Hong Kong — End-of-Day Verdict
3 DISPOSE (2513 500 @~1107 MKT, 0981 1000 @70.45 LMT, 7709 900 @~44.80 MKT - frees ~HK) + 2 ACQUIRE QUEUED (0823 900 @37.09, 1810 1600 @24.49 - ~HK, both conditional). 3 stop breaches: 2513/7709 disposed 50%, 1347 held-and-trailed (12wk +5.8% alive -> L2 not met). 2888 + 0388 queues withdrawn. +HK; HK -> ~HK.
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 |
|---|---|---|---|---|---|---|---|---|---|
| 2513 | Knowledge Atlas | 1107.00 | SELL 49.87% | 1107.00 | 1050.00 | 1178.59 (+6.5%) | 1145.18 (+3.4%) | 1215.15 (+9.8%) | Knowledge Atlas collapsed 28.5% in a single session and closed on its low, breaking our stop by 24% and leaving the position 32% underwater one day after entry. The signal has printed a stable, unwavering SELL-REDUCE all session on the book's top sell rank, so we bank half the position rather than hope. We keep the other half because the model still projects a 12-week recovery, and this was a violent one-day repricing rather than a confirmed thesis break. Basis: decisive stop-loss break (1107 vs 1450, -23.66%) + stable SELL-REDUCE x4 cv65; BreakdownDown_20D|sell cell +3 (n10, hit 0.667) -> adj 68. ACTION-PRICE DEVIATION (stated): model prints Sell Limit @1294.89 (+17% above the close) which would never fill -> the L3 anti-pattern; action price = anchor close 1107 at market. Portion 49.87% x 1000 = 500 (clean half; board lot unverified, new listing). No valuation amplify (Val NA). Est_* low-trust: pre-crash anchored, calib_1w -0.261 degenerate. Remaining 500 -> stop 1050 (below the 1085 low). MEMORY NOTE: this was a manual 1.635M entry on 07-16 made against a NO TRADE cv0 x4 model print. L7: also check hktech -> execute once. L8: execute at the 07-20 open; reprice if it gaps >1.5%. · news |
| 0981 | SMIC | 67.70 | SELL 49.87% | 70.45 | 65.35 | 68.40 (-2.9%) | 71.17 (+1.0%) | 69.14 (-1.9%) | SMIC broke down 10% as the TSMC results read through to the whole Hong Kong semiconductor complex, and the signal flipped outright bearish into the settled close after three sessions of merely avoiding it. With the model projecting the shares going nowhere for a quarter, we sell half into any bounce rather than wait for the stop to break. Hong Kong sell signals are the reliable side of this book's record. Basis: late-flipped close-print SELL-REDUCE cv64 (AVOID x3 -> SELL-REDUCE), L4 settled-close authoritative; BreakdownDown_20D +3 -> adj 67. Stop 65.35 NOT yet broken (-3.5% away) -> the basis is the label, not a stop break. Action price = the model's own Sell Limit 70.45 (eD +0.38, +4.1% above the close, reachable). BACKSTOP: if it closes below 65.35 before filling, exit at market instead - do not let the limit strand the trim (L3). VALUATION AMPLIFY SUPPRESSED: MoS -5.52 would bump 50%->100%, but G2 fires (no DCF anchor, FV_method=EPV, Rev_Growth +16.2% hyper-growth) AND the MoS rests on a pre-crash 75.2 price vs 67.7 today. Portion 49.87% x 2500 = 1247 -> floor to lot 500 = 1000. Remaining 1500 -> stop 65.35. L7: also in hksemicon/hktech -> execute once. · news |
| 7709 | XL2CSOPHYNIX | 44.80 | SELL 49.87% | 44.80 | 40.00 | 47.67 (+6.4%) | 44.96 (+0.4%) | 47.16 (+5.3%) | This 2x leveraged SK Hynix ETF fell another 21% and closed on its low, breaking a stop we had already widened once to give the bounce room. The bounce never came and the signal has flipped from oversold-reversal to outright breakdown. We halve the position into next week's SK Hynix earnings: half the leverage out of a binary event, half still on for it, since the model still sees a modest recovery. Basis: decisive stop-loss break (44.80 vs 50.00, -10.40%) on an ALREADY-widened stop on 07-16) + a breakdown-family SELL-REDUCE. NOTE adj 63 is BELOW the 65 floor - the basis is CLAUDE.md section 9's decisive stop-loss break, a standalone valid sell basis unaffected by the conviction floor. The overlay trigger is doubly met (close 44.80 << 101 AND breakdown-family), which literally authorizes a FULL exit, but the overlay is 12d stale and the leveraged-ETF rule sets full-exit at conv>=70 + 3 prints + price<stop -> cv60 misses -> 50% trim (7747 05-13 precedent: a clean exit missed a +5.8% bounce). Model Order_Type=Market, eD 0.00 -> no price deviation. Portion 49.87% x 1800 = 900 (exactly half, lot 100). CATALYST: SK Hynix Q2 07-22 - the catalyst-aware hold is overridden by its own decisive-stop-break carve-out. Remaining 900 -> manual stop 40.00 (the model's 9.08 is degenerate at -80%). L7: also in hklev -> execute once. · news |
