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 |
|---|---|---|---|---|---|---|---|---|---|
| 0992 | Lenovo Group Limited | 32.40 | BUY — QUEUED @31.56 | 31.56 | 29.71 | 32.41 (+2.7%) | 33.39 (+5.8%) | 35.44 (+12.3%) | Lenovo is the best buy setup this book has produced in weeks: a STRONG BUY on PullbackBuy_VWAP, the only buy signal family with a positive track record here, backed by the second-highest quality score in the portfolio (F-Score 7, fair-value rated, DCF 45.64 against a 31.56 entry). The shares jumped 6.1% today on a genuine cluster of AI product news - the Qira assistant rolling out across devices, a Frore AirJet ultra-slim notebook, and a Workato partnership - so the story is real, but paying up after a one-day pop is not. The order is therefore a limit 2.6% below the close, at the volume-weighted average price the pullback signal is built around: it fills only if the pop gives some back, and costs nothing if it does not. Twelve-week estimate is +12.3% with the DCF anchor well above it, so valuation and projection agree. Operational: adj 85 = cv86 val -3 bias +2, twenty points clear of the 65 floor. This is the 09-03 withdrawal's own re-arm firing exactly as written (fresh BUY print at conv >= 65) with the signal family upgraded from FailedBreakdown chase (-2) to PullbackBuy (+2), adj. Order_Type Buy Limit @31.56 from the live row, Entry_Distance_ATR -0.51 past the -0.25 no-chase line; the withdrawn card's @30.60 limit is NOT carried (provenance). Entry range 31.15-31.97. Size held at the 09-01/09-02 card level of (board lot 2,000) = HKdeliberately NOT escalated despite the stronger case - HK-11 escalates urgency, never size; model 0.7155 x 5,551,203 x 0.20 = HKwould consume all. Risk at stop HKL7 CROSS-BOOK: family holds (hktech 10,000, hksemicon 14,000) -> 30,000 on fill; hk is the governing parent, execute ONCE. Horizon: the 09-07 execution session carries a high-confidence two-sided NFP reaction, a further argument for the limit form. Auto-withdraw if conviction < 65 or on expiry 2026-09-09. · news |
| 0700 | TENCENT | 442.80 | SELL 85.46% | 442.80 | 430.00 | 438.78 (-0.9%) | 438.91 (-0.9%) | 442.08 (-0.2%) | Tencent has printed a sell signal on eight of the last nine sessions and today's rally did not change that. The stock rose 2.26% but the whole Hong Kong market rose 2.1%, so this was index drift rather than the company being re-rated - and it still closed below its own session average price and below both its 20-day and 200-day trend lines, the pattern of sellers meeting strength. The model sees no return over any horizon from here: one week -0.9%, four weeks -0.9%, twelve weeks -0.2%. We sell half the holding at 442.80, a better exit than the 433.00 we asked for yesterday, and keep the other half long the AI story, which is genuinely strong - a new model release, heavy demand for the Enflame chip listing, and an active HK buyback. This is a technical trim on one of the highest-quality businesses in the book, which is exactly why it is a half-trim and not an exit. Operational: adj 76 = cv76 + bias 0 (FailedBreakout_Down_20D n=65 hit_1w.508) + val 0 (MoS -0.362, not <= -3, no tier bump; sell side amplify-only per CLAUDE.md s9). Clears the 65 floor unaided. *** HK-11 MANDATORY ESCALATION: 3rd consecutive unexecuted session (09-02 @438.20, 09-03 @433.00) - issued ONCE as market-at-open and RETIRED thereafter. If unexecuted at the 09-07 open it is NOT re-proposed; re-arms only on a close below 430.00. *** Non-execution has HELPED here: 433.00 -> 442.80 is +HKon vs yesterday's card. Action price is the live Entry_Price (Order_Type Market, EDist 0.00), repriced up from 433.00; stale limits not carried. Entry range 439.86-445.74. Volume: 85.46% x 200 = 170.; nearest lot 200 would be a full exit a REDUCE cannot authorize, so rounded DOWN to 100 (50%), retaining; proceeds HKNo upward deviation - escalation changes urgency, never size. WARNING Avg_Cost recorded as 1.00 (Unreal_PnL_Pct +44,180%) so realised cannot be computed; backfill needed. Stop on the retained HELD at 430.00 (model prints 419.31, lower - never lowered). · news |
