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 |
|---|---|---|---|---|---|---|---|---|---|
| 8AZ | Aztech Global | 0.58 | WATCH | 0.58 | 0.56 | n/a | 0.65 (+12.1%) | 0.33 (-43.1%) | Aztech is cheap on our valuation work and deeply oversold, and for the third session running the model has cut its own score on exactly the same bounce setup - conviction 80 on the sixth of August, 68 last night, 63 tonight, while the shares fell another 1.7%. We are declining it for the third time, on the same two grounds: it fails our financial-strength screen, which bars a new position outright, and our own twelve-week view sits forty-three percent below the current price. The engine is steadily converging on the position we took two weeks ago. OPERATIONAL: WATCH is NOT a bullish actionable Action, so this row never enters the buy pool - adj 65 sits EXACTLY AT the 65 floor, so the floor is NOT the binding constraint and this must not be reported as a conviction near-miss (third night this caveat applies). Three blocks in precedence order: (1) Action label WATCH; (2) Quality_Pass False = hard gate on NEW entries, which also blocked adj 82 on 08-06 and cannot be out-scored; (3) T12 0.33 vs 0.580 close = -43.1%. adj 65 = cv63 + val_adj +2 (UNDERVALUED MoS +0.220 x20 = +4, HALVED for failed quality) + bias 0. RSI 18.3 (new low for the move), VWAP -10.44%, below EMA20/50/200. Gate PASSED clean at a 1.724% range on SGD 1.05m turnover - the cleanest pass on tonight's board. T12 UNPINNED tonight (0.34 -> 0.33) after eleven frozen sessions, which downgrades the prior 'floored or clipped' systemic finding to 'sticky'. REFUSAL LEDGER: 08-06 BUY cv80/adj82 leg, card 0.62, L8 exec 08-07 open 0.6200 vs 0.5800 tonight = -6.45% avoided (from -4.85%). RE-ARM: leg one (breakdown must fail first) remains SATISFIED; leg two is binding and unmet - FY2026 revenue stabilisation repairing Quality_Pass, THEN a settled favoured-cell BUY >= 65 with a forward view not forty percent below price. Do NOT re-derive as a live BUY on leg one alone (SGM-4). News: none found. · news |
| 5DD | Micro-Mechanics (Holdings) | 2.76 | WATCH | 2.76 | 2.72 | n/a | 2.91 (+5.4%) | 2.51 (-9.1%) | Micro-Mechanics did not move at all today, yet the model's confidence in it swung fifteen points back up - recovering the entire drop it took last night on an identically unchanged price. A score that round-trips that far while the shares stand still is an argument about the score's stability, not about the stock. It remains the weakest kind of setup in our playbook and is nowhere near a purchase. OPERATIONAL: WATCH cv43 -> cv58 (+15, full recovery of last night's -15) on a COMPLETELY UNCHANGED 2.760 close; RSI identical to 13 significant figures, which is arithmetically correct for a zero-change bar under Wilder smoothing - NOT a stale-data artefact. adj 51 = 58 + val_adj -7 (GROWTH_PRICED_IN MoS -1.426) + bias 0, fourteen short of the floor. Volume rebounded to 107,200 from 36,600 (2.93x); gate PASSED clean at a 2.536% range on SGD 296k. FailedBreakdown_Up_20D is L1's explicitly discounted cell (~0.25 hit, ~60% stop-first). RSI 49.2, T12 2.51 = -9.1% below close. Logged to systemic finding 6: a +/-15pt conviction swing on a flat bar suggests volume-derived pillar inputs dominate on ultra-thin tape. RE-ARM: settled PullbackBuy_VWAP or MomentumContinuation at RAW cv >= 65 with a positive T12; a recovering FailedBreakdown does not qualify (MSCAP-1). News: none found. |
| 575 | ASTI Holdings | 0.12 | WATCH | 0.12 | 0.11 | n/a | 0.13 (+8.3%) | 0.11 (-8.3%) | ASTI stopped falling and produced its first buy-side flag in weeks, but the flag arrives just as our own twelve-week projection for the stock has dropped BELOW today's price - the forward view has swung from twelve percent up a week ago to six percent down tonight. A bounce signal on a name our model expects to be lower in three months is not something we act on, and the score is twenty points short of our threshold in any case. OPERATIONAL: NEW WATCH cv55, up from NO TRADE cv0 after three sessions of clean 'Liquidity gate failed' - its range collapsed from 10.256% to 2.564% and the gate let it through on SGD 