Hong Kong — End-of-Day Verdict
3 ACQUIRE (0700 ADD 100 buy-stop @498.63 QUEUED; 0992 BUY 2000 @24.74; 3690 QUEUED 900 @85.45) + 3 DISPOSE (0005 SELL 6000/50%; 0981 SELL 1000/40%; 2888 SELL 150/43%). 0939 + 0823 profit-take arms declined as degenerate, ledgered optional, stops raised. 2513 WATCH <862 unchanged; 0522 advisory void; 7709 no card after -16.7%. No stop breaches.
Recommendations
2 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 |
|---|---|---|---|---|---|---|---|---|---|
| 0005 | HSBC HDGS | 168.20 | SELL 50% | 168.20 | 163.17 | 167.26 (-0.6%) | 166.67 (-0.9%) | 169.60 (+0.8%) | HSBC has run vertically into its H1 results and the run is stalling: the stock made a lower high (168.90 -> 168.70) on an unchanged close with RSI pinned at 80, and the model's conviction halved from in one session while its 12-week view (169.60) slipped below the current price. Banking half of a 42%-of-book position ahead of a binary earnings print is risk reduction, not a call on the results. ADR-0012 profit-take: armed (PROFIT_ARMED, T12_Progress 99.17) + 2 fire ticks - (b) conviction = -45 pts, (c) dead forward view T12 169.60 < 1.02x close 171.56 -> profit_take_2 = 50%. Catalyst-aware hold (HSBC H1 on 08-04, 1 session out) is satisfied via its own confirmed-lower-high exception. G1 inapplicable (cv 45 < 80). Valuation NA_CCY -> blind, no amplify, size x1.0. 6000/400 = exact. Remaining 6000 trails at 163.17 (raised from 158.65). Caveat: results may print before the 08-04 HK open - reprice per L8 if it gaps >1.5%; the sold half also forgoes the 08-13 ex-div (~HK, immaterial vs a +/-8% gap). · news |
| 0981 | SMIC | 62.20 | SELL 40% | 62.20 | 58.08 | 64.17 (+3.2%) | 68.05 (+9.4%) | 69.65 (+12.0%) | SMIC is in a confirmed downtrend - trading at 62.20, below both its 50-day (73.04) and 200-day (68.62) averages - and the model has printed STRONG SELL - REDUCE for five consecutive snapshots at the book's number-one sell rank. This exit family is the most reliable one-week signal this book has (three quarters of past calls worked), and the stock reports Q2 in three sessions, so taking size off ahead of a binary print on a bearish tape is the disciplined move. Adjusted 74 = 69 raw + 5 (RallySell_BearTrend|sell, n=14, hit_1w.75). CORRECTION to the 07-31 entry, which called this the '-5 cell (hit_1w.25, hit_4w.00)' and therefore held at adjusted 61 below the floor - the live scorecard and learnings section C both have it at +5; the 07-31 run inverted the sign, borrowing hit_1w.25 from BreakoutUp_20D|buy and hit_4w.00 from BreakdownDown_20D|sell. Engine portion is 85.46%; HK-9 caps a signal-only sell at a tactical 1-week de-risk, so 50% intent -> -> 500-lot rounding -> 1000 (40%). Valuation amplify available (MoS -4.39 <= -3 would bump a tier) but DECLINED per HK-9. Catalyst-aware hold does not block: the 'valid exits still fire' contrast clause covers a stable STRONG SELL - REDUCE. Est_* contradict the signal at every horizon (+3.2/+9.4/+12.0%) - systemic finding #2 recurring; trading the 14 graded outcomes over the estimator and sizing down to 40% because of the disagreement. Remaining 1500 stop 58.08 (raised from 57.00). · news |
| 2888 | STANCHART | 236.80 | SELL 43% | 236.80 | 222.09 | 235.43 (-0.6%) | 234.40 (-1.0%) | 231.24 (-2.3%) | StanChart is a genuine winner running out of road: 25.6% above its 200-day average with RSI 68, it just made a lower high (240.00 -> 237.20) on a lower close while the model's conviction dropped from an actionable BUY-ADD 73 to HOLD 52, and every forward estimate now sits below today's price. Unusually for this book the estimator agrees with the sell at all three horizons, so banking roughly half of an outsized gain is well supported. ADR-0012 profit-take: armed (PROFIT_ARMED, T12_Progress 102.4) + 2 fire ticks - (b) conviction -21 pts, (c) T12 231.24 < 1.02x close 241.54 -> profit_take_2 = 50% = -> 50-lot rounding -> 150 (43%). G1 does not protect (signal is FailedBreakdown_Up_20D [WATCH] cv52, not a fresh breakout >=80); G2/G3 do not block a fired profit-take. HSBC H1 on 08-04 is an ecosystem read-through catalyst for 2888 - the confirmed-lower-high exception clears it. Valuation NA_CCY -> blind, no amplify. Remaining 198 keeps stop 222.09 (printed 221.76 would LOWER the line - refused per trail discipline). · news |
