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 17 Aug 2026, 16:00:00 GMT+8

HK Semiconductors — End-of-Day Verdict

EOD 2026-08-17 (hksemicon)

Recommendations

0 to acquire · 1 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
0992Lenovo Group Limited33.02SELL 30%33.0230.91n/an/a47.05 (+42.5%)Re-issuing the Lenovo profit-take that never executed, and it is more warranted tonight than it was on Friday. The 30% trim decided at the 08-14 close was recorded for the Monday open and the Auto-Executor never saw it - the card was written with a non-vocabulary execution type - so the book still holds all. Monday opened at HK (+2.6% above the HKreference, i.e. a better price than the one we asked for), traded up to 34.94, then sold off all day to close 33.02, -1.73%, on the single day when every other name in this twelve-name semiconductor book rose (0148 +7.1%, 1347 +7.5%, 0981 +6.1%, 6869 +12.1%, 7709 +10.5%). Lenovo was the only decliner in its own sector on a broad sector rally two sessions after a blowout print - that is distribution, not consolidation. The model dropped the setup to NO TRADE at zero conviction on a consensus conflict, with the momentum pillar halving from, so the deterioration probe fires for a second consecutive session. We sell the minimum 30% - HKof realized gain on a +34% position - and let 70% keep running against a 12-week view of HK (+42.5%) that the record AI-server pipeline still supports; fair value reads FAIR_VALUE with a margin of safety of -0.61, so valuation neither argues for a deeper cut nor a smaller one. || OPERATIONAL: ADR-0012 profit-take, ARMED + FIRED on tick (b). Arm: PROFIT_ARMED on the anchor row, Gain_ATR 3.43 >= 3.0 (hand-check (33.02-24.56)/2.4636785 = 3.4339, Avg_Cost 24.56 reconciles with input/hksemiconportfolio.json so the HS-22 pre-fill artifact does not apply), T12_Progress 70.18 < 90 (fifth consecutive session at 70.18 - structurally pinned), +34.45% so HS-2 target-exhaustion does not apply. Ticks: (a) NO - PROFIT_ARMED prints alone for a third session, no RSI_DIV / CLIMAX_VOL / AVWAP_BREAK / EMA10_BREAK even on a -1.73% close against a +6% sector, which is pending HS-25 confirmed a third time; (b) YES - conviction, -67 points, the tick; (c) NO - 12wk 47.05 vs 1.02 x Close 33.68 = alive at +42.5%; (d) NO - 08-14 was HOLD so this is the FIRST NO TRADE print, and HS-4 requires two consecutive before a cv0 is deterioration in its own right. Honest caveat on tick (b): a NO TRADE row prints cv0 by construction (Pillar_Consensus Conflicted), so -67 overstates the signal-quality loss; the underlying deterioration is Momentum with Trend flat at 78, which is real but smaller than -67 implies. The tier is unaffected - 1 tick is 30% either way - and the direction agrees with Friday's independent read. 1 tick -> profit_take_1 = 30%, never 100% from a profit-take. G1 does not engage (NoTrade row, cv0, not a fresh BreakoutUp_* / MomentumContinuation_* at cv>=80) and tick (b) is exempt from G1 regardless; G2/G3 govern only the s9c valuation-elevate path, which is not the basis here. Valuation amplify NONE: MoS -0.6055 is far from the <= -3 tier bump and not UNDERVALUED, so val_adj 0 / size_mod 1.0. NOT concentration-originated - the 61.25% vs 60.0% max_position_pct breach (headroom -HK) is pure appreciation and per CLAUDE.md s9 blocks ADDs only; the card would be identical at 40% invested. Sizing: 30% x 44,000 = -share Lenovo board lot -> (31.8%), unchanged from the lapsed 08-14 card because the tier and the base position are both unchanged - this is a re-issue at a new price, NOT a second order and NOT an escalation. Proceeds HKcost basis HKrealized +HK (+34.45%) - the book's first realized gain; (58.24%); residual = HK = 41.76%. ACTION PRICE 33.02 = the anchor CLOSE, because a NO TRADE row prints Entry_Price / Stop_Price / Target_Price BLANK - stated as a deviation from the usual Entry_Price provenance and adopted because the alternative (carrying Friday's 33.60) is exactly the stale-limit republication the provenance rule forbids. Entry range 32.40-33.64 (+-0.25 x ATR 2.4637). ESTIMATES: 4wk and 1wk are n/a - Target_Price is blank on a NO TRADE row so there is no anchor to interpolate from and f (0.30 per L5, calib_1w -0.138 degenerate) has nothing to scale; Friday's 38.77 is deliberately NOT carried. 