This website is for educational purposes only and does not constitute investment advice. Always do your own research and assess your own risk tolerance before making investment decisions.

HK End-of-Day Verdict 27 Jul 2026, 16:00:00 GMT+8

Hong Kong — End-of-Day Verdict

2 ACQUIRE (both QUEUED GTC-limit) + 2 DISPOSE (both MARKET) - pending next open 2026-07-28

Recommendations

2 to acquire · 2 to dispose

TickerNameCloseActionAction 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
3690Meituan-W89.30BUY — QUEUED @80.3380.3375.4588.56 (+10.2%)89.04 (+10.8%)83.38 (+3.8%)Meituan is the book's best buy setup right now - a maximum-conviction pullback signal on the one Hong Kong buy pattern that actually grades well, and it ranks first in the book. The catch is that it has already run 10% above the price the model wants to pay, so we do not chase it: we leave a limit order at the model's entry and pay nothing unless the market comes back to us. Valuation is unavailable for this name, so nothing argues either way. Basis: STRONG BUY cv100 x4 stable, PullbackBuy_VWAP (HK-2 anchor cell, bias +3, n7 hit_1w.667), Buy_Rank 1 -> adj 103, the highest in the book. Valuation NA -> quality gate OFF, val_adj 0, size_mod 1.0. Extended eD -2.08 -> HK-4 no-chase -> gtc_limit @80.33 (fills only on a -10.0% pullback from the 89.30 close). Entry range 79.25-81.41 (+-0.25 ATR). Size: Portion 71.55% x HKequity x 20% buy_pct = HK -> /80.33 = 1,069 -> board lot 100 -> (~HK). Est_12W 83.38 is POSITIVE vs the 80.33 entry (+3.8%) but NEGATIVE vs the 89.30 close (-6.6%) - precisely why this is a queue and not a buy. Est_1W +10.2% / Est_4W +10.8% vs the entry, 80% conf. Stop 75.45 (-6.1%). expires_on 2026-07-30 (3 sessions); auto-withdraw if conviction drops <65. L7: also lives in hktech - execute once. · news
0939CCB - H8.99ADD — QUEUED @8.848.848.458.99 (+1.7%)9.02 (+2.0%)8.95 (+1.2%)China Construction Bank strengthened decisively through the session - from a hold to the book's strongest buy-add by the close - as it broke to a new 20-day high. We are adding to a position we already own, but only at a modest size and only on a small pullback, because breakout entries are the pattern this book has historically been worst at timing. Basis: HOLD cv80 -> BUY-ADD cv73 x2 -> STRONG BUY - ADD cv90 at the close, BreakoutUp_20D, Buy_Rank 2. Scorecard haircut -5 (BreakoutUp_20D|buy n8, hit_1w.25) -> adj 85, still well clear of the 65 floor. HK-1 confirmation requirements MET: 3 consecutive bullish ADD prints and RSI 67.9 < 70 (not yet overbought); HK-1 'never up-size' honoured - size_mod held at 1.0. Valuation NA_CCY -> blind, val_adj 0; the quality gate does not apply to an ADD in any case. Extended eD -0.78 -> no-chase -> gtc_limit @8.84 (1.7% below the 8.99 close), entry range 8.79-8.89. Size: Portion 70.29% x 50,700 held = 35,637 -> x0.5 book-note-B conservative scaling (HK buy cells grade weak) = 17,818, then capped at the book's standard buy ticket (70.29% x HKx 20% = HK) because the 0939 line is already the book's 3rd largest at HK -> 9,541 -> board lot 1,000 -> (HK). Est_1W +1.7% / Est_4W +2.0% / Est_12W +1.2% vs the 8.84 entry (80% conf) - a modest forward view; the trade is the breakout, not the projection. Stop 8.45 (-4.4%). FOMC 29-Jul is two-sided on bank NIM (horizon stance: hold-through, +-3.5%) - the limit sits below the market, so a hawkish gap fills us cheaper rather than worse. Held trail raised 8..45. expires_on 2026-07-30; auto-withdraw if conviction <65. L7: also lives in hkfin - execute once. · news
2513Z.AI (ex-Knowledge Atlas)1281.00SELL ALL1281.001287.24 (+0.5%)1200.34 (-6.3%)1257.71 (-1.8%)This position is down 22% and the model's own forward projections now sit below today's price at both the one-month and three-month horizons - the case for owning it has expired. On top of that, a block of previously locked-up shares becomes sellable tomorrow, which adds supply into an already weak tape. We exit the whole line at the open. Basis: AVOID cv60 x4 stable, FailedBreakout_Down_20D (HK's reliable down-trend sell family, bias +2, n45 hit_1w.579) -> adj 62. Forward view dead: Target_12Week 1257.71 sits BELOW the 1281 close, Est_4W 1200.34 (-6.3%), Est_12W 1257.71 (-1.8%); only Est_1W 1287.24 (+0.5%) is positive. L2/HK-3 admit this to the dispose pool despite the AVOID (not SELL-REDUCE) label. NOTE ON THE FLOOR: adj 62 is below the 65 fresh-issue bar - this is NOT a fresh origination but the RE-ISSUE of a carried, unexecuted pending dispose first raised 2026-07-21 (6th consecutive session; conviction has drifted while the thesis is unchanged), so it re-issues under L3/HK-5 execution discipline rather than clearing the origination floor. Valuation NA -> blind, no amplification available. Frees HKto. Trim-discipline gates G1-G3 do not apply: this is a loss-cutting trend exit on a broken name, not a profit-trim on a winner, and the PROFIT_ARMED flag on the row is spurious (systemic finding #1). Sells MARKET at the 2026-07-28 open; reprice if the open gaps >~1.5%. L7: also lives in hktech - execute once. · news
