A-Share Market Recap | 2026-08-31 Monday

Data read-time: everything is on the 2026-08-31 15:00 close basis; per-name quotes carry trade timestamps (Tencent qt.gtimg.cn, measured timestamps 15:30:57–15:31:32, confirmed to be post-close settlement data); the limit-up / limit-down / failed-limit pools are the East Money push2ex official pools on the 8/31 basis, and all three pools passed the tc == len(pool) check (limit-up 88/88, limit-down 11/11, failed-limit 30/30); sector moves and net inflows are the THS industry-sector close snapshot (ak.stock_board_industry_summary_ths, 90 sectors); index intraday lines are Sina 5-minute candles.

Scope boundaries of this piece (judge the strength of the conclusions accordingly): ① The reconciliation is given on two baselines: "vs the 8/28 close" and "vs today's open". All three major indices gapped down today (Shanghai Composite −0.65%, Shenzhen Component −1.35%, ChiNext Index −1.67%), and the pre-market list was published at 07:00, so the earliest actually executable price is today's opening price, not the prior close; reconciling only against the prior close systematically understates today's execution result (measured gap in this piece: 2.34pct). ② No block-trade (Dragon-Tiger) data today: the exchanges disclose it after 18:00, and this piece was written at 15:30, ahead of disclosure, so the Dragon-Tiger portion of §5 is left blank — this is not an interface failure. Last Friday's (8/28) Dragon-Tiger data was already added in this morning's pre-market piece. ③ No northbound flow data: since 2024-08-19 the daily net-buy figure for Stock Connect has stopped being disclosed, so this piece gives no northbound net-inflow number, and quotes no interface value that returns 0.0. ④ Advance/decline counts are the aggregate of the 90 THS industry sectors (3,181 up / 2,218 down, 5,399 names in total), excluding suspended names and some ST names not assigned to an industry sector; they do not exactly equal the total number of listed stocks and are used only to judge the up/down structural ratio. ⑤ Sector net inflow is on the THS basis, which differs from East Money's "main-force net inflow" basis and is not subtractable across pieces; East Money clist remains unavailable today, so there is no second independent basis to cross-check sector flows today. ⑥ No verifiable single same-day catalyst was obtained for today's media/film leadership, so this piece labels it honestly as "catalyst in doubt" and gives no specific attribution (see §3.2).

⚠️ Data-fetch failures and method notes (this section is not sent to clients):

I. One silent trap that must be recorded — getTopicZTPoolPrevious returns the same day, not the prior day

• Per the skill workflow, stock_zt_pool_previous_em / getTopicZTPoolPrevious is used to get "how yesterday's limit-up names performed today". Measured, date=20260831 returns tc=88, len=88, which is exactly the same set as today's limit-up pool from getTopicZTPool (88 names) (code-by-code comparison set==set is True, intersection 88).
• That is, the interface returns today's limit-up pool, not "how the 8/28 limit-up names performed on 8/31". Taking it at face value would produce the absurd conclusion that "all 88 of yesterday's 88 limit-up names hit limit-up again today".
• Switched to manual construction: take the date=20260828 limit-up pool (82 names, tc==len check passed) → pull today's price for each via Tencent quotes → compute the promotion rate myself. This is the same trap as zt-pool-previous-returns-same-day in memory; this is the second time stepping into it, and it should be hardened into "never use this interface".
• Two sibling traps recorded along the way: ① the failed-limit pool getTopicZBPool must use dpt=wz.ztzt — using wz.zbgc returns rc=206, and getTopicZBGCPool 404s outright; ② for the same getTopicZBPool, sort=fund:desc returns tc=30 but pool=[] (silent failure), and only sort=fbt:asc returned the full 30 rows. A sort parameter can empty the pool, so the tc==len(pool) check must be run every single time.

II. Interface failures

• East Money push2/clist has now thrown 502 Bad Gateway for a fifth consecutive full trading day (today both clist and ulist.np measured 502). Affected: sector quotes, sector fund flows, whole-market paginated snapshots, advance/decline counts.
• Working substitute routes (same as last Friday and this morning's pre-market piece; recommend hardening them): ① quotes via Tencent qt.gtimg.cn; ② limit-up / limit-down / failed-limit pools via push2ex (mind the sort-parameter trap above); ③ sector quotes and fund flows via ak.stock_board_industry_summary_ths(); ④ index intraday via Sina CN_MarketDataService.getKLineData.
• East Money push2his trends2 intraday interface gave RemoteDisconnected for three consecutive rounds this time (one round with 4 retries and one with 6), so it was switched to Sina 5-minute candles, and the data is normal.
• ak.stock_board_concept_summary_ths() raised KeyError: '涨跌幅' (upstream field change), so this piece uses no concept-sector ranking; instead the leadership was computed from "industry distribution of limit-up names + distribution of first-seal times", which actually carries more information.
• ak.stock_fund_flow_individual('即时') did not return within 120 seconds and was abandoned, so this piece has no stock-level main-force fund flow.
• ak.stock_lhb_detail_em('20260831','20260831') raised TypeError: 'NoneType' object is not subscriptable — this is a scheduling issue (15:30 is earlier than the 18:00 disclosure), not a broken interface; do not raise an alarm.

