Recommendations
0 to acquire · 3 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 |
|---|---|---|---|---|---|---|---|---|---|
| 3988 | Bank of China | 5.88 | SELL 30% | 5.88 | 5.79 | 5.90 (+0.4%) | 5.90 (+0.4%) | 5.89 (+0.3%) | Bank of China is still the book's biggest winner at +8.0%, but the trend has gone flat: the twelve-week model view now sits essentially on top of the current price, conviction has drifted lower for a second session, and the stop is only 0.6 ATR below the market with no room left to trail. We bank a further ~29% of the position into a dense macro fortnight - China inflation lands tomorrow and the Fed decides on 17 September - and hold the rest. This is a profit-take into a fading trend, not an exit; the signal is not bearish. Operational: armed genuinely on Gain_ATR 3.03 vs a real cost of 5.44 (HF-2 inapplicable, Est_12W +0.26% non-inverted); ONE fire tick only - (c) dead forward view, 5.89 < 1.02x5.875 = 5.9925 - so profit_take_1 = 30% minimum tier; (a) no (RSI_DIV cleared 09-07), (b) cv = -4, below the 20pt bar, (d) no. Volume 30% x 7,000 = 2,100 rounded to nearest 1,000-share board lot = 2,000 (28.6% effective). Valuation silent (NA_CCY, val_adj 0, no tier bump); bias 0 on FailedBreakdown_Up_20D|sell (n=8, hit_1w 0.50). Stop 5.79 UNCHANGED = max(live 5.79, engine 5.59, chandelier 5.588), cushion 1.45% / 0.59 ATR, tightest in the book. Caveat disclosed: tick (c) fired 5-of-5 across the sleeve for an 8th consecutive session and is non-selecting for HK banks; counterfactual optional_trim n=86 median +5.96% wh_4w [saved] argues against. Residual. L8: executes at the 2026-09-09 open; reprice if the open gaps >~1.5%. · news |
| 0939 | CCB - H SHS | 9.56 | SELL 30% | 9.56 | 9.37 | 9.58 (+0.2%) | 9.69 (+1.4%) | 9.71 (+1.6%) | China Construction Bank has reached 98.5% of the model's twelve-week objective at a +3.9% gain, and it is by some distance the most event-exposed name in this book: China's August inflation print lands tomorrow morning, with the credit data, the activity data and the loan-rate fixing all inside the next nine sessions. With less than 0.8 ATR of room to the stop in front of that, we take a third off and carry the balance through the data. Stated plainly, this is a trim taken for exposure rather than for deterioration - the model still sees upside from here and the name was today's only holding to close green - so it is sized at the minimum tier. Operational: the arm is GENUINE, not the 2888/3968 inversion artifact (T12_Progress 98.46 with Est_12W 9.71 = +1.57% NON-inverted, so HF-2 suppression is inapplicable; Gain_ATR 1.44 is below the 3.0 bar, which is why the T12 leg had to be checked). ONE fire tick, and the weak one: (c) 9.71 < 1.02x9.56 = 9.7512, by only 0.42%. Tick (b) fires the OTHER way - conviction ROSE, the only holding to gain conviction tonight. profit_take_1 = 30%. Volume: 30% x 3,000 = 900, below one 1,000-share board lot, rounded UP to the minimum executable lot = 1,000 (33.3% effective, +3.3pp tier overshoot, disclosed). Valuation silent (NA_CCY); bias 0. Stop 9.37 UNCHANGED = max(live 9.37, engine 9.06, chandelier 9.061), cushion 1.99% / 0.76 ATR. Most defeasible card in the verdict: counterfactual optional_trim n=86 [saved] argues directly against it, and a dispersion-aware tick (c) would not have fired it at all. Residual. L8: executes at the 2026-09-09 open. · news |
| 3328 | Bank of Communications | 7.86 | SELL 30% | 7.86 | 7.61 | 7.85 (-0.1%) | 7.90 (+0.5%) | 7.91 (+0.6%) | Bank of Communications is 99.4% of the way to the model's twelve-week objective at a +4.8% gain, with a forward view that is now flat rather than rising and conviction drifting lower for a second session. It is the smallest line in the book, it has the widest stop cushion, and we take the minimum profit-take tier off ahead of a fortnight carrying China inflation, US CPI and the Fed. Nothing here is bearish - the twelve-week estimate is still marginally above the price - this simply banks part of a completed move. Operational: arm GENUINE (T12_Progress 99.37 with Est_12W 7.91 = +0.64% NON-inverted, HF-2 inapplicable). ONE fire tick: (c) 7.91 < 1.02x7.86 = 8.0172; (b) cv = -4, below the bar; (a) and (d) do not fire. profit_take_1 = 30%. LOT OVERSHOOT DISCLOSED: 30% x 2,500 = = 0.75 of a board lot - flooring executes nothing (the 'never got cut' failure mode), so tonight's uniform rule is nearest lot with a floor of one lot, giving = 40% EFFECTIVE, a third above the intended tier. That is a sizing-fidelity defect in the lot logic, not a judgement, and it is filed as a systemic finding. Valuation silent (NA_CCY, no tier bump); bias 0. Stop 7.61 UNCHANGED = max(live 7.61, engine 7.51, chandelier 7.508), cushion 3.18% / 1.42 ATR, widest in the sleeve. F-Score 3 on tonight's refetch is the weakest in the book - advisory only, valuation never originates a sell. Residual. L8: executes at the 2026-09-09 open. · news |