| 0005 | HSBC HDGS(USD0.50) HK O'SEAS | 166.70 | HOLD | 165.42 | 160.17 | 167.62 (+1.3%) | 163.68 (-1.1%) | 164.69 (-0.4%) | HSBC printed the strongest buy signal in the book today and we are not adding to it. The position is already the largest holding at HK, 36% of the portfolio, and the signal is simply the Hong Kong market rising 2.1% showing up as a breakout - the stock's +1.96% actually lagged the index. More tellingly, the model that rates this a maximum-conviction buy simultaneously forecasts a lower price in four weeks (-1.8%) and twelve weeks (-1.2%). We raise the protective stop to 160.17 instead and keep the full holding. Operational: adj 97 = cv100 + bias -3 (BreakoutUp_20D|buy n=13 hit_1w.364) + val 0 (NA_CCY, gate off). Highest adj in the book and declined anyway on HK-1 - BreakoutUp_20D is one of the three momentum-chase families and the rule is categorical, never up-size; same add was declined at cv86 on 09-03. Entry 165.42 is a buy-stop BELOW the 166.70 close (EDist -0.49), already triggered and extended. Exit_Warning PROFIT_ARMED is VOID under the standing ruling - 0005 carries no Avg_Cost at all, and the arm is honoured only via Gain_ATR >= 3.0 on a real. ADR-0012 armed-but-not-fired response: TRAIL RAISED 159.95 -> 160.17 (model Stop_Price above the recorded line). |
| 0939 | CCB - H SHS | 9.76 | HOLD | 9.83 | 9.35 | 9.73 (-1.0%) | 9.99 (+1.6%) | 9.90 (+0.7%) | CCB flipped to a strong buy signal today, but this is the worst-performing signal type in the entire Hong Kong book - every single past instance was stopped out for a loss. The stock is also technically stretched (RSI 71.8) and the suggested order would mean paying above today's close to chase it, which is exactly how the previous losses happened. The position is already HK, the second-largest equity line, and the model only sees +1.4% over twelve weeks. We hold what we have and raise the stop to 9.35. Operational: adj 80 = cv90 + bias -10 (MomentumContinuation_Up|buy n=5, hit_1w.000, stop-first 1.000 - the worst cell in the book) + val 0 (NA_CCY, gate off). Clears the floor numerically and is declined on HK-1, which names this family the worst of the three chases and forbids up-sizing. Entry 9.83 is a buy-STOP above the 9.760 close (EDist +0.29) into an RSI already past the ~70 chase threshold HK-1's stop-out evidence set was built on. Exit_Warning PROFIT_ARMED VOID (no Avg_Cost). ADR-0012 armed-but-not-fired: TRAIL RAISED 9.23 -> 9.35. |
| 2888 | STANCHART | 237.40 | HOLD | 240.44 | 229.90 | 236.28 (-1.7%) | 231.78 (-3.6%) | 230.46 (-4.2%) | Standard Chartered broke to a 52-week high and on the technicals this was the most defensible addition on the board - the stock is not overbought and the order would only fill on further strength. We are still not buying, because the model expects the price to be lower at every horizon we care about: -0.5% in a week, -2.4% in a month, -2.9% in twelve weeks. Paying a 1.3% premium to enter a position the model thinks is heading down is not a return-maximising trade. We raise the stop to 229.90 and keep the existing. Operational: adj 90 = cv90 + bias 0 (BreakoutUp_52W|buy n=4 report-only, hit_1w.75 / hit_4w.50 - encouraging but below the sample floor, mentioned not applied) + val 0 (NA_CCY, gate off). NOT one of the three HK-1 chase families and RSI 57.71 is unextended; declined purely on the negative forward pair per section B's falling-knife rule. Entry 240.44 is a buy-stop 1.3% above market (EDist +0.58). Holding is = HKan odd lot against a 50-sh board lot. Exit_Warning PROFIT_ARMED VOID (no Avg_Cost). TRAIL RAISED 229.36 -> 229.90. Re-arm: a fresh ADD print with Est_4W turning positive. |
