Recommendations
1 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 |
|---|---|---|---|---|---|---|---|---|---|
| 2382 | Sunny Optical Technology (Group) Company Limited | 69.10 | BUY — QUEUED @66.31 | 66.31 | 63.24 | 68.93 (+4.0%) | 69.00 (+4.1%) | 67.82 (+2.3%) | Opening a new position in this Chinese optical-components maker on a pullback setup, but only at a lower price. The shares have turned up off a strong trend and print the book's highest-ranked buy signal, yet they have already run 4.2% past the level the model wants to pay - so we bid 66.31 rather than chase 69.10. The company earns its place: revenue grew 12.9% last year, it passes our quality screen with one of the strongest financial-health scores in the portfolio, it is buying back its own stock, and at our bid price it trades at a modest discount to what we think it is worth. The order sits for three sessions and then lapses. OPS: PullbackBuy_VWAP = the book's only positive actionable buy cell (bias +2, n=14) and the L1/HK-2 favoured family. Entry_Distance_ATR -1.03 -> queue, no chase. Quality gate PASSED (F_Score 7, Quality_Rank 4/31). val_adj 0 (MoS -0.026, at fair value). SIZE NOTE: model HK =, BELOW the - rounded UP to one minimum lot = HK (+54% vs model; 22.1% of the 600k config base, 2.5% of the real HK book). There is no smaller tradeable size; drop the card rather than shrink it if the oversize is unacceptable. CAUTION: Est_12W 67.82 is +2.3% vs the 66.31 limit but -1.85% vs the 69.10 close - the forward case exists ONLY if the pullback fills. Re-price or withdraw at each EOD per the provenance rule; auto-withdraw if cv<65 or after 09-17. · news |
| 3750 | CATL (Contemporary Amperex Technology) | 547.50 | SELL 85.46% | 547.50 | 513.60 | 547.21 (-0.1%) | 548.64 (+0.2%) | 543.93 (-0.7%) | Cutting two-thirds of our holding in this EV-battery maker because the price finally broke the line we drew six sessions ago. We said we would sell only on a close below 556.49 and held through five sessions of noise - including the strongest sell signal the name had ever printed, which we deliberately ignored and which then faded. Today it closed at 547.50, and the rule fires. The stock is 7.6% below its 200-day average with momentum deeply oversold, the model sees no upside at any horizon from one week to twelve, and the shares trade about 13% below the price institutions paid in the April share sale, which is real overhead supply. We keep. OPS: the 556.49 re-arm pre-committed 09-03 FIRES (close -1.62%, 0.53 ATR - decisive). ALSO L2: decisive stop break + negative 12wk (Est_12W 543.93 = -0.65% vs close). ADJ CONVICTION 59 IS BELOW THE 65 FLOOR - this card is carried by the pre-commitment + L2, NOT by conviction; stated explicitly rather than papered over. Bias +4 (RallySell_BearTrend|sell, n=29, hit_1w.696 - the book's best exit cell). Valuation amplify WITHHELD under G2: EPV 302.80 vs DCF 724.51 on a 547.50 price = 2.4x spread, unreliable anchor. PROFIT_ARMED is an ARTIFACT (position -27.17%, Gain_ATR -12.05) - no ADR-0012 derivation. Residual stop REBASED to 513.60 (2.0 ATR) now that 556.49 is broken and acted on. HK-11: session 1 of a new cycle - escalate once if unexecuted, then retire to WATCH 513.60. · news |