| 0700 | TENCENT | 490.40 | WATCH — queued order lapsed | — | — | n/a | n/a | n/a | Queued buy @498.63 lapsed — closed 490.40 below the level; awaiting reprice at the next end-of-day run. Tencent broke out on a +3.2% session and the model escalated to STRONG BUY - ADD at the book's number-one buy rank, with RSI still a moderate 61.5 - room to run rather than an exhausted move. We add only on continued strength via a buy-stop above today's close, so the order fills only if the breakout is real, and we keep it deliberately small because the stock is still trading below its 200-day average and reports Q2 on 12 August. Adjusted 78 = 86 raw - 3 val (GROWTH_PRICED_IN, MoS -0.5502) - 5 bias (BreakoutUp_20D|buy, n=9, hit_1w.25). Model order type IS a Buy Stop @ 498.63, eD +0.51 (1.7% above close) -> no-chase satisfied and it is the fill-on-strength instrument HK-4/HK-10 prescribe. HK-1 confirmation for a chase family met: two consecutive buy prints (15:30 BUY-ADD cv66 -> 17:00 STRONG BUY-ADD cv86) with RSI < 70. The standing overlay bearish override on 0700 is invalidated BY ITS OWN TERMS ('invalidated when signal turns up') - and the overlay is 29 days stale regardless. Sizing 70.29% x HK x 20% = HKx1.0 -> -> 100-lot -> = HK (200 would be 18% over budget). WARNING: Est_1W/4W (486.73/489.73) sit BELOW the 498.63 fill - the estimator does not support this entry near-term; it only fills if the market disagrees, hence the small size. 12wk 509.41 exceeds the DCF anchor 481.75 - projection assumes growth beyond priced-in. Event Horizon flags Tencent Q2 (12 Aug) as the window's widest ATR-normalised band (+/-7% = +/-2.03 ATR) on a top-weight name - queue expires 08-06, well before, but the combined 300-sh position must be re-assessed before 08-12. · news |
| 0992 | Lenovo Group | 24.74 | BUY | 24.74 | 23.56 | 24.94 (+0.8%) | 25.53 (+3.2%) | 26.60 (+7.5%) | Lenovo is the only fresh buy in the book that clears the quality screen (F-Score 7, passes the FCF/quality gate) and it is buyable at the model's own entry rather than chased - the price is exactly at the trigger with RSI a calm 57, four consecutive BUY prints today, and the stock well above both its 50-day and 200-day averages. Independent valuation work puts it near fair value, and buying at the 04 Aug open also captures the HKdividend going ex on 05 Aug. Adjusted 67 = 73 raw - 2 val (FAIR_VALUE, MoS -0.3925) - 4 bias (FailedBreakdown_Up_20D|buy, n=10, hit_1w.30). Quality gate PASSES (Quality_Pass True, F=7) - the only new-entry candidate that does; 9988 is blocked at F=3. eD 0.00 exactly at the model entry; HK-1's confirmation demand for this chase family is met by 4 consecutive prints. CAUTION (news-sourced, real): +27% over 7 days, +114.8% over 30 days, +161.8% YTD - parabolic. De-sizing is delivered by the 2000-share board lot, which already rounds HKdown to HK (a 42% cut); do NOT round up to 4000. Do not misread the ~0.34 markdown on 05 Aug as a signal break - it is the ex-dividend. Note the 4wk band [15.62-42.20] is extremely wide at c80 - low confidence in magnitude, direction only. · news |
