Recommendations
0 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 |
|---|---|---|---|---|---|---|---|---|---|
| 0700 | TENCENT | 441.00 | SELL ALL | 441.00 | 466.94⚠ | 443.44 (+0.6%) | 432.22 (-2.0%) | 430.52 (-2.4%) | Tencent's Q2 delivered 11% revenue growth but profit of RMB56bn missed consensus and fell 4% sequentially, and the market's verdict was about what that growth costs: operating capex jumped 190% year-on-year to RMB51.8bn and free cash flow went negative at -RMB13.8bn. The shares broke decisively below our risk line on the first close after the print and the 12-week view has turned negative, so we take the full exit rather than pay to find out how long the AI spending cycle runs. Exit basis: confirmed stop-loss break -5.56% (441.00 vs 466.94) plus negative 12wk (430.52) = L2 / CLAUDE.md section 9 signal-based exit, pre-committed on the 08-12 card which required a confirmed close below 466.94 on 13-Aug. Label NO TRADE on all 7 prints; scorecard NoTrade|sell +6 (n=8) supports acting. Close below EMA20 466.24 / EMA50 463.69 / EMA200 502.57, RSI 40.67. Valuation FAIR_VALUE MoS -0.486 did not originate the sell and was immaterial (no tier bump). PROFIT_ARMED on this row is meaningless - Avg_Cost 1.00 is a data artifact (+44,000%) and was ignored. Full exit not a trim: a stop-loss break is a valid full exit under the trim-discipline gates. Board lot 100. · news |
| 0981 | SMIC | 67.55 | SELL 85.46% | 67.55 | 63.71 | 68.13 (+0.9%) | 71.05 (+5.2%) | 71.42 (+5.7%) | SMIC has broken below its 20-, 50- and 200-day moving averages simultaneously with a strengthening sell signal, while renewed US-China trade tension and rare-earth export controls keep pressure on Chinese foundries and margin compression remains the standing risk to the utilisation-driven thesis. We cut 80% of the position as a one-week de-risk and keep a residual stake behind a raised trail. Signal FailedBreakout_Down_20D, STRONG SELL - REDUCE, Sell_Rank 1, conviction rising across sessions; adj 77 = 76 + 1 (cell n=59, hit_1w.564, ACTIONABLE - note the cell SOFTENED from +2 to +1 on this re-score). Valuation MoS -4.657 would bump the tier but the amplify is SUPPRESSED by G2 (DCF_per_share blank, Quality_Pass False). HK-9: this is a 1-week instrument - the positive 4wk/12wk points are the acknowledged cost of the trim, which is why are retained. Sizing 85.46% x 2,500 = at an ASSUMED (VERIFY). HK-11 thread position #2 (re-armed 08-12 on the fresh signal after the price-only re-arm < 58.86 failed to fire); already market-at-open - if unexecuted again, 08-14 is #3 = escalate-once-then-retire. Remaining 500 trail raised 61.72 -> 63.71 (2.0xATR off the 70.30 high). · news |
| 7709 | XL2CSOPHYNIX | 33.58 | SELL 50% | 33.58 | 28.00 | 34.73 (+3.4%) | 33.55 (-0.1%) | 35.12 (+4.6%) | This 2x leveraged SK Hynix tracker has printed a stable sell signal for seven straight snapshots and closed on its low, down 6.8% from the day's high. The structural memory upcycle is intact - shortage extending into 2027, DRAM pricing up ~30% sequentially - which is why this stays a trim rather than an exit, but the specific catalyst we gave it extra rope for has moved against us: SK Hynix fell 5% after approving ~$38bn for two new fabs and deferring its shareholder-return update to Q3. We take half off a product sitting 70% below cost. Signal FailedBreakout_Down_20D, STRONG SELL - REDUCE, Sell_Rank 