III. ⚠️ Search contamination (this time all of it is the "same day but different read-time" type)

• When searching for the 8/31 closing wrap, almost everything returned was the 09:30 opening broadcast, saying "the real estate sector exploded, 5i5j sealed limit-up at the open with over RMB 2.5 billion of sealing orders, gold and semiconductors pulled back". That was correct at the open and completely reversed by the close: at the close semiconductors led with +2.36% and net inflow of RMB 8.376 billion, the largest of the whole market, while real estate was −0.48% with net outflow of RMB 3.716 billion, the largest of the whole market. Had the search summaries been copied over, this piece's leadership judgement would have been written entirely backwards. This is a variant of "cross-market data must be stamped with a read-time" within the same market on the same day — an intraday snapshot mistaken for a close conclusion.
• Searching "film & media 8/31 catalyst" returned items published on 8/11, 8/18, 7/29 and 1/29 respectively, with no date in any of the headlines — the same trap as "recurring column headlines hide the publication date". All were discarded, which is why this piece gives no catalyst attribution for media.
• 5i5j's "over RMB 2.5 billion of sealing orders" comes from the opening broadcast; the official pool's sealing capital at the close is RMB 459 million. The two do not contradict each other (intraday sealing orders get consumed), but when quoting it one must state that it is on the close basis, otherwise an order of magnitude is conjured out of thin air.

IV. What this piece did not do (listed honestly)

• No sub-agent cross-verification (timeliness takes priority for a post-close recap, and every conclusion here is a recomputable market fact, not a valuation judgement).
• No first-hand catalyst was obtained for the media leadership; it is explicitly labelled as in doubt in §3.2 with the reason for not attributing it, and summer box office was not used to manufacture causality.

---

0. One-Sentence Recap

Today was a day where "the conclusion at the open and the conclusion at the close were exactly opposite", and the pre-market list got it right where it mattered most and got it wrong where it was most confident. All three major indices gapped down (Shanghai Composite −0.65%, ChiNext Index −1.67%, STAR 50 at one point −2.51%), so the pre-market call of "likely to gap up" was wrong on direction; the August manufacturing PMI released at 09:30 came in at 49.8%, up 0.6pct month-on-month (July 49.2%), clearly better than the "continues to weaken" worry set out pre-market, yet the market ignored it entirely all morning — the Shanghai Composite closed the morning session at −0.20% at 11:30, the real push happened in the afternoon, and it closed at 3,986.30 (+0.86%), at the high of the day, just 0.34% from the 4,000 mark. Sentiment temperature is judged as warming but structurally torn: 88 limit-ups, 11 limit-downs, 30 failed limit-ups, a failed-limit rate of 25.42%, clearly worse than 8/28's 16.33%; the highest streak fell from 8/28's 7 consecutive limit-ups to 6.

The pre-market list has to be scored in three parts, and the conclusions differ a lot. The 5 recommendation slots averaged +0.40% on the 8/28 close basis (underperforming the Shanghai Composite's +0.86%), but +2.74% if executed at today's opening price (outperforming by 1.88pct) — all 5 gapped down or opened roughly flat, a full 2.34pct difference; without that column today's scorecard is simply wrong. At the single-stock level, 国科微 (GOKE Microelectronics, 300672) +5.11% and 伟测科技 (Wise Testing, 688372) +6.25% delivered, 长鑫科技 (ChangXin Memory, 688825) −1.01% and 圣邦股份 (SG Micro, 300661) −0.44% did not, and 万辰集团 (Wanchen Group, 300972) −7.90% was an outright wrong call. Meanwhile the Pass list was today's most valuable output: the 14 Pass names averaged −0.70%, underperforming the Shanghai Composite by 1.56pct, with only 35.7% advancing versus 58.9% for the whole market; among them 香江控股 (Xiangjiang Holding, 600162) opened +10.07% and briefly hit limit-up, then closed −7.28% (open→close −15.76%), 万科 A (China Vanke A, 000002) opened +5.70% and closed −0.63%, 深中华 A (Shenzhen China Bicycle A, 000017) ended its 7-streak at −2.10%, and 千金药业 (Qianjin Pharmaceutical, 600479) −8.85% — the first-hand evidence listed item by item pre-market was borne out item by item by today's price action.