| 3690 | MEITUAN-W | 93.45 | BUY — QUEUED @85.45 | 85.45 | 81.19 | 93.08 (+8.9%) | 92.46 (+8.2%) | 85.05 (-0.5%) | Meituan prints the book's single most reliable buy pattern - a pullback-to-VWAP setup that has worked roughly seven times in ten at the one-week horizon - but the stock has already run 8.6% past the level the model wants to pay. Rather than chase it we leave a good-till-cancelled limit at the model's own entry: it fills only on a genuine pullback, and costs nothing if it does not. Adjusted 77 = 73 raw + 0 val (Val_Signal NA -> gate off) + 4 bias (PullbackBuy_VWAP|buy, n=8, hit_1w.714 - HK-2's positive anchor). Model order type IS a Buy Limit @ 85.45; eD -2.12 makes a market buy a hard chase, so the GTC limit is the no-chase instrument. HONEST CAVEATS: (1) the limit is 8.6% below the close and expires in 3 sessions - it will very likely expire unfilled; (2) T12 85.05 < close 93.45 (-9.0%), a negative 12-week view - but FROM A FILL at 85.45 the estimates are +8.9% (1wk) / +8.2% (4wk) / -0.5% (12wk), a profile that matches the cell exactly (hit_1w.714, hit_4w.25), so treat this strictly as a 1-week instrument per HK-9, not a 12-week thesis; (3) a fill at 85.45 would be BELOW the 200-day (91.23); (4) this is the FOURTH consecutive queue on 3690 (07-22 @77.22, 07-27 @80.33, 07-28 @81.64 all expired unfilled) - the expiry mechanism is working as designed but four no-fills is logged as a systemic finding. Sizing 70.59% x HK x 20% = HK -> -> 100-lot -> 900 = HK. · news |
| 0939 | CCB - H SHS | 9.20 | NO ACTION | 9.20 | 8.78 | 9.17 (-0.3%) | 9.18 (-0.2%) | 9.25 (+0.5%) | Not trimming CCB. A mechanical reading of the profit-take rule would cut half the position, but the trigger is an artefact rather than a real profit warning - the 'armed' flag fires because the 12-week target has drifted toward the price, not because the position has a verified gain, and a quiet no-signal state is simply how a large mainland bank normally behaves. Raising the protective stop instead. ADR-0012 mechanical read: armed (T12_Progress 99.46) + 2 ticks - (c) T12 9.25 < 1.02x9.20 = 9.384, (d) 5 consecutive NO TRADE prints -> would be a 50% trim = = HKthe second-largest banked-dollar candidate. DECLINED: the arm derives entirely from the degenerate T12_Progress >= 90 limb (systemic finding #-31 - PROFIT_ARMED prints on held names including 7709 at -69%); Avg_Cost is blank so the Gain_ATR limb is unverifiable; and both ticks are structurally weak on a low-beta bank (NoTrade/'consensus conflict' is 0939's normal state, and a 1.5% 12wk shortfall is noise). The us book gated this as US-26 (require Gain_ATR > 0 AND Unreal_PnL_Pct > 0); hk has NOT adopted it and lessons are not auto-shared - carried as a finding for the next hk learnings review. Ledgered optional so the counterfactual grader prices what not trimming cost. Stop raised 8.60 -> 8.78 (2.0x ATR chandelier; printed Stop_Price is blank, HK-12). |
| 0823 | LINK REIT | 39.14 | NO ACTION | 39.14 | 37.84 | 39.14 (+0.0%) | 39.40 (+0.7%) | 39.44 (+0.8%) | Not trimming Link REIT. The profit-take flag fires on the same artefact as CCB, and this position is barely 2.5% above its 200-day average - there is no meaningful profit here to take. Raising the stop instead. ADR-0012 mechanical read: armed (T12_Progress 99.24) + 1 tick - (c) T12 39.44 < 1.02x39.14 = 39.92 -> would be a 30% trim = = HKDECLINED on two grounds: same degenerate T12_Progress arm, and it ranks 4th on banked dollars behind 2888 so the <=3 DISPOSE cap displaces it regardless. Conviction ROSE this session (NO TRADE 0 -> HOLD 66), the opposite of deterioration. Ledgered optional for the counterfactual grader. Stop raised 37.70 -> 37.84 (printed model stop = 2.0x ATR chandelier). |