7709XL2 CSOP Hynix 2x ETF54.50SELL 50%54.5044.0056.60 (+3.9%)53.77 (-1.3%)56.43 (+3.5%)This leveraged Hynix ETF is down 52% and the model escalated it today from avoid to its strongest sell rating, ranking it the number one thing to reduce in the book. It also faces a binary earnings event in two days that a doubled-up ETF turns into a roughly 14% swing either way. We halve the position at the open to take that risk down, and keep the other half - a full exit is not yet justified. Basis: escalated AVOID cv58 (Fri) -> STRONG SELL - REDUCE cv68 x4 stable, Sell_Rank 1, FailedBreakout_Down_20D (bias +2, n45) -> adj 70, clears the floor. Portion: the model prints 85.46% but the leveraged-ETF rule caps a low-conviction reversal at 50% - a full exit needs cv >=70 PLUS 3 prints PLUS price < Stop_Price, and cv is 68 with the close at 54.50 against a 32.30 model stop, so two of three legs fail (the 2026-05-13 full exit on a similar print missed a +5.8% bounce). @ ~54.50 frees HKHONEST TENSION: Est_1W 56.60 (+3.9%) and Est_12W 56.43 (+3.5%) argue mildly AGAINST trimming; the case rests on the stable STRONG SELL - REDUCE label from HK's most reliable sell family, the -52.4% drawdown, the 2x leverage into a binary print, and this being the 4th consecutive unexecuted re-issue (L3/HK-5: untaken HK trims graded 'saved', median +2.78%). Est_4W 53.77 (-1.3%) is the one negative leg. Valuation NA_ETF -> blind, no amplification. RESIDUAL : trail held UNCHANGED at 44.00 (0.95 ATR below the close) rather than tightened - the horizon's tighten-trail stance is already discharged by the 50% cut, and a tighter stop inside a +-14% event band would knife the stub out on noise. Sells MARKET at the 2026-07-28 open; reprice if the open gaps >~1.5%. L7: also lives in hklev - execute once. · news
0005HSBC Holdings162.80HOLD157.50160.95163.13168.14HOLD: conviction wobbled -> across the session and the signal label drifted through three families (AnticipationUp_VCP -> BreakoutUp_52W -> MomentumContinuation_Up), with RSI at 76.3 (overbought). That is deterioration, but the label is still HOLD, not SELL-REDUCE, and Est_12W 168.14 is positive (+3.3%) - no signal-based sell basis exists. HSBC H1 on 04-Aug is a hold-through catalyst; do not pre-trim into the print. Response is a trail raise, not a trim: 157..50 (chandelier 2.0x ATR = 162.80 - 5.26 = 157.54). PROFIT_ARMED on the row is spurious (systemic finding #1) and Avg_Cost is blank, so no genuine profit-take can be evaluated on the book's largest line (HK) - systemic finding #2. Watch: a close below 157.50, or a second sub-60 conviction print alongside a negative 12wk, turns this into a dispose candidate. · news
0522ASMPT161.60HOLD156.50164.51174.19178.15HOLD - and explicitly NOT a dispose. The 14:00 print was a full-conviction STRONG SELL - REDUCE (cv65, Portion 85.46%, Sell_Rank 2) but it de-escalated to NO TRADE cv0 for the next three prints as the name rallied ~+3.1% into the close. L4 (the settled close is authoritative) governs: the sell case did not confirm, so it is withdrawn rather than actioned. The position is -17.7% but the forward view is the book's strongest - Est_4W 174.19 (+7.8%) and Est_12W 178.15 (+10.2%). GROWTH_PRICED_IN MoS -4.43 is extreme, but with NO early Exit_Warning code on the row there is no valuation-elevate path (CLAUDE.md 9c), and G2 would suppress the amplify here in any case. The interim on 29-Jul is a hold-through catalyst. Trail carried at 156.50 (the 161.60 close is 3.3% clear). Flagged to systemic finding #5: a sell-label swing within an hour on a 3% move is an engine stability defect. · news