| 0823 | Link Real Estate Investment Trust | 37.84 | SELL 49.87% | 37.84 | 36.74 | 37.83 (-0.0%) | 37.78 (-0.2%) | 37.88 (+0.1%) | Halving our position in this Hong Kong shopping-mall and car-park REIT after it broke down through the level we had flagged as our line in the sand. We held this name through five sessions of marginal weakness because the breaks were trivial; today's 2.6% fall is not - it closed 2.2% below our stop, dropped under its 200-day average, and flipped to an outright sell signal for the first time. The business itself is shrinking, with revenue down 2% year on year, and the next fortnight puts US inflation data and a Federal Reserve decision - which Hong Kong's currency peg forces the city to follow the same morning - directly on a rate-sensitive landlord. The model sees the price flat at every horizon, so we give up nothing by cutting. We keep half. OPS: the 09-10 hard line ('a close below ~38.40 arms L2 on the spot') is BREACHED at 37.84. Fresh SELL - REDUCE cv64, Sell_Rank #3; bias +1 (BreakdownDown_20D|sell, n=19 - refreshed live per HK-7, was 0 in the D-bis block). Adj 65 clears the floor exactly. VALUATION AMPLIFY SUPPRESSED: the row prints UNDERVALUED MoS +9.69 while FV_per_share 37.8437 EQUALS Price 37.84 to 4dp, with a negative EPV of -35.71 and Quality_Pass False - internally contradictory, degenerate, so no down-tier (it would have cut the trim from 1,400 to). Avg_Cost blank -> PROFIT_ARMED invalid, no ADR-0012. Residual stop REBASED to 36.74 (1.8 ATR). HK-11: session 1 of a new cycle. · news |
| 0700 | Tencent Holdings | 428.40 | SELL 49.87% | 428.40 | 415.64 | 424.40 (-0.9%) | 424.61 (-0.9%) | 418.59 (-2.3%) | Trimming half of a small Tencent position - the second and final time we will propose this sale. The shares remain below the 430 level we set as our exit trigger, have now printed a sell signal for four straight sessions, and sit 13% below their 200-day average, just 4% off a 52-week low. The model expects the price lower at one week, four weeks and twelve - the only name we are selling today where the forecast actively agrees with the decision rather than being sold into. Against that, Tencent's AI-chip affiliate Enflame listed in Shanghai today and the stock rose slightly, which is a genuine positive but not enough to override a trigger that has already fired. The position is tiny either way; we keep, and we note the shares are not expensive - our own discounted-cash-flow estimate sits 19% above today's price, which is why this is the minimum possible trim and not more. OPS: HK-11 ESCALATION, session - the 09-10 card was not filled at the 09-11 open. BIAS IS 0, NOT the +4 carried yesterday: the signal CHANGED FAMILY to BreakdownDown_52W, whose cell is n=3 report-only (below the sample floor) - no credit, no haircut (HK-7). Card clears the 65 floor on raw conviction alone, with zero margin. Valuation argues trim-the-minimum (DCF 508.07 = +18.6% vs spot) and we are already at the 100-sh board-lot minimum. Avg_Cost placeholder 1.00 (Unreal_PnL 42,740%) -> PROFIT_ARMED invalid, no ADR-0012. Stop RAISED 408.41 -> 415.64. Stale 68-day overlay's standing 'close below 430 -> exit' is still numerically true and carried as CORROBORATION ONLY. RETIREMENT: if unexecuted at the 09-14 open, 0700 is RETIRED to WATCH 415.64 and NOT proposed again; re-arms only on a close below 415.64 OR a fresh REDUCE at cv>=65 following a rally back above 443.06. · news |
| 2513 | Z.AI CO., LTD (KNOWLEDGE ATLAS) | 793.00 | HOLD | 793.00 | 682.51 | 851.78 (+7.4%) | 911.75 (+15.0%) | 898.54 (+13.3%) | Holding, under protest. We proposed selling most of this AI-model developer on three consecutive sessions and none of those sales was executed; under our own rules the idea is now retired and will not be raised a fourth time. Not acting cost about HKtoday alone as the shares fell another 3.2%. The position is down 51% and is still the second-largest thing we own. It becomes actionable again only if the price falls below 682.51 or a strong sell signal returns after a rally. OPS: HK-11 RETIREMENT TAKES EFFECT. Neither re-arm branch met - close 793.00 is 16.2% ABOVE the 682.51 line; SELL - REDUCE prints but cv60 < 65 and there was no rally back above 913.50. Non-execution mark today -HK ( x 26.00), 4th consecutive costly session. Position 15.2% of the book, Avg_Cost 1,635.00, -51.50%, Gain_ATR -9.12. Model Stop_Price 567.67 would LOWER the line 17% - refused (HK-12). NOTE: T12_Progress 88.25 is the ONLY non-degenerate 12wk projection in the book and it is strongly positive (+13.3%) - the retirement is expensive and the tension is recorded, not hidden. · news |