The single most accurate call was the stratification of the property chain. Pre-market it was written explicitly that "the policy is negative for developers as a group; no recommendation slot is given to any developer, and the only link with clean logic is the second-hand housing brokerage". Today's result: 我爱我家 (5i5j Holding, 000560) sealed a one-line limit-up at 09:25 (2 consecutive limit-ups), with sealing capital of RMB 459 million at the close versus turnover of RMB 186 million, a seal/turnover ratio of 2.47 — one of the most solid seal qualities among all 88 limit-up names; while all three developer leaders gapped up and sold off — 保利发展 (Poly Developments, 600048) opened +3.08% and closed −4.24% (open→close −7.10%), 招商蛇口 (China Merchants Shekou, 001979) opened +3.07% and closed −8.80% (open→close −11.52%), 万科 A (000002) opened +5.70% and closed −0.63% (open→close −5.99%); the real estate sector was −0.48% for the day, 25 up / 61 down, with a net outflow of RMB 3.716 billion, the largest in the market. On day 1 of policy pricing, the market's answer matched the pre-market judgement item for item.

The single most wrong call was the internal split within semiconductors. Pre-market §2.1 used Friday's US session ("MU −0.27% and WDC −0.55% held, while NVDA −4.57% and AMAT −4.29% slumped") to make a falsifiable inference: A-share memory would be stronger than compute. Today the A-share answer was the opposite: the compute group averaged +2.69%, the semiconductor design/equipment group +3.61%, while the memory group managed only +0.78%; 浪潮信息 (Inspur Electronic Information, 000977) hit limit-up with turnover of RMB 9.601 billion, the largest in the market, 海光信息 (Hygon Information Technology, 688041) +3.43%, 中科曙光 (Sugon, 603019) +2.85%, while memory leader 长鑫科技 (688825) −1.01% was the weakest name in the memory group and 澜起科技 (Montage Technology, 688008) −3.78%. The sector direction was right (semiconductors +2.36% led, with net inflow of RMB 8.376 billion, the largest in the market), the sub-industry direction was inverted. The assumption that "overnight US sub-industry divergence can be mapped onto the identically named A-share sub-industry" was falsified on the spot today.

And the pre-market piece missed two whole legs, one of which was material sitting in its own document. The first is banks: +1.38% for the day, net inflow of RMB 6.729 billion, second in the market, 38 up / 4 down, and the only sector to gap up in a morning where almost everything gapped down (齐鲁银行 (Qilu Bank, 601665) opened +3.19% and closed +6.68%, 中国银行 (Bank of China, 601988) closed +5.17%, 工商银行 (ICBC, 601398) closed +2.67%). Pre-market, banks were written into the "branches involved" column of the 36 property-credit measures, but no further step of inference was taken — longer development-loan tenors, the lead-bank system, and mortgages pushed to after completion filing all correspond to improving expectations for bank asset quality and duration, and that was the cleanest and earliest-priced leg of this document in the A-share market. The second is media: Film & Cinema +6.50% (20 up / 0 down) and Culture & Media +4.66% (81 up / 1 down) were the top two gainers in the whole market, with a limit-up ladder of 13 or more names, 中文在线 (COL Digital Publishing, 300364) at limit-up with turnover of RMB 3.641 billion, and 华策影视 (Huace Film & TV, 300133) +15.77%. Of the 22 news items pre-market, 0 touched media.

In one line: of the three strongest legs today, the pre-market piece got one right (the broad semiconductor direction + second-hand housing brokerage), got one inverted (memory vs compute), and missed two entirely (banks, media); and what actually salvaged today's scorecard was "no recommendation slot for developers" plus that Pass list. Tomorrow's tone is "the index got through, but the failed-limit rate is deteriorating" — it closed at the high of the day, 0.34% from 4,000, while the core leaders of the afternoon push all sealed only after 13:46, which is grabbing chips, not absorbing supply.

---

1. Market Overview

1.1 Indices and Turnover

Combined turnover of the two markets: RMB 2,131.031 billion (Shanghai 1,014.293 + Shenzhen 1,116.738; the aggregate of the 90 THS industry sectors is 2,145.358 billion, the difference coming from sector-assignment basis, the order of magnitude being the same).

Basis of the turnover column: for the Shanghai Composite, Shenzhen Component and ChiNext Index, Tencent quotes return the turnover of all stocks in that market/board (not of the index constituents), so "Shenzhen Component RMB 1,116.738 billion" should be read as the whole Shenzhen market and "ChiNext Index RMB 528.235 billion" as all ChiNext stocks; the CSI 300 and CSI 500 rows are on a constituent basis and cannot be added to the first three rows.

All six indices "gapped down and closed at the high of the day" — that is the single most important formation fact of today. The STAR 50 went from an intraday low of −2.51% to a close of +1.34%, an intraday range of 3.85pct, making it the amplifier of market-wide sentiment.

