HK Financials — End-of-Day Verdict
1 ACQUIRE (0939 BUY @8.96, half-size) x 2 DISPOSE (3968 SELL FULL EXIT @48.00; 2888 SELL / 30% profit-trim @235.80); 0005 08-17 buy WITHDRAWN on conviction collapse
Recommendations
1 to acquire · 2 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 |
|---|---|---|---|---|---|---|---|---|---|
| 0939 | CCB - H SHS | 8.96 | BUY | 8.96 | 8.82 | 8.91 (-0.6%) | 9.02 (+0.7%) | 8.94 (-0.2%) | China Construction Bank is the strongest technical print in the book and the clearest expression of a mainland-bank bid that showed up across the sector today while the HK/international banks and the insurers were sold. It closed at 8.96, up 0.7% on the session, above every one of its six moving averages, with the trend reading at 77 out of 100, RSI a healthy 57, and the entry sitting exactly at the close so there is no chasing involved. The valuation layer is blind here - CCB reports in renminbi against a Hong Kong dollar quote, so the model's fair-value output is a currency artefact and carries no weight either way. Against that, the model's own forward estimates are lukewarm: it sees the shares essentially unchanged over one week and twelve weeks, and a policy rate decision lands in two sessions that could squeeze mainland bank lending margins. We are buying, but at half the size the signal alone would justify. || OPERATIONAL: conviction 68 clears the 65 floor unaided (val_adj 0 NA_CCY, bias 0 - FailedBreakdown_Up_20D|buy n=1 ungraded, report-only). Buy_Rank 1, Position_Status NEW, Entry_Distance_ATR -0.00 = NO CHASE, Pillar_Consensus Consolidation_Bullish, Trend 77 / Momentum 41 / Volatility 35, above EMA5/10/20/50/100/200 = 8.874/8.865/8.815/8.681/8.543/8.284, VWAP +0.06%, 4.3% below the 9.36 52W high. TWO BOOK GATES TRIP, BOTH AT NOISE MAGNITUDE AND OFF-PATTERN, AND THE SIZE IS HALVED FOR IT: HF-4 fires (Est_1W 8.91 < entry 8.96, but only -0.28 ATR, and Est_4W 9.02 is ABOVE entry, so this is a single-leg trip versus the double-below signature of HF-4's 2888 07-16 evidence case); HF-3's inversion leg fires (Est_12W 8.94 < Close 8.96, -0.11 ATR) but HF-3 targets OversoldBounce/ReversalUp falling knives and this name is the opposite - RSI 57, above all six EMAs, near its 52W high. NEW SIZING RULE APPLIED (learnings candidate): both HF-3 and HF-4 tripping, even marginally, halves the size. Arithmetic: = 26.5% Portion x HKequity x 20% buy_pct = HK; 0.5x gate modifier = HK; / 8.96 =; board lot 1,000 -> = HK = 2.68% of book. HKand max_position_pct 60% neither bind. Stop 8.82 = -1.56% / 0.79 ATR - honoured without argument. Tipping factor: no_chase_queue counterfactual n=7, median +1.60% @1wk / +11.32% @4wk labelled COST, against 82% of capital sitting idle - this book's habitual withholding is now the more expensive error. Est_* conf 80: 1wk 8.91 [8.57-9.36] -0.56%; 4wk 9.02 [8.27-9.92] +0.67%; 12wk 8.94 [7.72-10.61] -0.22%. Target_Price 9.31 (+3.91%) is the risk objective, distinct from the 4wk estimate. Entry range 8.92-9.00 (Entry_Price +/- 0.25 ATR); REPRICE if the 08-19 open gaps outside 8.83-9.09 (+/-1.5%, L8). No DCF annotation - NA_CCY makes the 126.01 DCF a CNY-vs-HKD artefact. L1 discount applied and overridden on the counterfactual; L6 no post-earnings gap (interims are 08-28, ahead not behind). · news |
| 3968 | China Merchants Bank | 48.00 | SELL ALL | 48.00 | 48.10⚠ | 47.99 (-0.0%) | 48.40 (+0.8%) | 49.15 (+2.4%) | We are closing the China Merchants Bank position in full. The risk line we wrote down a week ago was a close below HK, and the shares have now closed below it three sessions running - today at 48.00. More importantly, what was previously just an absence of any signal has turned into an outright bearish one: the model now reads the stock as a failed upside breakout and rates that read at 62 out of 100 conviction, with the exit strategy marked immediate. The shares sit below all six of their moving averages and nearly 2.5% below their volume-weighted average price. This is a sound bank - it is the highest-quality business in the book on our own screens - but the price action has broken and a pre-committed stop is not something to re-argue on the third day it triggers. We take a small loss of about 1.3% and move the capital on. || OPERATIONAL: ORIGINATION = (i) pre-committed stop break, close 48.00 vs stop 48.10 = -0.21% / -0.13 ATR, THIRD consecutive session below (08-14 -0.29%, 08-17 -0.58%, 08-18 -0.21%); the 08-13 card's written rule 'FULL EXIT of the residual on ANY close below 48.10' has now fired three times. (ii) NEW AND STRONGER THAN YESTERDAY: the label has turned from NO TRADE cv0 to a formal bearish FailedBreakout_Down_20D at cv62, Pillar_Consensus Consolidation_Bearish, Exit_Strategy IMMEDIATE, Action_Rationale 'Held; bearish signal (conv=62); do not add', Trend 36. This is an affirmative bearish read, not merely the absence of a bullish one. (iii) Structural breakdown intact: close 48.00 below EMA5 48.11, EMA10 48.39, EMA20 48.42, EMA50 48.02, EMA100 48.09, EMA200 48.32 - all six. VWAP -2.48%, RSI 47.5. SCORECARD, STATED WITHOUT LAUNDERING: NoTrade|sell has just crossed the actionable floor at n=5, hit_1w 0.0, bias -10 - a 0-for-5 record on the exact exits this book kept issuing on this very name. Today's row is FailedBreakout_Down_20D, not NoTrade, so that -10 does not mechanically apply and adj_conviction is unhaircut at 62; the family change is NOT being used as a laundering device and the 0-for-5 is carried as an explicit caution. It is overruled because the origination is a pre-committed risk line, not a signal-family bet. L2 does NOT fire for a FIFTH session - Est_12W 49.15 is +2.40% above the close, so the negative-12wk leg fails; this exit rests on the stop plus the bearish label, NOT on L2. HF-2: PROFIT_ARMED printed but SUPPRESSED for a sixth session on the arm criterion itself - Gain_ATR -0.85, the position is AT A LOSS, T12_Progress 97.66 is a near-flat-target artefact. This is NOT a profit-take. HF-5 cleared: no 3968 ex-dividend in the 08-18/08-31 horizon window, last ex-date 2026-07-02 (33 sessions past). Counter-evidence recorded and overruled: breach is shallower than yesterday and inside one ATR; forward estimates still mildly positive; 3968 is Quality_Pass TRUE / F-Score 5 / MoS_FV +0.245, the book's best business; optional_trim counterfactual n=50 median +1.60%/+5.96% saved. Note the sector tension honestly: mainland banks were BID today (0939 +0.67%, 3328 +0.62%, 3988 +0.28%) and 3968 rose +0.38% with them - but from a base below every EMA, and it is the one mainland bank that has not repaired its structure. Action price 48.00 = anchor settled close (the AVOID row's Entry_Price 48.00 agrees); range 47.81-48.19; REPRICE if the 08-19 open gaps outside 47.28-48.72. Board lot 500 -> = 4 clean lots. Proceeds HK, realised -HK (-1.34%), cumulative realised -> +HK. The exit steps out of a 12wk band whose downside is 39.05 (-18.6%). FOURTH re-issue - the auto-executor has not run in eight sessions. · news |
| 2888 | STANCHART | 235.80 | SELL 30% | 235.80 | 228.47 | 236.17 (+0.2%) | 233.78 (-0.9%) | 228.32 (-3.2%) | We are taking a 30% profit-trim on Standard Chartered, banking roughly HKof gain and keeping three-quarters of the position. The shares are up 3.1% on our cost and have run to within 1.8% of their 52-week high, but today they made a lower high and closed down 0.9%, and the model's conviction in the setup halved from in a single session. Yesterday we wrote down an explicit condition - trim if conviction prints 45 or below - precisely so this would not become a judgement call on the day. It printed 43. With Federal Reserve minutes landing tomorrow and a run of macro events through month-end, taking a partial profit off a name near its highs is the disciplined move; the remaining stay in with the stop unchanged at 228.47. || OPERATIONAL: ADR-0012 TICK ACCOUNTING - ARMED via PROFIT_ARMED, but through T12_Progress 103.28 = Est_12W 228.32 BELOW the 235.80 close, the HF-2 inversion artefact (Gain_ATR 1.55 < 3.0, so progress is the only arming route). Ticks: (a) NO - PROFIT_ARMED is the sole code on the row; (b) YES - conviction = -22, clearing the 20-point threshold; (c) technically yes (228.32 < 1.02 x 235.80 = 240.52) but EXCLUDED AS COLLINEAR with a T12_Progress-derived arm for the SEVENTH consecutive session, per HF-2 - one inverted 12wk number must not both arm and fire the same trim; (d) NO - label HOLD. ONE INDEPENDENT TICK -> profit_take_1 = 30%. SECOND, INDEPENDENT ROUTE TO THE SAME ANSWER: yesterday's written re-arm trigger (i) 'conviction printing <= 45 again' -> 43 <= 45 -> FIRES. Both routes agree, which is why this is a mandatory