| 0981 | SMIC | 67.20 | HOLD | 67.20 | 69.63⚠ | 67.80 (+0.9%) | 69.54 (+3.5%) | 68.38 (+1.8%) | SMIC has now closed below its protective stop for four sessions running and the breach has widened to 3.5%, the deepest of the run, while the signal escalated to a strong sell. We are still not trimming, because the model's twelve-week view remains positive (+1.8%) and the sell signal's own confidence score is too low to act on - our exit rule requires both a stop break and a negative forward view, and only one of the two conditions is met. This is honest but uncomfortable: a risk line the position keeps ignoring is a problem the current rules cannot terminate, so we add an explicit backstop below. Operational: adj 58 = cv55 + bias +3 (RallySell_BearTrend|sell n=23 hit_1w.650, the book's best exit cell) + val 0 (sell side amplify-only). SEVEN points below the 65 floor, so it fails independently of L2. L2 leg 2 fails for the FOURTH consecutive session (Est_12W +1.8%, Est_4W +3.5%, Est_1W +0.9%). Pre-committed 09-03 re-arm honoured as written, not loosened. Position HKStop HELD at 69.63 (model prints 61.35, -8.7% - never lowered; HK-12 uselessly wide). *** NEW INTERIM PRE-COMMITMENT: a FIFTH consecutive session below 69.63 with the STRONG SELL label intact ALONE originates a mandatory 30% trim, regardless of the 12wk sign - see systemic finding 3, the hk analogue of ADR-0020 that HK-12 declined to import. *** Existing re-arm also stands: a close below 69.63 with Est_12W turning negative. · news |
| 1347 | Hua Hong Semiconductor | 109.70 | HOLD | 109.70 | 105.00 | 116.42 (+6.1%) | 122.78 (+11.9%) | 127.55 (+16.3%) | Yesterday we declined to trim Hua Hong because it was joining the Hang Seng Index today and index inclusion normally brings forced buying from tracker funds. The opposite happened: the stock fell 8.4% on the inclusion day itself, costing HKThe buying had evidently already been done in advance by active investors, leaving only sellers on the day. That call was wrong and is recorded as such. Today the signal is no longer a sell (it reads AVOID at low confidence) so there is nothing to act on, and the model now sees the drop as overdone, projecting +6.1% in a week and +16.3% over twelve weeks. The next scheduled check is Monday, the post-inclusion unwind session. Operational: adj 42, twenty-three points below the 65 floor - not actionable regardless. The 09-03 decline had three stated grounds and only one (the catalyst) failed: the exit had ALREADY been retired under HK-11 on independent grounds, so the catalyst rule was not the sole thing holding the position. Horizon rated it up[+2.0,+9.0] HIGH confidence hold-through - wrong in sign; see systemic finding 1 on index-inclusion stance defaults. Pre-committed re-arm is the 2026-09-07 EOD (post-inclusion unwind) or earlier on a close below 105.00; horizon independently flags 09-07 as down/tighten-trail for this name - that check MUST be made. Stop HELD at 105.00 (model prints 126.94, +15.7% ABOVE the close - inverted, HK-12, not adopted). Quote verified 109.700/-8.43%/prev 119.800; no dated news on allowed domains explains the drop, consistent with a pure flow unwind. · news |
Outcome & track record
Accuracy at the time of this record.
198 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 →
-0.261 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.