Recommendations
1 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 |
|---|---|---|---|---|---|---|---|---|---|
| PLTR | Palantir Technologies Inc. | 179.94 | BUY — QUEUED @156.52 | 156.52 | 149.34 | 157.55 (+0.7%) | 167.20 (+6.8%) | 256.41 (+63.8%) | We are placing a limit order to buy Palantir, but only at a much better price than Friday's close. Palantir is the strongest name on our board — our highest conviction reading since this book began, on our most reliable signal type, backed by a quarter that grew revenue 93% with US commercial sales up 149% and full-year guidance raised. A technical piece published that same Friday flagged as the level that opens the next leg higher, and the shares closed at, just through it. The problem is price: at roughly 80 times next year's earnings the stock is expensive on any measure we apply, and it closed 15% above the average price it traded at during the session. So we buy at the model's own entry of — the session's volume-weighted average — or we do not buy at all. Half normal size, risking 0.3% of the book. || Ops: adj cv 68 = 78 raw − 10 val_adj (bias 0, UT-8 tenth week). Quality gate PASSES (F 8, QRank 5). val_adj is NOT an EPV artifact — vs the DCF anchor 46.62 the MoS is still −2.86, caps at −10 either way (relevant to 08-21 systemic finding 8). UT-6/L1 no-chase binds hard: Entry_Distance_ATR −3.27, +14.97% above VWAP, most extended row on the board — QUEUE, never chase. Fill probability over 3 sessions is LOW (limit is −13.0% below close); a non-fill is the order working, not a failed decision. Sizing: 66.95% × × 20% =; × 0.5 size mod (MoS ≤ −3) =; ÷ 156.52 = = = 6.68% of. Defined risk = 0.31%. Entry range 154.73–158.31 (±0.25 ATR). f = 0.096; 1wk is non-informative at this calibration. 12wk 256.41 ⚠ exceeds DCF anchor 46.62 by 5.5× — carries no decision weight. CELL WARNING (printed, not applied): own-book PullbackBuy_VWAP|buy is 1-for-4 at 1wk, 0-for-1 at 4wk, below the n≥5 floor. L6 does not bind (Q2 print is weeks old, not a gap-chase). Expires 2026-08-26; withdraw or reprice at the next EOD per the queue-expiry rule. · news |
| DELL | Dell Technologies Inc | 442.08 | SELL 50% | 442.08 | 406.44 | 447.03 (+1.1%) | 493.63 (+11.7%) | 572.13 (+29.4%) | We are selling half our Dell holding at Monday's open to honour a stop-loss that broke four sessions ago and has not healed. Dell fell decisively through its risk line at and has closed below it every session since, including Friday, when a 1.7% bounce still ended below the day's average price and below its 20-day trend line. This is the same trade we decided twice already and that has not reached the market either time — our automated executor has not run in four sessions — so it goes out as a plain market order rather than being written down a third time. We keep the other half: the twelve-week view is still 29% higher, quality metrics are the best of our three holdings, and Dell reports earnings in early September with a $51bn AI backlog and consensus looking for 110% earnings growth. Selling the broken half and keeping the half exposed to that print is the whole idea. || Ops: SECOND re-fire under UT-3/L3 (armed 08-20 for the 08-20 open, escalated MOO 08-21, neither filled; ustech_executor_state.json last_run_date 2026-08-18 = 4 sessions stale, portfolio still reads DELL 713, trades_detected empty). Execution-gap accounting reported honestly: the delay HELPED this time — ref 434.78 → 442.08 = +/sh = +on, and +/sh / +vs the original 08-20 arm of 437.55; realized loss shrinks − → −. That is a favourable coin-flip, NOT a vindication — the same lag on 08-19's −6.64% session would have cost L2 INCOMPLETE (stop leg only; T12 572.13 = +29.42% positive, consensus Consolidation_Bullish, conv is +2 not a −20 collapse) → size stays 50%, never 100%. Valuation amplify (MoS −3.178 ≤ −3 would bump SUPPRESSED by G2 for a 3rd session: T12 572.13 vs DCF 485.61 = +17.8% exceeds-anchor, FV_method EPV, EPV 112.78 vs DCF 485.61 = 4.31× intra-model spread — val_adj 0. ADR-0020 did NOT fire: peak_stop 455.32 vs live 406.44 = 10.74% erosion with zero size reduction, 4.3 pts from the 15% us-family trigger; executing re-bases the watermark legitimately. Proceeds; realized −; residual = = 16.27% of, open risk to 406.44 = = 1.31%. RESIDUAL STOP HELD AT 406.44 (ratchet: 2.0×ATR below the 444.43 high = 392.88, lower, so no raise; engine Stop_Price 429.19 = 0.50×ATR is too tight and hand-overridden for the 3rd consecutive session). FQ2 earnings date CONTESTED 09-01 (Zacks coverage this week) vs 09-03 (08-21 read) — plan on 09-01, which shortens the residual's window by two sessions. 12wk 572.13 ⚠ exceeds DCF anchor 485.61. · news |