2, adj 77 = 76 + 1. TWO SIZING LAYERS: (1) leveraged-2x-ETF rule caps the engine's 85.46% at 50%, never a full exit on a reversal-family signal (memory feedback_leveraged_etf_exits, seeded by the 7747 exit of 2026-05-13 that missed a +5.8% bounce) - STANDS; (2) yesterday's news override 50% -> 30% is RELEASED back to 50% because the SK Hynix shareholder-return catalyst it rested on has been deferred. HK-12: model Stop_Price 16.91 against a 33.58 close (-49.6%) is DEGENERATE and rejected as a risk control (ATR 8.33 = 24.8% of price; the 2.0xATR chandelier at 19.43 is equally useless) - the remaining 900 carry the 28.00 WATCH line as their numeric stop. HK-11 thread position #2. Board lot 100. · news |
| 0005 | HSBC Holdings | 163.00 | NO ACTION | 163.00 | 159.50 | 162.32 (-0.4%) | 161.12 (-1.2%) | 164.45 (+0.9%) | HSBC is trading essentially at its 12-week target, which mechanically arms a profit-take worth roughly HK - the largest single candidate in the book. We are not taking it: this exit was formally retired on 06-Aug with a re-arm at a close below 159.50, the shares closed at 163.00, and the signal itself is a rising-conviction bullish hold. The stop stays at the watch line. Armed (T12_Progress 99.12) + tick (c) fires (12wk 164.45 < 1.02 x 163.00 = 166.26) -> 30% = HKNOT taken per HK-11 (retired 08-06, re-arm 'close below 159.50' NOT fired). Label HOLD conviction 60 and RISING on bullish FailedBreakdown_Up_20D supports the restraint. Stop HELD at the 159.50 watch line (tighter than both the 158.30 model stop and the 158.30 chandelier). 14-AUG EX-DIVIDEND (second interim, record 14-Aug, pay 25-Sep): the mechanical gap is NOT a stop break. Logged optional so the cost of this discipline stays measurable. |
| 0823 | Link REIT | 38.60 | NO ACTION | 38.60 | 37.86 | 38.79 (+0.5%) | 38.68 (+0.2%) | 39.65 (+2.7%) | Link REIT arms a 30% profit-take on a repeated no-trade print, but it is the smallest of the day's candidates on dollar terms and all three forward estimates are positive, so it loses the slot to three stronger sells rather than being gated out. Stop held. Armed (T12_Progress 97.35) + tick (d) (7 consecutive NO TRADE) fires a 30% trim = HKDISPLACED ON THE STEP 7 RANKING, not gated: 30% x 2,800 x 38.60 = HKof dollar bleed ranks BELOW all three cards taken (0981 HK0700 HK7709 raw HKat the engine's 85.46%). Stated explicitly because the 08-12 run excluded it 'on G3' - under ADR-0012 as written G3 does NOT block a fired profit-take, so ranking is the honest reason. Supporting context: 1wk 38.79, 4wk 38.68, 12wk 39.65 (+2.7%) all positive; NoTrade_Reason is the low-information 'consensus conflict + low conviction'. Stop HELD 37.86 (tighter than the 37.51 chandelier). THIRD consecutive session displaced - HK-11 position #3; retire if displaced again. |
| 0522 | ASMPT | 175.80 | HOLD | 175.80 | 161.13 | 177.85 (+1.2%) | 193.38 (+10.0%) | 201.60 (+14.7%) | ASMPT carries the best forward view in the book - a +10% four-week and +14.7% twelve-week estimate - and is the one holding with no profit-take arm at all, so we simply hold and ratchet the trailing stop up behind it. No arm (T12_Progress 87.20 < 90, Gain_ATR -1.92) despite an AVOID label. Trail raised 151.79 -> 161.13 (2.0xATR off the 182.50 high). HK-12: model Stop_Price 175.80 EQUALS the close exactly - a zero-width stop, degenerate, rejected. AMAT FQ3 printed after the US close 13-Aug; HK read-through lands 14-Aug. |