DISPOSE card and not prose - optional-trim language is banned (ADR-0012). GATES: G1 does not protect - the signal is FailedBreakdown_Up_20D cv43, nowhere near the fresh-breakout-at-cv>=80 pattern G1 guards, and tick (b) fires regardless of G1 under the softened rule. G2/G3 govern the valuation-elevate path only and do not block a fired profit-take. CATALYST-AWARE HOLD RESOLVED CLEANLY: the FOMC-minutes stance is tighten-trail, and the carve-out permits a trim on a CONFIRMED LOWER HIGH - today's high 238.60 is below the 08-13 high and the close 235.80 is below the 238.00 prior close, off the 240.00 52W high. Confirmed. VALUATION BARRED: MoS_FV -7.51 is a GBP/USD-reporting-vs-HKD-price artefact on an NA_CCY row -> val_adj 0, NO tier bump; the trim stays at the minimum 30% tier. SIZE: 30% x 1,100 =; board lot 50; yesterday's pre-commitment wrote explicitly and is HONOURED ->, residual 750. Proceeds HK; realised on the slice (235.80 - 228.63) x 350 = +HK. STOP HELD AT 228.47, NOT RAISED: the model's printed Stop_Price today is 226.56, BELOW the watermark, so the ratchet holds and the printed value is not adopted; a highest-close chandelier (238.80 - 2.0 x 4.6183) gives 229.56, but stacking a stop-raise on top of a trim would double-de-risk a name 1.8% off its 52W high into a two-sided macro print. The trim IS the de-risk. Cushion 235.80 vs 228.47 = 1.59 ATR. ADR-0020 Trail Erosion: 228.47 vs peak_stop 228.47 = 0.00%, not applicable (and default-off outside us). COUNTER-EVIDENCE RECORDED AND OVERRULED: optional_trim counterfactual n=50, median +1.60% @1wk / +5.96% @4wk SAVED - this book's trims have historically been wrong, and Trend_Score is still 77 with the close above all six EMAs and VWAP +2.13%. Overruled because a fired tick plus a written pre-commitment is not an optional trim. HONEST SCORING OF YESTERDAY: withdrawing the 08-14 trim at 238.00 has cost -0.92% in one session; relative to the original 08-14 ref of 234.20 the withholding is still +0.68% ahead. The -> conviction oscillation across four sessions is itself the finding, and is logged as systemic. News: no dated headline for 08-17/08-18; the HKD 0.204 dividend went ex on 2026-08-06, eight sessions past and already in the price, so HF-5 does not apply; chairman-succession chatter (Sajid Javid named a candidate) is governance colour, neutral, no override. Action price 235.80 = anchor settled close = the row's printed Entry_Price; range 234.65-236.95; REPRICE if the 08-19 open gaps outside 232.26-239.34. Re-arm for the NEXT tranche: trim a further 30% on any second Exit_Warning code alongside PROFIT_ARMED, or on conviction <= 35; a close below 228.47 is a FULL-EXIT basis. · news |
| 3988 | Bank of China | 5.32 | WATCH | 5.32 | 5.26 | 5.33 (+0.2%) | 5.38 (+1.1%) | 5.36 (+0.8%) | Bank of China is the closest near-miss in the book - a constructive setup that is simply three points short of the conviction we require to commit capital. We are watching it, not buying it. || OPERATIONAL: WATCH cv62 vs the 65 floor; val_adj 0 (NA_CCY), bias 0 -> adj 62. WATCH is not an actionable label. Trend 70, above all six EMAs (5.291/5.292/5.284/5.232/5.125/4.939), RSI 53.4, Est_12W 5.36 vs close 5.32 NON-INVERTED (+0.75%), Est_1W 5.33 > entry 5.32 so HF-4 CLEARS - materially cleaner on both book gates than the 0939 acquire, but 3 points short on the one number that decides. Trail ->. Re-arm: BUY label with conviction >= 65 and Est_1W_Point still above entry. |
| 3328 | Bank of Communications | 7.34 | WATCH | 7.34 | 7.22 | 7.32 (-0.3%) | 7.40 (+0.8%) | 7.37 (+0.4%) | Bank of Communications rose with the mainland-bank complex today but the setup is still seven points below our commitment threshold. No action. || OPERATIONAL: WATCH cv58 vs the 65 floor; val_adj 0 (NA_CCY), bias 0 -> adj 58. Trend 64, above all six EMAs, RSI 54.0, Est_12W 7.37 vs close 7.345 non-inverted, but Est_1W 7.32 < entry 7.34 so HF-4 would fire even if the floor cleared. Trail ->, rising. Recorded as the third mainland-bank near-miss - the sector bid is real and only 0939 converted it into an actionable print. |
Outcome & track record
Accuracy at the time of this record.
12 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 →
0.145 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.
50 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.