1.2 ⚠️ The Push Happened in the Afternoon, Not at the PMI Release

Pre-market, "the August PMI released at 09:30, the same minute as the open" was listed as the largest single unknown variable of the day. The result:

• August manufacturing PMI 49.8%, up 0.6pct month-on-month (July 49.2%), better than the "continues to weaken" scenario set pre-market; the production index at 50.4% and the new orders index at 50.6% (+2.1pct) both returned above the expansion line, equipment manufacturing 51.4%, high-tech manufacturing 52.9% (National Bureau of Statistics / Chinanews¹).
• But the market did not react to it all morning. Shanghai Composite intraday: 09:35 −0.44% → 10:00 −0.36% → 10:30 −0.18% → 11:00 −0.38% → 11:30 −0.20%, underwater the entire morning, unable to turn green in the two hours after the PMI release.
• The turn green and the push all happened in the afternoon: 13:30 +0.29% → 14:00 +0.43% → 14:30 +0.64% → 15:00 +0.86%. The STAR 50 was even more extreme: still −1.61% at 11:30, −0.19% at 13:30, +1.02% at 14:30, closing +1.34%.

How to read this (this one is useful for tomorrow): do not wire "good data + index up" straight into causality. The intraday path shows the market stayed negative for two hours after the PMI landed, and the real driver appeared in the afternoon. Writing today up as "a PMI-beat-driven rebound" would be a self-consistent narrative falsified by its own data. This piece makes no such attribution — the verifiable features of the afternoon push are in §3.3 (an intraday rotation of money from property-policy names into tech/media), and the single piece of news that triggered that rotation could not be sourced verifiably, so it is honestly left blank.

1.3 Limit-Ups/Downs and Sentiment Temperature

Today's limit-up streak distribution: 1 name at 6, 2 at 5, 1 at 4, 4 at 3, 10 at 2, 70 first-time limit-ups.

Sentiment temperature: warming, but structurally torn, not euphoric. Three points supporting "warming": the indices closed at the high of the day, advancers were 58.9%, and 88 limit-ups exceeded last Friday. Three points supporting "torn": ① the failed-limit rate deteriorated from 16.33% to 25.42%, willingness to chase the next board is receding; ② first-time limit-ups were 70/88 = 79.5%, an extremely flat ladder, with only 18 names on consecutive limit-ups; ③ the top streak was downgraded — 深中华 A (000017), the 7-streak name, was −2.10% today and broke. "Index makes a stage high + failed-limit rate deteriorates in step" is a combination to be wary of: money is moving into weights and large-cap tech, not into the relay trade.

---

2. Reconciliation of the Pre-Market List

Note on the baseline (the most important methodological point in this section): the pre-market list was published at 07:00, so the earliest executable price that day is the opening price produced by the 09:25 call auction, not the 8/28 close. The whole market gapped down today, so the two baselines give directionally consistent but very differently sized conclusions, and this section gives both. "Today's chg%" = versus the 8/28 close; "open→close%" = versus today's opening price, i.e. the result of executing the list at the open and closing out at the close.

2.1 The 5 Recommendation Slots (pre-market §9.1, all labelled "watch closely")

Hit rate of the recommendation slots:

⚠️ One-line self-criticism (the methodological point most worth remembering today): on a day when the whole market gaps down, reconciling only against the prior close turns a passing list into a failing one. The opening changes of the 5 recommendation slots were +0.34%, −1.03%, −2.00%, −4.25% and −4.31% — not one gapped up by more than 0.34%; the "do not chase a gap up" risk flagged repeatedly pre-market simply did not materialise today, and the real risk was at the other end — whether you dare execute the list after a gap down. On the prior close it is +0.40%; executed at the opening price it is +2.74%. Tomorrow morning's pre-market list must permanently add a "versus the opening price" column, and state in the body that "executing the list on a gap-down day is cheaper".

2.2 "Watch Only" — 6 Names (pre-market §5, §7)

The "watch only" group averaged −0.54% (3/6 advancing). Split by the pre-market's own stratification, the two categories still separate:

• Category 1 "logic does not hold / poor quality", 4 names: 强一 +3.87%, 商络 +3.54%, 万润 −7.45%, 招商蛇口 −8.80%, average −2.21% — the drag comes entirely from the last two, and those two are precisely the two with the hardest quantitative evidence given pre-market (Q2 sequential −49.8%, ex-non-recurring of only RMB 63 million).
• Category 2 "position too high / already priced", 2 names: 我爱我家 +9.85% (one-line limit-up), 保利发展 −4.24%, average +2.81%.

Memory validated once again: a name Passed on the grounds of "position too high" does not mean it will fall. 我爱我家 (000560) was the best performer of the entire list today, and the pre-market logic on it (the second-hand housing brokerage is the only beneficiary link in the policy that needs no extrapolation) was entirely correct — it was just that "not a good entry" got executed as "do not participate". This is another sample of "chip metrics size the position, they do not set the direction" — the correct handling is to cut the position size, not to remove it from the list.