| 1347 | Hua Hong Semiconductor | 111.70 | HOLD | 111.70 | 105.00 | 115.22 (+3.2%) | 121.51 (+8.8%) | 118.76 (+6.3%) | Holding this Chinese chip foundry. We proposed trimming it yesterday, the sale was not executed, and today the sell signal has decayed to a neutral 'avoid' reading with no trade instruction attached - so there is nothing left to escalate. Writing a sell order off a signal the model is no longer giving would be inventing a decision. The idea is retired; it returns only if the price breaks 105.00 or a fresh sell signal appears with real conviction. OPS: HK-11 - the pre-committed second escalation CANNOT be written (Action AVOID, 0.0, decayed from STRONG SELL - REDUCE cv70). Per the 3750/09-04 precedent (escalation expires -> retirement takes effect), 1347 RETIRES NOW to WATCH 105.00. Non-execution mark -HKModel Stop_Price 125.00 prints 11.9% ABOVE the close - INVERTED, rejected (HK-12); manual 105.00 stands. MoS -18.05 computed off EPV 5.86 vs a 111.70 price with a blank DCF - degenerate and moot. Position -30.19%, HK (6.4%). Re-arm: close below 105.00 OR fresh REDUCE at cv>=65. |
| 0981 | SMIC (Semiconductor Manufacturing International) | 63.15 | HOLD | 63.15 | 62.78 | 63.60 (+0.7%) | 64.17 (+1.6%) | 67.55 (+7.0%) | Holding China's largest chip foundry, though it is the strongest sell candidate in the book by rank. We retired this sale two sessions ago after repeated proposals went unexecuted, and the conditions we set for reviving it have not been met - the price is still fractionally above our 62.78 trigger. The margin has narrowed to about half a percent, so a single weak session fires this exit. OPS: HK-11 retirement BINDS for a 3rd session. Sell_Rank #1, STRONG SELL - REDUCE cv68 stable 4/4, adj 72 - FLOOR-CLEARING AND NOT ACTIONED. Neither branch met: close 63.15 is 0.59% above 62.78 (narrowing from 1.07% yesterday); the fresh REDUCE does not follow a rally above 69.63 - the name FELL 65.00 -> 63.15. Model Stop_Price 58.26 would lower - refused (HK-12). Position / HK (3.0%), RSI 37.3, close 8.3% below EMA200. A close below 62.78 fires this immediately. |
| 0005 | HSBC Holdings | 163.30 | HOLD | 163.30 | 163.18 | 163.33 (+0.0%) | 162.47 (-0.5%) | 162.87 (-0.3%) | Holding the portfolio's largest position, but on the narrowest margin in the book - the shares closed twelve cents above the stop-loss level we have set, having drifted lower for a second session. The signal is a neutral hold with fading conviction and the model sees the price flat for the next three months. HSBC's chief financial officer announced her departure this week, which changes nothing about the trade but is one more reason not to extend this name any benefit of the doubt. If it closes below 163.18 we act, mechanically, with no discretion - this is 37.5% of the portfolio going into US inflation data and a Federal Reserve decision that Hong Kong's currency peg forces the city to match the same morning. OPS: stop HELD at 163.18; model Stop_Price 158.23 would LOWER it - refused (HK-12). Clearance 0.07%. / HK = 37.5% of book. HOLD cv55 decayed from 73. PROFIT_ARMED with a BLANK Avg_Cost -> invalid, no ADR-0012. Horizon: 09-14 CPI (HIGH), 09-17 FOMC + HKMA (HIGH), 09-16 retail sales, 09-18 blackout lift. News classification: WATCH (CFO exit; originates nothing). · news |