| 0823 | Link REIT | 39.22 | ADD — QUEUED @37.09 | 37.09 | 35.60 | 39.49 (+6.5%) | 39.44 (+6.3%) | 38.73 (+4.4%) | Link REIT is the book's strongest and steadiest buy signal - a four-print BUY-ADD on the one Hong Kong buy pattern that actually grades well - and it rose while the rest of the market was liquidated. It has run too far above its entry to chase, so we queue a limit at the model's pullback price and pay nothing unless the market comes to us. Valuation is neutral, so nothing argues against adding on weakness. Basis: stable BUY-ADD cv70->->73, PullbackBuy_VWAP (+5 actionable, n6 hit 0.75), Buy_Rank 1 -> adj 78 (highest in the book). Extended eD -2.86 -> no-chase -> GTC-limit @37.09 (fills only on a ~-5.4% pullback). Quality gate N/A: this is an ADD to a held 2800 (the gate blocks new entries only; QP=False would otherwise bite). val_adj 0 (MoS -0.02). Size: Portion 63.97% x 2800 = 1791 -> x0.5 book-note-B conservative scaling (HK buy cells grade weak) = 896 -> lot (~HK). Est_12W 38.73 is POSITIVE vs the 37.09 entry (+4.4%) but NEGATIVE vs the 39.22 close - precisely why we queue rather than chase. The 12wk exceeds the 37.84 DCF anchor by 2.4% (mild). The held trail stays 36.50 (tighter than the model's 35.60). Declined on 07-16 as extended; today it is stronger (adj and still extended -> queue. expires_on 07-22; auto-withdraw if conviction <65. L7: may also sit in a sub-book -> execute once. · news |
| 1810 | Xiaomi | 26.88 | BUY — QUEUED @24.49 | 24.49 | 22.77 | 26.91 (+9.9%) | 27.63 (+12.8%) | 27.35 (+11.7%) | Xiaomi is the only unheld candidate in the book that passes our quality screen - a 25% revenue grower with a solid balance sheet - and it is the one large-cap Hong Kong tech name that did not break in today's rout. The buy pattern is the book's most reliable, but the shares sit well above the model's entry, so we queue a patient limit nearly 9% lower and only get involved if the selling reaches it. Basis: stable BUY cv62 x4, PullbackBuy_VWAP (+5 actionable), Buy_Rank 6, val_adj -1 (MoS -0.21) -> adj 66 (clears the floor by ONE point). QUALITY GATE PASSES - Quality_Pass=True (F-Score 6, Rev_Growth +25.0%); the only unheld name that does (0316/0388/1211/9988 all fail). Extended eD -1.58 -> no-chase -> GTC-limit @24.49 (fills only on a ~-8.9% pullback). Size: = 70.59% x HK x 20% buy_pct = HK; cap 60% x 600k = HK not binding; /24.49 = 3459 -> x0.5 book-note-B scaling = 1730 -> lot (~HK); HK not binding. The 12wk 27.35 exceeds the 15.77 DCF anchor by +73% -> projection assumes growth well beyond priced-in. CAVEAT: adj 66 clears by a single point and a buy-limit under a falling tech tape can catch a knife - hence the 0.5x size and the -8.9% patience; the no_chase_queue counterfactual just flipped to 'saved' (-7.47%, n11). expires_on 07-22; auto-withdraw if conviction <65. L7: may also sit in hktech -> execute once. · news |
| 1347 | Hua Hong | 138.70 | HOLD | — | 132.00 | 143.71 | 144.29 | 146.73 | Hua Hong fell 12% on the TSMC read-through and broke its stop for the second session running, but the model still projects a recovery from here and prints no sell signal, so we hold with the stop reset below today's low rather than sell into the low. This is the position most likely to be cut next. TOP EXIT-WATCH. NO TRADE cv0 x4 (8 consecutive NO TRADE prints across two sessions; NoTrade_Reason verbatim: 'Consensus conflict + low conviction'). Close 138.70 vs stop 148.00 = -6.28% breach on an ALREADY-reset stop; -13.31% vs the 160 cost. NOT in the dispose pool: no SELL label, Portion 0, and valuation (MoS -20.9) cannot originate a sell (section 9). L2 NOT met - Est_12W 146.73 = +5.8% ALIVE. Reset stop (below the 134.30 low). HARD EXIT TRIGGER next session: a fresh breakdown-family SELL print OR Est_12W flipping negative -> dispose immediately. L7: also in hksemicon/hktech. · news |
| 0005 | HSBC | 156.90 | HOLD | — | 152.62 | 154.40 | 157.11 | 159.05 | HSBC was an island of calm on a brutal day, finishing flat, but the signal has weakened and the shares are getting stretched, so we lock in more of the move by raising the stop rather than touching the position. Conviction (-15, short of the 20-pt collapse tick, so no profit-take tick fires). RSI 70.4 stretched. Est_12W 159.05 (+1.4%, flat). PROFIT_ARMED is SPURIOUS (no Cost recorded -> gain unverifiable; T12_progress arm). Raise stop 149..62 (the model's level). H1 results ~07-30. L7: also in hkfin. · news |
Outcome & track record
Accuracy at the time of this record.
106 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.
21 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.