2.3 The Pass List, 14 Names (pre-market §6)

Pass list statistics (14 names):

One-line self-criticism: the Pass list was the most effective part of today, and the reason it worked is that every line was hung on first-hand evidence. Of the 9 correct calls, 5 rested on evidence the companies themselves wrote in their weekend announcements (深中华 A disclosing no AI revenue, 千金药业 disclosing that consolidation accounted for 22.05%, 锦龙股份 stating the counterparty was not yet determined, 昊华科技 stating it was still at the planning stage, and 香江控股's related-party transaction announcement). "The company negating its own narrative" was the strongest single Pass signal today.

The 4 wrong calls (锦龙, 商络, 金健, 沐曦) share one feature: their Pass reasons were all "poor quality" rather than "narrative negated". The fact of 锦龙股份's restructuring was not negated, only that "the counterparty is not yet determined"; 商络电子's profit is real, it just does not generate cash. In a sentiment market where first-time limit-ups make up 79.5%, "poor quality" is not a same-day reason to sell. This is the same as "financial quality is not a short-term ranker" in memory — tomorrow morning's Pass list should split "narrative negated by first-hand evidence" and "poor financial quality" into two explicit columns, with the latter only downweighted, not removed.

2.4 The 5 "Earnings Blow-Up" Names Named in Pre-Market §8.4

The 5 averaged −2.31%, 3 down and 2 up.

Self-criticism: the rule "an unpre-announced earnings blow-up = day one of pricing = it falls" was only 60% right today, and the two that were wrong were very wrong. 中控技术 (688777) and 泰格医药 (300347) both gapped down then rose in a straight line, a "bad news exhausted" formation rather than "day one of pricing". What the two have in common is that the negative was extremely large and industry-wide (CRO, industrial control), and their share prices had already fallen for a long time beforehand. The three correct calls (禾迈, 中信博, 孚能) belong to the PV/storage chain, sectors still in a downtrend. The fix for tomorrow morning: "earnings blow-up" cannot stand alone as a reason to avoid; it must be layered with a second dimension, "is the sector still in a downtrend". Blow-up + sector in a downtrend = avoid; blow-up + sector already washed out = possibly a contrarian signal.

---

3. Verification of Today's Leadership

3.1 Actual Performance of the Three Pre-Market Branches

⚠️ Detailed falsification of point 1 (the thing most worth recording in this piece):

Pre-market §2.1 used Friday's US session to make an explicitly falsifiable comparative inference: "MU −0.27% and WDC −0.55% — memory held, while NVDA −4.57% and AMAT −4.29% — compute and equipment slumped → mapping SMH −3.47% wholesale onto 长鑫 is wrong; memory will be stronger than compute". Measured in A-shares today:

The inference was inverted, and cleanly so: the compute group beat the memory group by 1.91pct, and the design/equipment group beat the memory group by 2.83pct. The mapping assumption "an overnight US sub-industry holds up → the identically named A-share sub-industry will be strong" does not hold today. One possible explanation (not treated as a conclusion here, only as a hypothesis to verify tomorrow): A-share semiconductors today traded the domestic substitution + interim-report ex-non-recurring endogenous thread rather than the global cycle thread — the best performers 瑞芯微, 国科微, 伟测 and 海光 are all in the domestic design chain, with no correspondence to the identically named US sub-industries. If that hypothesis holds, then the information content of the overnight US session for A-share semiconductors should be marked down wholesale, rather than sliced more finely.

3.2 ⚠️ Unexpected Leadership Missed Entirely Pre-Market, No. 1: Media / Film / Publishing

These were the top two gaining sectors in the whole market today, and 0 of the 22 pre-market news items touched it.

The limit-up ladder (13 names, ordered by first-seal time, showing a clear all-day fermentation path):

Large gainers that did not hit limit-up: 华策影视 (300133) +15.77% (open +0.58%, open→close +15.11%, turnover RMB 1.908 billion), 华智数媒 (Huazhi Digital Media, 300426) +12.74%, 光线传媒 (Enlight Media, 300251) +6.87%, 横店影视 (Hengdian Film, 603103) +4.71%.

⚠️ Catalyst: no verifiable first-hand same-day source was obtained, so this piece makes no attribution. The "media surge" items returned by search were published on 8/11, 8/18, 7/29 and 1/29, with no date in either the headline or the URL — the classic "recurring column headlines hide the publication date" contamination, all discarded. The only public information carrying a same-day date: the 2026 summer box office season closed, with total box office (including pre-sales) as of 8/30 13:00 exceeding RMB 12.34 billion, surpassing the 2025 summer season's final total, a second consecutive year of growth; Kung Fu Women's Football topped the chart with over RMB 2.3 billion (Guangming Online 8/31²). But this piece does not treat that as the attribution, because the formation does not match: all 13 limit-up names opened flat or down (华策影视 opened +0.58%, 中文在线 opened +1.72%, 华智数媒 opened −0.16%), and 100% of the gain was produced intraday. Summer box office is information that was disclosed progressively throughout August, and it cannot explain the "flat open → intraday push → afternoon acceleration" formation. All that is verifiable is the formation itself: a complete all-day ladder starting at 09:34, spreading at 10:25–10:42, and re-accelerating at 13:00–13:22 — day 1 of fermentation, driven by money rather than news.