675k. adj 45 = 55 + val_adj -10 (GROWTH_PRICED_IN MoS -2.385, clamped) + bias 0. Closed 0.117 UNCHANGED. THE DISQUALIFIER IS NOT THE ARITHMETIC: Target_12Week has fallen below the close - four-session sequence 0.15 -> 0.13 -> 0.12 -> 0.11 against a price of 0.133 -> 0.125 -> 0.117 -> 0.117, so the forward view travelled from +12.8% on 08-18 to -6.0% tonight. WORSE, the 4-week Target_Price prints 0.13 (+11%), ABOVE the 12-week number - an internal inconsistency logged as systemic finding 2 (KLAC/LMT pattern; candidate assertion in core/predictions.py). Entry_Distance_ATR +0.39 means the 0.12 entry sits 0.39 ATR ABOVE the close and has not triggered. RSI 41.8, VWAP -7.21%. Valuation drift caveat: -12.0% against the 08-18 surface, still the least reliable label in the book. News: none found. |
| S71 | Sunright | 0.46 | WATCH | — | — | n/a | n/a | 0.26 | Sunright is the cheapest name in this mandate on our valuation work and today it effectively did not trade at all - twelve hundred shares changed hands, about five hundred and fifty dollars' worth, with the price identical at every point of the session. Our own twelve-week projection still puts it more than forty percent below today's price. A stock this untradeable is arguably uninvestable at any size this book would, cheap or not. OPERATIONAL: O = H = L = C = 0.460 on = SGD 552 turnover, the thinnest print in this book's history by two orders of magnitude. THIS ROW IS TONIGHT'S GATE FALSIFIER AND IT CORRECTS THE RECORD: it FAILED the liquidity gate on a 0.000% realised range, which no pure-range gate can do. The single-variable range hypothesis this book has been narrowing for a week - bracket (2.913%, 3.191%) - is FALSIFIED OUTRIGHT, not tightened, and the repeated prior conclusion that the 'liquidity' LABEL was refuted at both ends is RETRACTED: the label is sound. A two-condition spec (fail if range wide OR turnover below a floor) explains all nine clean rows tonight AND retro-explains every prior anomaly - the 08-20 5DD-pass/S71-fail 'inversion' resolves because S71's range that day was 5.435%, and AWX's SGD 31.1m fail was a 4.560% range. Turnover floor now bracketed at (SGD 552, SGD 101,016]; the range bracket SURVIVES at (2.913%, 3.191%) once S71 is reclassified as a turnover-floor fail. Shadow-test the compound specification, NEVER hand-tune (systemic finding 1). UNDERVALUED MoS +0.459 gives the largest positive val_adj in the book (+9) and it changes nothing: no chart, no signal, T12 0.26 = -43.5%. RSI 36.8, VWAP -3.85%. SGM-5 in its purest form. RE-ARM: settled favoured-cell BUY >= 65 with a POSITIVE forward view - the negative T12 is the binding objection, not the valuation. News: none found. |
| M14 | InnoTek | 0.44 | WATCH | — | — | n/a | n/a | 0.25 | InnoTek was the only name in the mandate to rise today, up nearly five percent - and tonight we finally have the explanation for the collapse it is bouncing from. First-half profit fell sixty-four percent, to a hundred and forty-five thousand dollars, and the shares dropped over ten percent on the release about a week ago. That is precisely the window in which the watch-level signal we declined at fifty-one cents went underwater. Our refusal was not luck: the discipline kept us out of a company whose earnings deteriorated by two-thirds within days. Today's bounce is the classic reflex after that kind of drop, and our twelve-week view is still forty-two percent lower. OPERATIONAL: closed 0.435 (+4.82%) on (0.82x) - the ONLY advancing name on the board; still NO TRADE cv0, clean 'Liquidity gate failed' at a 5.747% range on SGD 747k (range condition, turnover fine). RSI recovered 24.5 -> 31.6 but VWAP -11.74% remains the weakest in the book. T12 0.25 = -42.5% vs close. GROWTH_PRICED_IN MoS -0.330, val_adj -2, QPass False. NEWS RESOLVES A STANDING 'UNEXPLAINED' FLAG: 1H FY2026 attributable profit SGvs SGa year earlier (-64%); shares plunged >10% at the close on the release. L6 applies literally - do NOT chase the gap. REFUSAL LEDGER: the declined 08-12 WATCH cv68 card (0.51, L8 exec 08-13 open 0.510) marks -14.71% AVOIDED, DOWN from -18.63% - the bounce shrank the book's largest-ever favourable leg by four points in one bar, the honest entry in tonight's table and the same volatility that flipped PCT's sign five times in five sessions (MSCAP-12: recorded, never re-litigates the floor). The leg is nonetheless now the BEST-EVIDENCED of the five, because a 64% profit decline is attached to it rather than a chart pattern. · news |