12wk 47.05 (+42.5%) stands and EXCEEDS the DCF anchor of HKby 3.55 - annotated, not modified. The scan's own forward envelope is Est_Low 31.54 - Est_High 39.67. What the 14,000 step out of if the 12wk is right: ~HKscorecard_bias 0: the NoTrade|sell cell exists at n=1 and graded 0-for-1 on both horizons - that is the 07-20 exit HS-7 flags as good_sell:false / missed_up +14.94%, the one prior time this book sold on a cv0 NO TRADE row and was wrong. Report-only (n=1), and materially different: 07-20 was a 100% dump near flat, tonight is a 30% profit-lock at +34% with 70% left running. Recorded as the standing caution on this card. COMPANION TRAIL RATCHET 30.65 -> 30.91 = 2.13x ATR (2.4637) off the unchanged 36.16 52-week high, the book's standing spacing; it sits HK / 0.83 ATR below today's Low of 32.96 so it is outside the bar (Systemic #11 clear; the HS-23 gap test does not engage, range/ATR = 0.80 < 2). Cushion 6.39% / 0.86 ATR - honestly tight, and under HS-3 a decisive close below it is an unconditional FULL exit of the residual; affordable only because today's cut removes from that distance (residual exposure to the trail HKvs HKon full size at 30.65). Pending HS-7 (partial-not-full on a shallow breach of a positive-12wk wide-ATR name) is INERT and NOT applied. Risk at trail on the residual: 30,000 x 30.91 = HKvs cost HK = +HK / +25.86% locked; with the realized gain that is HKof the HKunrealized either banked or stop-protected = 83.0%, up from 77.8% intended on Friday. Peak Stop re-bases 30.65 -> 30.91; ADR-0020 erosion 0%, NOT fired, and the cut satisfies the size-reduction test regardless. HS-3 breach test: NO breach (33.02 vs the 30.65 trail in force). HS-3a moot - the model prints no Stop_Price at all tonight; twelfth consecutive session governed by the hand-maintained trail. NEXT RATCHET, objective test per the HS-24 anti-drift discipline: tighten to the then-computed 2.13x ATR chandelier on any new 52-week high above 36.16; otherwise HOLD 30.91 - no downward drift, no discretionary skips. PRE-COMMITMENT: a second consecutive NO TRADE print at the 08-18 close fires tick (d) alongside tick (b) = 2 ticks = profit_take_2 = 50% of the then-residual, mandatory, no further judgement required. NEWS neutral, no override: no new same-day headline (the Yahoo quote-news page for 0992 has now failed or returned stale content on seven consecutive sessions, so this is search-indexing plus three article fetches, two of which returned February/April 2026 content). The standing two-sided case is unchanged - record Q1 revenue US94bn +43% y/y vs a US-22.4bn consensus, AI revenue +60% to US3bn = 35% of group, AI server pipeline US$54bn +157% q/q, against global memory supply constraints management itself expects to persist through the year (CEO Yang: shortages were 'accurately anticipated' and memory is locked for 2026). The one valuation-flavoured item, dated 08-12 and therefore pre-print, has Lenovo at 23.6x P/E against a tailored fair 18.3x and 160% one-year returns - mildly supportive of trimming, not new, and not an override. No 0992 event anywhere in the fresh 08-17 horizon window; the nearest read-through touching this book is NVIDIA Q2 FY2027 on 08-26 (0981/1347/0522/7709/7747/6869, band +-14%), which does not name 0992. L8: executes at the 2026-08-18 open; 33.02 is the close-ref, reprice if the open gaps >~1.5% (outside ~32.53-33.51) and ACT on the reprice rather than letting the order lapse - letting it lapse is precisely the failure this card is correcting, and it has now cost the book HKagainst Monday's 34.48 open. execution is written as 'market' (NOT 'market_on_open') so scripts/paper_executor.py can actually see it; the 08-14 card is marked superseded/withdrawn in this same ledger so only one order is ever live. · news

Outcome & track record

Accuracy at the time of this record.

13 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 →
-0.138 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.

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.