3.3 ⚠️ Unexpected Leadership Missed Pre-Market, No. 2: Banks — and the Material Was in the Pre-Market's Own Document

This is the least excusable miss, because the material was entirely in the pre-market's own document. Pre-market §1 item 3 listed the "branches involved" of the Opinions on Reforming and Improving Real-Estate Credit Management… (the 36 measures) as "property / banks / construction", but nowhere in the piece was the bank side carried forward one step; §2.2 only covered "developer ROE compression" and "second-hand brokerage benefits". In fact this document's provisions for banks are the densest of all: Article 12 on the lead-bank system plus closed-loop management, Article 13 extending development-loan tenor (up to 7 years for completed-property sales) with first principal repayment pushed to after completion filing, Article 20 allowing mortgages of up to 40 years, Article 22 requiring mortgages to be disbursed only after completion filing, and Article 5 prohibiting loan proceeds from being used to pay land transfer fees. What these provisions jointly point to is that bank exposure to real estate shifts from "lend early, long risk exposure" to "lend late, collateral already built", while asset duration lengthens. That is a combination that improves asset quality and pricing power on both sides at once. "Banks gapping up in a morning when the whole market gapped down" is the hardest piece of evidence — it was priced in the call auction, not chased intraday. The permanent fix for tomorrow's pre-market: every single "branch involved" listed by a policy document must get at least one sentence on the benefit/harm pathway. Listing it without carrying it forward is equivalent to missing it.

3.4 Sector Panorama

Top 10 gainers (THS industry sectors, close basis)

Top 8 decliners

Industry distribution of today's limit-up names (88): General Equipment 7, Auto Parts 6, Real Estate Development 4, Home Furnishings 4, Decoration & Renovation 4, Film & Cinema 4, Digital Media 3, Electronic Components 3, Packaging & Printing 3, Telecom Equipment 3, Publishing 3, Farming 2.

Note one "rose but money is leaving" structure: the Telecom Equipment sector was +1.56% yet saw a net outflow of RMB 3.426 billion (the 2nd largest outflow in the market), while its turnover of RMB 138.369 billion was the 2nd largest in the market. At the same time 星网锐捷 (Star-net Ruijie, 002396), 移远通信 (Quectel, 603236) and 美格智能 (MeiG Smart, 002881) all hit limit-up. High turnover + sector up + large net outflow is a classic distribution structure and needs guarding against tomorrow. The pharma chain was the weakest overall: Chemical Pharmaceuticals −1.47% (31 up, 123 down), Traditional Chinese Medicine −1.74% (9 up, 57 down), Biologics −2.01%, Pharmaceutical Distribution −1.53%. This was the most consistent negative leadership today, and pharma was not mentioned once pre-market (only 泰格医药 was cited as a blow-up case, and it was even up 5.90% today).

---

4. Limit-Up Ladder / Top Streaks

4.1 Full List of the Consecutive-Limit Ladder (18 names)

⚠️ The quality of a top streak must be viewed divided by turnover (memory: seal orders must be divided by turnover). 海鸥住工 (002084)'s 6 boards is today's highest, but turnover was only RMB 40 million and the turnover rate 1.0% — a seal/turnover of 8.31 looks extremely strong, but it is in fact a shrinking-volume one-line board with no real money involved and no ability to drive a sector. The same applies to 新赛股份 (600540, 3 boards, turnover RMB 40 million) and 福建金森 (002679, 2 boards, turnover RMB 59 million). All three are "height without size" lone survivors. The one actually pushed by money is the 4-board 锦龙股份 (000712): turnover of RMB 3.119 billion, turnover rate 27.44%. But its sealing orders were only RMB 52 million, seal/turnover = 0.017, and it broke the limit 3 times intraday. High turnover rate + extremely low seal ratio + repeated failed limits is the formation of relay traders handing stock to each other, not of money locking up. The best seal quality of the day was 我爱我家 (000560): sealing orders of RMB 459 million ÷ turnover of RMB 186 million = 2.47, sealed one-line at 09:25, 0 failed limits all day. It is the only name in the market simultaneously satisfying "has policy logic + one-line seal + zero failed limits + sealing orders significantly larger than turnover".

4.2 Performance Today of the 8/28 Limit-Up Names (82) — Computed Manually

The data in this section was constructed manually. The official getTopicZTPoolPrevious interface returned today's limit-up pool rather than "how yesterday's limit-ups performed today" (exactly the same set as today's 88-name limit-up pool), and was discarded. This section instead: take the 82 names in the 8/28 limit-up pool → pull today's quotes for each → compute manually. Excluding 雪天盐业 (Snowsky Salt, 600929), suspended today, and 花溪科技 (Huaxi Technology, 920895), which has no Tencent quote, the effective sample is 80 names.