| I07 | ISDN Holdings | 0.71 | WATCH | — | — | n/a | n/a | 0.84 | ISDN's unexplained jump three sessions ago now has an explanation: the company launched two new precision-motion products at a trade show in Taipei on the nineteenth. That is a marketing announcement, not an order book or an earnings release - which is exactly why the move is unwinding rather than holding. The shares have now fallen for three straight sessions and given back most of the spike. Waiting was correct, and we now know why it was correct rather than merely that it was. OPERATIONAL: closed 0.715 (-4.03%) on (1.11x), third consecutive decline, 0.775 peak -> 0.715. Gate reason flipped back from masked to a clean 'Liquidity gate failed' at a 5.594% range on SGD 4.40m, continuing the masked/clean alternation. RSI cooling 61.3 -> 54.9; VWAP distance collapsing +5.83% -> +1.36% (from +9.68% at the peak). T12 0.84 = +17.5%, one of only four positive forward views left in the book, but eroding (0.87 -> 0.90 -> 0.89 -> 0.84). GROWTH_PRICED_IN MoS -0.742, val_adj -4, QPass False. NEWS: ISDN Precision System launched the ApexTwin MFF dual-gantry motion system (back-end packaging) and a 33mm vacuum low-profile precision module at Automation Taipei 2026 on 2026-08-19 - dates exactly to the spike. Classified CONFIRMS the wait-it-out read. · news |
| AWX | AEM Holdings | 9.36 | WATCH | — | — | n/a | n/a | 10.65 | AEM fell a further three percent, a fourth consecutive decline, and its momentum has now turned outright neutral for the first time in this stretch. The business news is fine - the company went ex-dividend on the nineteenth, reaffirmed full-year revenue guidance of four hundred and sixty to five hundred and ten million dollars, and the stock is still up around ninety percent this year - so this is a strong name giving back an extended run, not a company in trouble. That distinction is the whole content of our two standing refusals here, both of which improved again tonight. OPERATIONAL: closed 9.360 (-3.01%) on; clean 'Liquidity gate failed' at a 4.915% range on SGD 42.4m - the deepest tape in the book, and the range condition (not turnover) is what fires, per tonight's corrected two-condition gate spec. RSI 49.4, BELOW 50 for the first time in the stretch. T12 10.65 = +13.8%, eroding in every session (12.22 -> 11.90 -> 11.38 -> 10.65). GROWTH_PRICED_IN MoS -10.844 clamps val_adj to -10 (MSCAP-2 saturation - identical clamp to 575's MoS -2.385 despite a 4.5x difference); QPass False. NEWS: ex-dividend SGD 0.024 on 2026-08-19 (payable September) - about 0.25% of the share price, so it CANNOT account for tonight's -3.01%; FY2026 guidance S-510m intact, Q1 revenue +36% y/y, net margin 12.3% vs 3.9%. Classified NEUTRAL. REFUSAL LEDGER: both legs improved again - 08-12 floor card 9.63 / L8 exec 08-13 open 10.00 marks -6.40% avoided (from -3.50%); 08-13 chase card 10.98 / exec 08-14 open 11.00 marks -14.91% avoided (from -12.27%). L1's no-chase rule remains the highest-value discipline in this book. THIS IS THE NVIDIA READ-THROUGH NAME: NVIDIA Q2 FY2027 lands 08-26 (SGX reaction 08-27), 3 sessions out. · news |
| PCT | PC Partner Group | 2.76 | WATCH | — | — | n/a | n/a | 3.45 | PC Partner remains the strongest fundamental case in this mandate - the only name our twelve-week view still sees meaningfully higher, at about twenty-five percent up, on a fair valuation that passes our financial-strength screen - and for the eleventh session in twelve it has been blocked by the tradability filter. It is untouchable while it trades this erratically, and its forward view is eroding alongside everything else here. OPERATIONAL: closed 2.760 (-1.43%) on; clean 'Liquidity gate failed' at a 3.623% range on SGD 5.07m - a range-condition fail under tonight's corrected two-condition spec, well above the (2.913%, 3.191%) bracket. RSI 46.6, VWAP -4.46%. T12 3.45 = +25.0%, BEST IN BOOK, but eroding hard (4.09 -> 4.05 -> 3.66 -> 3.45). FAIR_VALUE MoS -0.020, val_adj 0, QPass True. REFUSAL LEDGER: the 08-06 PullbackBuy_VWAP cv62 leg (card 2.74, L8 exec 08-07 open 2.86) marks -3.50% AVOIDED, improving from -2.10%; sixth different reading in six sessions on an unchanged decision (+12.9 -> +7.00 -> +2.10 -> +3.50 -> -2.10 -> -3.50). Valuation drift caveat: -7.8% cumulative against the 08-18 surface. RE-ARM: settled PullbackBuy_VWAP/MomentumContinuation, gate-cleared (range under ~2.9%), RAW cv >= 65. News: none found. |