| AAOI | Applied Optoelectronics, Inc. | 124.82 | HOLD | 124.82 | 111.93 | n/a | n/a | 112.17 (-10.1%) | We are holding Applied Optoelectronics without adding or trimming, and it is the least comfortable hold in the book. The shares gave back half of Thursday's bounce, closing down 3.3% at, and our model's twelve-week projection has now sat below the current price for four straight sessions and keeps falling — in plain terms, the model sees no upside from here. What it does not see is a broken risk line: our stop sits 11.5% below the price, and our exit rule requires both a stop break and a dead forward view before it fires. So the rule holds and we do nothing. The underlying business case is intact — revenue up 140% year on year, a multi-year Amazon supply agreement, and capacity scaling toward 900,000 units a month by late 2027 — against continued losses and real execution risk. The position is small at 4.4% of the book and is down 16.8%. || Ops: L2 INCOMPLETE — forward-view leg armed since 08-19 (T12 112.17 = −10.14% below close, monotonic decay 135.97→116.18→120.41→112.17), stop leg 11.51% away → AND-gate correctly blocks the trim. NO TRADE cv0, NoTrade_Reason verbatim: 'Consensus conflict + low conviction'; regime Ranging, consensus Indeterminate; below EMA20 127.08 and EMA50 129.76, above EMA200 106.52; closed −0.16% vs VWAP 125.02 on 8.02m sh. UT-9 controls the arm for a TENTH session: PROFIT_ARMED prints while Gain_ATR −1.61 on a −16.80% position — arm-on-loss, INERT, never a profit-take. Valuation can neither originate nor amplify: Val_Signal NA, FV_per_share −3.38 (negative EPV — the screen cannot value the company), Quality_Pass False, F 4. ADR-0020: peak_stop = live stop = 111.93, 0% erosion, no fire. Position = 4.41% of; open risk to stop = 0.46%; unrealized −. News classified confirms-hold but flagged NOT tape-anchored — the pieces found quote price references against a live 124.82, so treated as undated background, not fresh catalyst. FLAGGED TO THE LEARNINGS REVIEW (systemic finding 7): a sustained dead-12wk leg (N consecutive sessions with Target_12Week < Close) is a candidate deterioration tick in its own right; not acted on here. · news |
| NVDA | NVIDIA Corporation | 214.72 | HOLD | 214.72 | 206.05 | n/a | 224.95 (+4.8%) | 230.42 (+7.3%) | We are holding Nvidia through Wednesday's earnings rather than trimming ahead of them. Our conviction reading on the name has fallen hard — from over four sessions — and we are naming that rather than ignoring it. But none of our three grounds for selling is present: the price is 4.2% above our stop, the twelve-week view is still 7% higher, and the signal has not turned bearish. Our rule for a holding with earnings within a week is to hold or tighten the trailing stop through the event, never to pre-trim into it. The setup itself is favourable: Nvidia guided to $91bn of revenue, roughly double a year ago, and has beaten estimates each of the last four quarters — though what matters is the forward guidance number, not the beat. The honest risk is that the options market implies a 6.75% move, which puts our stop inside the range a merely ordinary bad reaction would reach. Position is small at 3.7% of the book. || Ops: catalyst-aware hold (CLAUDE.md §9) — NVDA Q2 FY2027 08-26 after close = 3 sessions out, inside the ≤5-day window; trim requires a CONFIRMED lower high or a decisive stop break, neither exists. Conviction → = −29 over four snapshots but −12 vs the prior EOD, short of the −20 collapse tick — no fire. No exit basis on any origination class: no stop break (+4.21% above 206.05), T12 230.42 = +7.31% positive, no SELL label. UT-9: PROFIT_ARMED INERT (Gain_ATR −2.36 on a −5.33% position). NO mandatory trail-raise — the standalone-PROFIT_ARMED 2.0×ATR tighten computes to 208.51 (high 218.74 − 2×5.117) but derives from an inert arm, and 208.51 is no less deep inside the band than 206.05; engine Stop_Price 204.49 is lower and rejected by the ratchet. 