Broken-streak list (≥2 consecutive limit-ups on 8/28, no limit-up today):

⚠️ One clear pattern: of the 24 consecutive-limit names from 8/28, the 10 that issued risk-warning / abnormal-fluctuation announcements over the weekend were almost wiped out today — 深中华 A −2.10%, 千金药业 −8.85%, 香江控股 −7.28%, 中安科 −7.44%, 昊华科技 −0.56%, 风范股份 −4.30%, 金辰股份 (Jinchen Machinery, 603396) −3.26%; only 沃特股份 (002886, 3 boards), 海鸥住工 (002084, 6 boards) and 新赛股份 (600540, 3 boards) held up. Pre-market listed "a weekend risk announcement = the most direct coolant for today" as a core judgement, and it holds completely today — one of the most solid correct calls in this piece.

4.3 The Core Leaders (high-turnover limit-up names)

⚠️ This is the point most in need of flagging today: of the top 5 core leaders by turnover, 4 first sealed only after 13:00. 浪潮信息 13:51, 大族激光 14:14, 星网锐捷 13:46, 中文在线 13:00, 红板科技 14:16 — only 瑞芯微 (09:37) sealed in the morning session. Afternoon sealing + seal/turnover generally below 0.06 (浪潮 0.042, 大族 0.052, 星网 0.047, 红板 0.024) means these boards were "grabbed up into the close", not "locked up in the morning and sealed all the way". Such boards have a high probability of gapping up tomorrow, but also the greatest absorption pressure at the open. The control group is 我爱我家 (000560): one-line at 09:25, seal/turnover 2.47, zero failed limits. Both are limit-ups, but their certainty tomorrow is not in the same league.

4.4 The Failed-Limit Pool (30 names)

Top 8 by turnover:

The property chain took 4 of the seats in the failed-limit pool (香江控股, 特发服务, 京投发展, 世联行), all with high turnover. This is consistent with the judgement in §3.1: the property chain today was "spike in the morning, distribute intraday".

---

5. Fund Flows

5.1 Northbound Flows

No data, and this is not a failure on our side. Since 2024-08-19 Stock Connect has stopped disclosing daily net buying, so any interface returning "northbound net inflow of RMB 0.0" is a dead field rather than a real reading, and this piece does not quote it. (The "RMB 42.0 billion" returned by the same interface is the daily quota ceiling, not an inflow figure, and is likewise not quoted.)

5.2 Dragon-Tiger List

No data today — a scheduling reason, not an interface failure. The exchanges disclose the Dragon-Tiger list after 18:00, and this piece was written at 15:30, ahead of disclosure. Last Friday's (8/28) list was already added in this morning's pre-market piece. Tomorrow morning's pre-market piece will add today's (8/31) Dragon-Tiger list.

5.3 Sector Flows (THS basis, close snapshot)

Top 10 net inflows

Top 10 net outflows

Three readings: ① Semiconductors' +RMB 8.376 billion is first in the market, and its turnover of RMB 260.856 billion is also first in the market. Volume and money are both here — this is the least disputable leadership of the day. ② Real estate's net outflow of RMB 3.716 billion is the largest in the market, and yet it led the market at the open today. This is the most direct quantitative evidence of "the conclusion at the open being the opposite of the conclusion at the close" — on day 1 of policy pricing, the answer money gave was outflow. ③ Two "rose but is bleeding money" sectors should be flagged: Telecom Equipment (+1.56% with −RMB 3.426 billion, 2nd largest turnover in the market) and Brokerages (+0.49% with −RMB 1.763 billion). The gains are held up by a handful of limit-up names (星网锐捷, 移远通信, 美格智能 / 锦龙股份), while the sector as a whole is being sold.

5.4 Data Not Obtained (listed honestly)

• Stock-level main-force net inflow: the fetch timed out without returning, so this piece has no stock-level fund flow data, and the "core leaders" judgement in §4 substitutes three recomputable metrics: turnover + seal/turnover ratio + first-seal time.
• Concept-sector ranking: the fetch failed because of an upstream field change, so this piece uses no concept sectors, and the leadership judgement was instead computed from "the industry distribution of the 88 limit-up names + the distribution of first-seal times".
• Today's sector flows come from a single THS basis only, with no second independent source for cross-verification.

---

6. Outlook for Tomorrow

6.1 Judgement on Leadership Continuation

6.2 Key Focus for Tomorrow (branch + stock + verification point)

Write the verification point first, derive the conclusion after (memory: the verification point takes priority over the headline conclusion). Each line below gives a reading that can be adjudicated on the spot during tomorrow's session.