| 2513 | Knowledge Atlas | 1317.00 | HOLD | 1317.00 | 1041.04 | 1301.65 (-1.2%) | 1288.17 (-2.2%) | 1390.51 (+5.6%) | Knowledge Atlas closed up 9% with conviction rising and its twelve-week view still well above the price, so the profit-take arm does not fire and the right move is a raised trailing stop, held through the Hang Seng index review it is a candidate for. Armed (T12_Progress 94.71) but NO tick fires: (c) does not fire (12wk 1,390.51 > 1.02 x 1,317 = 1,343.34), conviction RISING, label AVOID not NO TRADE, no second warning code. MANDATORY trail-raise 989.29 -> 1,041.04 (2.0xATR chandelier off the 1,324 high). CATALYST HOLD: Hang Seng quarterly review announcement 21-Aug, 2513 an inclusion contender. HK-12: model Stop_Price 1,599.96 is 21% ABOVE the 1,317 close - degenerate, rejected. |
| 0992 | Lenovo Group | 34.90 | NO ACTION | 29.98 | 23.57 | 35.42 (+18.1%) | 34.31 (+14.4%) | 37.34 (+24.5%) | Lenovo posted a record quarter - US9bn of revenue, up 43% year-on-year - and is the single highest-quality name in our buy pool on fundamentals. We are not buying it here: the shares are up roughly 115% in a month, RSI is 79, and the close sits 16% above the model's entry trigger on the day of the print. Chasing a post-earnings gap of that size is the one setup this book has repeatedly lost money on. We wait for a pullback. Adj 84 = 90 - 3 (FAIR_VALUE MoS -0.6055 -> clamp(round(-3.03))) - 3 (BreakoutUp_20D|buy, n=10, hit_1w.333, ACTIONABLE). THE ONLY buy-pool name that PASSES the quality gate (F-Score 7, FCF +HK16bn, Quality_Rank 2.0 - best in the book). DECLINED on extension x3: Entry_Distance_ATR -1.94 (~8x the -0.25 no-chase bar), RSI 78.97, post-earnings gap (L6). HK-1 marks BreakoutUp_20D a chase family. The printed Entry_Price 29.98 is DEGENERATE - a buy-stop 14% below the market fills instantly at 34.90, so the provenance rule forbids publishing it as a queue. RE-ARM: fresh BUY at a settled close with RSI < 70 and Entry_Distance_ATR >= -0.25. COST OF THE DISCIPLINE: the 08-11 buy-stop @29.29 expired 08-12 (high 29.18, 0.4% short); Lenovo OPENED 29.96 today and closed 34.90 - carrying that queue one more session was worth +HKon the 2,000-share size then contemplated. Logged optional for counterfactual grading. · news |
| 9988 | BABA-SW | 121.90 | NO ACTION | 119.35 | 114.44 | 120.97 (+1.4%) | 119.95 (+0.5%) | 117.41 (-1.6%) | Alibaba screens as the highest adjusted-conviction buy in the pool on the book's most reliable entry pattern, but it fails the quality screen outright on a negative free cash flow of -HK7bn and an F-Score of 3, the shares have already run past the model's limit, and every forward horizon points lower with the price still below its 200-day average. We pass. Adj 76 = 73 - 3 (GROWTH_PRICED_IN MoS -0.6871) + 6 (PullbackBuy_VWAP|buy, n=13, hit_1w.80, ACTIONABLE). BLOCKED on three independent grounds: HARD QUALITY GATE (Quality_Pass False - F-Score 3, FCF -HK7bn, Quality_Rank 23.0), NO-CHASE (Entry_Distance_ATR -0.59), and a NEGATIVE forward view (4wk 119.95 -1.6%, 12wk 117.41 -3.7%) with the close under EMA200 127.19. Unchanged from 08-12 - textbook HK-B falling knife. Logged optional for counterfactual grading. |
Outcome & track record
Accuracy at the time of this record.
165 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 →
-0.249 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.