2888Standard Chartered226.00HOLD215.07226.49227.57227.60HOLD: firming through the session, cv -> on FailedBreakdown_Up_20D, with the 226.00 close well above both the EMA50 (212.36) and EMA200 (186.11). The forward view is flat-to-slightly-positive (Est_12W 227.60, +0.7%) - not a buy case, and no sell case either. H1 lands 29-Jul alongside FOMC; hold through the print, no pre-trim. Trail raised 211..07 (the model's own Stop_Price), locking in the session's firming while leaving 4.8% of room - inside the +-1.83 ATR event band but acceptable on a small HK line. Avg_Cost blank (systemic finding #2), so and any cost-based arming are unknowable. · news
0823Link REIT39.10HOLD37.5939.0739.3838.77HOLD with a tightened trail. The signal is stable and mildly constructive (HOLD cv60 -> 63, FailedBreakdown_Up_20D, the 39.10 close above both EMA50 38.24 and EMA200 38.07), but the forward view is flat-to-slightly-negative (Est_12W 38.77, -0.8% vs close; Est_4W 39.38, +0.7%). The actionable point is the event: a live ~35% FOMC hike tail transmits mechanically through the HKD peg to Link REIT's financing cost and discount rate, and a +-3.5% scenario on this name is +-1.87 ATR - wider than a standard stop. Per the horizon's tighten-trail stance the response is a stop raise, not a trim: 37..59 (the model's own Stop_Price, 3.9% below the close). GROWTH_PRICED_IN MoS -0.03 is effectively neutral - no valuation input either way. · news
0700Tencent443.00HOLD430.00439.57436.53437.20HOLD, but this is the book's closest non-actioned sell case. A 4th consecutive AVOID cv60 on FailedBreakout_Down_20D with a negative forward view (Target_12Week 437.20 and Est_4W 436.53 both BELOW the 443.00 close), trading under both EMA50 (457.88) and EMA200 (505.79). What holds it: the 07-05 overlay override sets a specific exit trigger - a decisive close below 430 - and at 443.00 that floor is intact, so the L2 stop-break leg is not satisfied. The line is also small at (HK). Avg_Cost is a 1.00 placeholder, which is why the row shows a nonsense +44,200% reading (systemic finding #2). Watch 430: a decisive close below it is a mandatory full exit the next session, no further confirmation needed. · news
0981SMIC70.65HOLD66.0071.3575.7074.58HOLD: flat NO TRADE cv0 x4 ('Consensus conflict + low conviction'), with the 70.65 close below the EMA50 (74.90) but above the EMA200 (68.70) - a genuine no-man's-land, which is exactly what the engine is saying. The forward view is positive (Est_4W 75.70 +7.1%, Est_12W 74.58 +5.6%), so there is no exit basis. GROWTH_PRICED_IN MoS -5.18 is extreme but carries NO early Exit_Warning code, so the valuation-elevate path does not open (CLAUDE.md 9c). Q2 on 06-Aug is hold-through - an add is only considered AFTER the print on signal confirmation, never anticipatorily. Stop carried at 66.00 (6.6% below). Avg_Cost blank (systemic finding #2). · news
1347Hua Hong Semiconductor148.60HOLD145.00151.92148.45162.63HOLD: flat NO TRADE cv0 x4, only -7.1% underwater, and the 12-week view is the second-best in the book (Est_12W 162.63, +9.4% vs the 148.60 close). GROWTH_PRICED_IN MoS -19.88 is the most extreme valuation reading in the portfolio, but with NO early Exit_Warning code on the row there is no valuation-elevate trim (CLAUDE.md 9c) - and G2 (DCF-reliability) would suppress the amplify on a hyper-growth foundry regardless. The PROFIT_ARMED flag is spurious (systemic finding #1); the position is at a loss. Q2 on 06-Aug lands in the same session as SMIC's, so the book takes a correlated +-9-10% draw rather than two independent ones - that is a sizing observation, not a sell reason. Hold-through; trail carried at 145.00 (2.4% below the close - tight, and deliberately so into a binary print). · news
3750CATL635.00HOLD488.81631.74609.88616.85HOLD: flat NO TRADE cv0 x4, the position -15.5% underwater, the 635.00 close below the EMA50 (654.06) but comfortably above the EMA200 (584.13). The forward view has turned mildly negative (Est_4W 609.88 -4.0%, Est_12W 616.85 -2.9%), which is half of the L2 exit test - but the other half is absent: the close is 30% above the 488.81 stop, nowhere near a break. FAIR_VALUE MoS -1.24 is unremarkable and provides no amplification. No signal-based sell basis, so no trim. Watch: a bearish label (SELL-REDUCE or a breakdown-family print) landing while the 12wk stays negative would complete the L2 case without needing the stop to break. · news

Outcome & track record

Accuracy at the time of this record.

122 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.
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.

This website is for educational purposes only and does not constitute investment advice. Always do your own research and assess your own risk tolerance before making investment decisions.