| 558 | UMS Integration | 2.77 | WATCH | — | — | n/a | n/a | 3.38 | UMS held up better than most of the mandate again, down under two percent, and it still carries the second-best twelve-week outlook here at about twenty-two percent up. It also remains blocked on two fronts at once - richly priced against our valuation work and failing our financial-strength screen - so it cannot clear even on a good chart. Nothing about tonight changes that. OPERATIONAL: closed 2.770 (-1.77%) on; clean 'Liquidity gate failed' at a 3.610% range on SGD 17.68m - a range-condition fail, comfortably above the bracket. RSI 60.0, cooling from 63.4 but STILL THE MOST EXTENDED in the book; VWAP +10.11%. T12 3.38 = +22.0%. GROWTH_PRICED_IN MoS -4.559 clamps val_adj to -10; QPass False = the SGM-4a double block. RE-ARM: requires a valuation AND quality re-rate, not a chart event. News: none found. NVIDIA Q2 FY2027 on 08-26 (SGX reaction 08-27) is a read-through for this name - 3 sessions out. |
| E28 | Frencken Group | 2.58 | WATCH | — | — | n/a | n/a | 2.18 | Frencken fell for a second session and our indicators are still disagreeing about it, so we cannot even see whether the tradability filter would have blocked it. More to the point, our own twelve-week projection for the shares has now dropped to about fifteen percent BELOW today's price and is deteriorating in a straight line - so even a clean buy signal here would be arguing against our own forward view. OPERATIONAL: closed 2.580 (-1.15%) on (0.85x); reason string 'Consensus conflict + low conviction' for a SECOND straight session, so gate status is masked and the row is EXCLUDED from the bracket; its 2.713% range would most likely have passed, which is a loss - a clean pass there would have raised the (2.913%, 3.191%) lower bound. RSI 45.6 and falling. T12 2.18 = -15.5%, deteriorating fast (2.33 -> 2.39 -> 2.25 -> 2.18). FAIR_VALUE MoS -1.341 -> val_adj -7, which per MSCAP-2 over-penalises a fair-valued small-cap; flagged, not overridden. QPass True. News: none found. NVIDIA Q2 FY2027 on 08-26 (SGX reaction 08-27) is a read-through - 3 sessions out. |
| BN2 | Valuetronics Holdings | 1.02 | WATCH | — | — | n/a | n/a | 0.97 | Valuetronics stayed quiet for a second session - unchanged on the day, with the bearish signal it carried on and off for three weeks still extinguished. We hold none of it and are long-only, so this changes nothing either way; it simply means the board has no bearish signal-carrier left at all, for the second night running. OPERATIONAL: NO TRADE cv0 for a second session; closed 1.020 UNCHANGED on (4.13x, reversing last night's 78% volume collapse). Reason 'Consensus conflict + low conviction', so gate status is masked and the 1.961% range is EXCLUDED from the bracket - again a loss, since a clean pass there would have loosened the lower bound. NOTE: BN2's 2.913% clean pass on 08-19 remains the highest clean pass ever recorded here and is the standing lower bound of the (2.913%, 3.191%) range bracket, which SURVIVES tonight only because S71's 0.000%-range fail is correctly reclassified as a turnover-floor fail rather than a range fail (systemic finding 1). RSI 49.7, VWAP +0.75%, T12 0.97 = -4.9%. GROWTH_PRICED_IN MoS -0.136 -> val_adj -1. QPass False. Long-only + = no action either way (SGM-4b). News: none found. |
Outcome & track record
Accuracy at the time of this record.
0 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.
0.000 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.
50 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.