price 927.03 == close (Entry_Distance_ATR -0.0), so no buy-stop exists; carrying the old 1,006.49 trigger would violate the action-price provenance rule. Anticipate precondition 1-of-3. val_adj suppressed to 0 under note B. |
| — | Sandisk Corporation | — | NO ACTION | — | — | n/a | n/a | n/a | Sandisk has the best financial quality in this mandate and is the only name our valuation work calls fairly priced rather than expensive - but it was the single worst performer on a day the sector rose 1.6%, and the chart is going the wrong way. Quality does not override price action. We keep watching for a pullback entry that has not yet appeared. OPERATIONAL: WATCH cv54, 2nd session, -3.50% on a +1.59% median day. Quality_Rank 1/20, the mandate's sole FAIR_VALUE row. adj 54-7=47, 18 short. The 09-10 note ('watch for a PullbackBuy_VWAP re-entry') remains unfilled - it prints WATCH, not a pullback-buy. |
| — | Qualcomm Incorporated | — | NO ACTION | — | — | n/a | n/a | n/a | Qualcomm finally produced a breakout signal, and we are letting it go. Our conviction in it fell from in the same session the breakout appeared, the buy level sits above the current price so nothing would fill without chasing, it is the most overbought name on our board, and it fails our quality screen. A setup losing conviction into its own breakout is one to leave alone. OPERATIONAL: WATCH cv51, entry 185.21 with Entry_Distance_ATR +0.39 (unfilled buy-stop), RSI 66.29 (highest on the board), Quality_Pass FALSE. adj 51-6=45, 20 short. Note the 12wk (177.00) prints BELOW the close (181.97) - a dead forward view on the breakout row. |
Outcome & track record
Accuracy at the time of this record.
45 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.