| 2513 | KNOWLEDGE ATLAS | 940.00 | HOLD | 940.00 | 862.00 | 986.07 (+4.9%) | 961.99 (+2.3%) | 1029.53 (+9.5%) | Holding the watch line on Knowledge Atlas. The stock fell another 4.8% to 940 but has not reached the level that would reopen the exit case, and this name was already re-proposed for sale on eight consecutive sessions without ever being acted on - re-issuing a ninth card would repeat the exact failure the book learned from. The signal has in any case decayed from a strong sell to a neutral AVOID. HK-11 governs: the card was escalated once (07-28) and retired to a WATCH line; it is NOT re-issued. Re-arm trigger is a SETTLED CLOSE below 862.00 - today's 940.00 is 9.1% above it, so not triggered. Signal decayed STRONG SELL cv77 -> AVOID cv63; per HK-9 that decay is NOT a re-entry case either. Position = HK, -42.5% vs cost 1635. Printed Stop_Price 1354.99 is ABOVE the close - HK-12 degenerate, rejected. Hang Seng quarterly index review 14 Aug is the one dated catalyst on this name. |
| 7709 | XL2CSOPHYNIX | 35.44 | HOLD | 35.44 | 33.00 | 37.52 (+5.9%) | 35.70 (+0.7%) | 36.77 (+3.8%) | No new card on the leveraged Hynix tracker despite a brutal -16.7% session. It carries no actionable sell label, has not broken its risk line, and the standing rule for 2x leveraged ETFs is to trim rather than dump - a full exit here would repeat the May 7747 error. Session -16.65% (42.52 -> 35.44), the worst held name; position -69.06% vs cost. AVOID cv59 is not an actionable sell label; L2 not met (no stop break, T12 36.77 marginally positive at +3.8%). HK-11 retired this name's card after it was mass-declined 07-22..07-28. Hard risk cap RETAINED at 33.00 - stated explicitly as a cap, not an ATR-derived stop: ATR 21.07 on a 35.44 close is 59% of price, so any ATR-faithful line is unusable, and the printed Stop_Price 77.57 is ABOVE the close (HK-12 degenerate). |
| 0522 | ASMPT | 150.30 | HOLD | 150.30 | 133.00 | 157.58 (+4.8%) | 166.91 (+11.1%) | 168.48 (+12.1%) | The 07-31 advisory half-sale on ASMPT is now void rather than overdue - it was a retroactive backfill card that could never be executed, and the sell signal behind it has since decayed to a neutral AVOID across four flat prints today. Not re-issued. Prior card: SELL @151.30 (advisory, backfill, not_executed); today STRONG SELL cv76 -> AVOID cv61. HK-11: not re-issued. HK-9: the decay is not a re-entry case either. Stop 133.00 holds (close 150.30, +13.0% above). Position -23.45% vs cost 196.35; T12 168.48 = +12.1% positive. Applied Materials FQ3 on 13 Aug is the read-through catalyst. |
| 1347 | Hua Hong Semiconductor | 126.80 | HOLD | 126.80 | 118.00 | 132.91 (+4.8%) | 136.20 (+7.4%) | 132.83 (+4.8%) | Holding Hua Hong through its Q2 results on 06 August. The signal is a neutral AVOID rather than a sell, the stop is intact, and every forward estimate is positive - there is no exit case to act on, and pre-trimming into a binary print without one is not supported. AVOID cv44 is not an actionable sell label; close 126.80 is 7.5% above the 118.00 stop; T12 132.83 = +4.8% positive so L2 is not met. PROFIT_ARMED prints on this row at -20.75% vs cost with Gain_ATR -2.19 - a textbook degenerate arm, no profit-take originated. BreakdownDown_20D|sell carries bias 0 (n=15, hit_4w.00 - decayed out of its +3 per HK-8). Catalyst: Hua Hong Q2 on 06 Aug, plus TSMC July revenue 10-13 Aug and AMAT FQ3 13 Aug as read-throughs. |
| 3750 | CATL | 625.00 | HOLD | 625.00 | 580.00 | 621.03 (-0.6%) | 609.11 (-2.5%) | 589.24 (-5.7%) | Holding CATL. No signal in either direction across five consecutive snapshots and the stop is well clear, so there is nothing to act on. NO TRADE cv0 x5 ('Consensus conflict + low conviction'); close 625.00 vs stop 580.00 (+7.8%). PROFIT_ARMED prints at -16.87% vs cost with Gain_ATR -5.11 - degenerate arm, ignored. Forward view is mildly negative (T12 589.24 = -5.7%) but with no sell label and no stop break neither L2 nor a profit-take fires. Macro-driven: China services PMI 05 Aug, trade data 07 Aug, CPI/PPI 10 Aug, activity data 17 Aug. |
Outcome & track record
Accuracy at the time of this record.
132 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 →
-0.283 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.
49 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.