| 0939 | China Construction Bank - H | 9.75 | HOLD | 9.75 | 9.35 | 9.72 (-0.3%) | 9.88 (+1.3%) | 9.85 (+1.0%) | Holding. This Chinese bank was the highest-conviction name in the portfolio yesterday and the signal engine has gone completely blank on it today, citing conflicting indicators - a sharp deterioration worth naming even though it is not a sell instruction. The price barely moved, the stock sits 15% above its 200-day average and 4.3% above our stop, and it remains the steadiest large holding we have. No action, but we are watching it. OPS: CONVICTION COLLAPSE in one session. NoTrade_Reason verbatim: 'Consensus conflict + low conviction'. Flagged as deteriorating per the conviction-collapse rule; a NO TRADE is not a sell and the stop is 4.3% away. / HK (9.5%), 3rd-largest line. PROFIT_ARMED with a BLANK Avg_Cost -> invalid, no ADR-0012 trim. No model stop printed; manual 9.35 stands. |
| 2888 | Standard Chartered | 241.80 | HOLD | 241.80 | 233.08 | 239.46 (-1.0%) | 239.25 (-1.1%) | 241.07 (-0.3%) | Holding, unchanged on the day. We withdrew a proposed top-up of this bank yesterday when its signal collapsed, and the conditions for reinstating it are not met - the reading has recovered in strength but it is still a 'hold', not a buy, and we will not manufacture a purchase out of that. Raising the protective stop slightly. OPS: re-arm required 'a fresh BUY-family label at cv >= 65'; cv66 clears the NUMBER but HOLD is not a BUY-family label -> not armed. Model stop 233.08 RAISES the recorded 232.95 -> adopted (raises always permitted). / HK (1.6%). PROFIT_ARMED with blank Avg_Cost -> invalid. |
| 0522 | ASMPT | 161.80 | HOLD | 161.80 | 155.02 | n/a | n/a | n/a | Holding this semiconductor-equipment maker. The signal is a weak 'avoid' with no trade instruction, the position is down 17.6% and getting worse, but the price is still 4.4% above the level at which we would act. No trigger, no trade. OPS: WATCH 155.02 held, not re-armed on either branch. AVOID cv49. Model Stop_Price 179.78 prints 11.1% ABOVE the close (161.80) - INVERTED, rejected (HK-12), 3rd consecutive session this row emits an invalid stop. Position / HK (3.4%), Avg_Cost 196.35, -17.60% (from -15.97%). MoS -4.07 on a blank DCF - degenerate. |
| 7709 | XL2CSOPHYNIX (CSOP 2x Long Hynix daily) | 42.88 | HOLD | 42.88 | 34.40 | n/a | n/a | 51.36 (+19.8%) | Holding this twice-leveraged ETF and tightening its safety net. The sell signal that appeared two sessions ago has now faded twice over, which is exactly the whipsaw behaviour that leveraged products produce and the reason we do not exit them on a single weak reading. What has improved is the stop: the model finally produced a sensible protective level, so we raise ours from a near-useless 28.00 to 34.40, meaningfully reducing how far this can fall before we act. OPS: HK-11 retirement + the leveraged-2x rule (full exit needs cv>=70 AND 3 prints AND price < Stop_Price) both block a sell. Signal decayed S.SELL-RED cv64 -> AVOID cv60 -> HOLD cv49 across three sessions - retirement vindicated again. STOP RAISED 28.00 -> 34.40: 19.8% below close = 2.0 ATR on a 9.9%-ATR instrument, versus the inherited line's degenerate 34.7% width. A raise is always permitted (HK-12 forbids only lowering). Position / HK (1.5%), Avg_Cost 114.56, -62.6%. |
Outcome & track record
Accuracy at the time of this record.
213 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 →
-0.241 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.