206.05 HOLDS. ADR-0020: peak_stop = live stop = 206.05, 0% erosion. Implied band from 08-28 contracts ±6.75% off 214.72 = – — THE STOP IS INSIDE THE BAND, which is the whole risk statement. REVISIT AT THE 08-25 CLOSE, the last session before the print. L6 stands: no add before 08-26; the horizon's add-after-print read-through to MU/KLAC/NBIS/NET is a POST-print licence only. Position = 3.73% of; open risk = 0.15%; unrealized −. Consensus 8bn (+96.4% YoY), EPS, 75% non-GAAP GM, China DC compute excluded from the guide; four-quarter beat history 5.42/6.58/4.84/3.96%. · news |
| SNOW | Snowflake Inc. | 332.78 | WATCH | 332.78 | 314.20 | n/a | 358.01 (+7.6%) | 474.21 (+42.5%) | We passed on Snowflake for the second week running, and it fails on two independent counts. Our raw conviction of 69 clears the entry bar, but the shares look expensive enough on our valuation work to cut that to 59 — below the threshold — and separately the company fails our financial-quality screen, which blocks any new position outright. The shares rose 3.6% to and would have been a straightforward market buy rather than a limit order, so nothing about the price stopped us; the screens did. Worth noting the signal type also degraded this session to one our own records treat with suspicion. Snowflake reports around 2 September. || Ops: adj 59 = 69 raw − 10 val_adj (MoS −1.946, bias 0) → fails the 65 floor; Quality_Pass False (F 5, QRank 20.5) blocks new entries independently — either alone is sufficient. Entry_Distance_ATR 0.0 = market order, no chase problem. Signal DEGRADED PullbackBuy_VWAP (3 sessions) → FailedBreakdown_Up_20D, which sits on the L1/UT-12 discount list — the family change argues against forcing it. Closed +7.28% above VWAP 310.19, above EMA20 314.78. Ledgered direction:none so the counterfactual grader prices what the gate cost or saved. |
| MU | Micron Technology, Inc. | 966.78 | WATCH | 966.78 | 940.19 | n/a | 1073.12 (+11.0%) | 1078.41 (+11.5%) | Micron was the name we came closest to buying last session, missing by three points, and we wrote down exactly what would have to happen for us to buy it. Neither condition was met — conviction fell rather than rose, the shares slipped 0.8%, and Micron dropped off our buy list entirely. We are recording the miss because a condition that is correctly specified and then correctly not met is as much a decision as one that fires. Micron remains the highest-quality name on our board and the most likely candidate to return after Nvidia reports on Wednesday, which our event calendar flags as a direct read-through to memory makers. || Ops: 08-21 re-arm was written as 'raw conviction ≥ 72 at the current MoS, OR a pullback lifting MoS_FV above −1.0.' Actual: conviction, action BUY → WATCH, MoS_FV −1.349 (unchanged), close −0.77% to 966.78 — neither leg met, name out of the buy pool. adj 47 = 54 − 7. Vindicates writing re-arms on price AND conviction rather than conviction alone (08-20 systemic finding 7). Quality_Pass True, F 8, Quality_Rank 1 — still the single highest-quality row in the book; the gate would pass. Relevant to 08-21 systemic finding 8: MU's val_adj −7 derives from an EPV FV_method with a 1.79× EPV/DCF spread (EPV 432.15 vs DCF 772.94) — the proposed halved-haircut fix WOULD have mattered here, where it would NOT have mattered for PLTR. Horizon: NVDA 08-26 carries an explicit add-after-print read-through to MU; L6 says the disciplined entry is after the print, not before it. |
Outcome & track record
Accuracy at the time of this record.
19 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 →
0.096 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.