| 2888 | STANCHART | 239.40 | HOLD | 239.40 | 232.73 | 237.71 (-0.7%) | 233.13 (-2.6%) | 230.51 (-3.7%) | Standard Chartered is the book's largest position and tonight it threw its biggest warning: the model's conviction in the setup fell from while the share price barely moved. That is the engine withdrawing a pattern rather than the market repricing the business, and with the shares still trending near their highs and comfortably above our protective level, the right response is to tighten the stop rather than sell into strength. We raise the stop to 232.73 and hold the full through the Fed and HKMA decisions on 17 September. Operational: HF-2 SUPPRESSES the profit-take arm - T12_Progress 103.86 is manufactured by an Est_12W 3.71% BELOW the close (inversion artifact), and Gain_ATR 2.48 is under the 3.0 bar, so the position is not genuinely armed and the ADR-0012 fire rule never runs, even though tick (b) fires hard (cv -29, past the 20pt bar) and tick (c) fires too. HF-2's required independent bearish basis is absent: label still bullish (AnticipationUp_VCP), close 1.53 ATR clear of the raised stop, no formal SELL. Flagged as the verdict's most defeasible hold, and the 6th consecutive session in which HF-2's mechanism is applied past the letter of its oversold-family clause. Stop = max(live 232.00, engine 232.73, chandelier 230.70) -> 232.73 (raised from 232.00), cushion 2.79% / 1.53 ATR - this executes the Event Horizon's tighten-trail stance into the 16-17 Sep FOMC/HKMA pair and satisfies the stale overlay's hold-unless-below-stop override. If HF-2 gains a conviction-collapse override clause at the next learnings review, this becomes a 30% trim. · news |
| 3968 | China Merchants Bank | 51.90 | HOLD | 51.90 | 50.88 | 51.75 (-0.3%) | 52.13 (+0.4%) | 51.74 (-0.3%) | China Merchants Bank did nothing today - it closed unchanged to the tick and the model's read on it was equally unchanged. There is no deterioration to act on and no fresh strength to add to, so we hold the full at a +6.7% gain with the stop where it is. Operational: HF-2 suppresses the profit-take arm (T12_Progress 100.31 is an inversion artifact, Est_12W 51.74 = -0.31% vs close; Gain_ATR 2.67 below the 3.0 bar), and tonight - unlike 09-07 - NO genuine tick fires against it: tick (b) does not fire (cv, only the non-selecting tick (c) does. Yesterday's uncomfortable 'held on HF-2's letter against two live deterioration ticks' has therefore resolved itself; the line is simply dormant and correctly held. Stop = max(live 50.88, engine 49.47, chandelier 49.469) -> 50.88 UNCHANGED, cushion 1.97% / 0.84 ATR. Valuation silent (NA_CCY); bias 0. · news |
| 0966 | China Taiping Insurance | 21.36 | WATCH | 21.36 | 21.08 | 21.37 (+0.0%) | 20.87 (-2.3%) | 20.42 (-4.4%) | China Taiping screens as the one genuinely cheap name in this book, but we are not buying it. The buy signal that had it ranked first for five sessions withdrew today as the shares fell another 2.3%, the company fails our fundamental quality screen, and the model's own twelve-week view now sits 4.4% BELOW the current price - a cheap valuation attached to a deteriorating trend is not an entry. We keep it on the watch list. Operational: label went BUY cv60 -> WATCH cv46, ending a five-session run of quality-gate blocks; adj conviction 46 + val_adj 5 = 51, now well below the 65 floor in any case. Even had the label survived: quality gate LIVE and FAILING (UNDERVALUED is a non-NA_* row with Quality_Pass False, F-Score 5) = hard block on a new entry; HF-3 inversion Est_12W 20.42 vs 21.36 close = -4.40%, the deepest in the scan for a second session. Fresh valuation refetched 2026-09-08 (MoS_FV +0.785 -> +10 halved to +5 for the quality failure). Re-arm: a bullish label PLUS Quality_Pass True or a non-inverted Est_12W. · news |
| 0005 | HSBC HDGS(USD0.50) HK O'SEAS | 164.90 | BUY-STOP | 164.90 | 164.16 | 165.04 (+0.1%) | 163.23 (-1.0%) | 163.56 (-0.8%) | HSBC lost its setup entirely today - the model's conviction in it fell by 35 points to a level that no longer supports any kind of entry, and its twelve-week view has slipped below the current price. We stay out, and yesterday's decision to cancel the standing buy order rather than chase it lower now looks right: the shares traded down to within fifteen cents of that cancelled limit today, which would have filled us straight into the conviction collapse. Operational: BUY - ANTICIPATE cv81 -> WATCH cv46 (-35). The anticipate print gate has now FAILED as well (ANTICIPATE in 1 of the last 3 snapshots, needs 2) and cv 46 is below both the 55 anticipate floor and the 65 acquire floor, so the slot is empty on availability, not on a gate. HF-3 inversion widened: Est_12W 163.56 vs 164.90 close = -0.81% (was -0.25% on 09-07). val_adj 0 (NA_CCY row despite MoS -10.9 - currency-mismatched, no force). Note for the counterfactual grader: the 09-07 withdrawal of the 164.75 gtc_limit records as a SAVE against the queue_expiry class's standing [cost] reading. Re-arm: a fresh BUY/ANTICIPATE label at cv>=55 plus a non-inverted Est_12W. · news |
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 →
5.000 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.