6.3 What to Avoid Tomorrow

1. "Shrinking-volume lone survivors" among the high consecutive-limit names — 海鸥住工 (002084, Shenzhen main board, 6 boards) had turnover of only RMB 40 million and a turnover rate of 1.0%; 新赛股份 (600540, Shanghai main board, 3 boards) turnover RMB 40 million; 福建金森 (002679, Shenzhen main board, 2 boards) turnover RMB 59 million. Height without size and without a sector — once the board opens there is no absorption.
2. Relay names with a high turnover rate and an extremely low seal ratio — 锦龙股份 (000712, Shenzhen main board, 4 boards): turnover rate 27.44%, seal/turnover 0.017, 3 failed limits; 沃特股份 (002886, Shenzhen main board, 3 boards): turnover rate 31.44%, seal/turnover 0.021, 7 failed limits; 捷荣技术 (002855, Shenzhen main board, 5 boards): seal/turnover 0.06, 5 failed limits.
3. Developers as a group — 保利发展 (600048), 招商蛇口 (001979), 万科 A (000002) (Shanghai/Shenzhen main board ±10%). All three gapped up and sold off (open→close −7.10%, −11.52%, −5.99%), and the sector's net outflow of RMB 3.716 billion was the largest in the market. Day 1 of the policy already gave a rejection; do not push the same policy again on day 2.
4. Home furnishings / decoration & renovation in the property chain — of the 13 property-chain limit-up names, 11 first sealed between 09:25 and 09:54, and only 2 did so in the afternoon, both of very poor quality (万通发展 600246 seal/turnover 0.08, 梦百合 603313 failed the limit 5 times). This is the classic "priced once in the morning, no relay in the afternoon".
5. The pharma chain — Chemical Pharmaceuticals (31 up, 123 down), Traditional Chinese Medicine (9 up, 57 down), Biologics, Pharmaceutical Distribution.
6. Precious Metals — −3.44%, worst in the market, 1 up / 13 down.
7. Sectors that "rose but where money is leaving" — Telecom Equipment (+1.56% with a net outflow of RMB 3.426 billion, 2nd largest turnover in the market), Brokerages (+0.49% with a net outflow of RMB 1.763 billion).
8. The property chain within the failed-limit pool — 香江控股 (600162), 特发服务 (300917), 京投发展 (600683), 世联行 (002285). All spiked and distributed today.

6.4 Input Notes for Tomorrow Morning's Pre-Market List

The 6 points below are changes today's scorecard must carry into tomorrow morning's pre-market piece; the first 3 are methodological and the last 3 concern coverage.

1. [Method · must fix] Both the reconciliation and the recommendations need a "versus the opening price" column. Today's recommendation slots were +0.40% on the 8/28 close and +2.74% executed at today's opening price, a difference of 2.34pct; all 5 gapped down while pre-market predicted "likely to gap up". The pre-market piece is published at 07:00, so the executable price is always the next session's opening price, not the prior close.
2. [Method · must fix] Stop mapping "overnight US sub-industry divergence" onto identically named A-share sub-industries. Pre-market §2.1 inferred from MU/WDC holding and NVDA/AMAT slumping that "A-share memory would be stronger than compute", and measured today compute at +2.69% overtook memory at +0.78%, the exact opposite direction. A-share semiconductors today traded the endogenous thread of domestic substitution + interim-report ex-non-recurring numbers. The information content of the overnight US session for A-share semiconductors should be marked down wholesale, rather than sliced more finely.
3. [Method · must fix] "Earnings blow-up = day 1 of pricing = it falls" must be layered with a second dimension, "is the sector still in a downtrend". Today's 5 blow-up names were 3 down and 2 up, 中控技术 (688777) +6.15% and 泰格医药 (300347) +5.90% both gapped down and rose, while the three correct calls were all on the PV/storage chain that is still in a downtrend.
4. [Coverage · must add] Every single "branch involved" listed by a policy document must get at least one sentence on the benefit/harm pathway. Pre-market wrote "banks" into the branches-involved column of the 36 property-credit measures but never carried it forward, and the result was missing the sector with the 2nd largest inflow today and the only one to gap up in a gap-down morning. The material was entirely in its own document.
5. [Coverage · must add] The media/film/publishing line got 0 coverage among the 22 pre-market news items, and it was the top-two gaining sector today. Tomorrow morning needs a dedicated search of weekend and overnight information on the culture & media direction.
6. [Structure · keep] Keep stratifying the Pass list, but change it to three columns. Today the Pass group averaged −0.70% with 35.7% advancing versus 58.9% for the whole market, making it the most effective part of the piece; of the 9 correct calls, 5 rested on evidence of the company negating its own narrative in its own announcement — the strongest single Pass signal, which should get its own column. The 4 wrong calls (锦龙, 商络, 金健, 沐曦) were all reasoned on "poor financial quality", and in a sentiment market where first-time limit-ups make up 79.5%, financial quality is not a same-day reason to sell; it should only be downweighted, not removed.

---

⚠️ Risk disclosure: this recap is a post-close review of information and observations only and does not constitute investment advice. Data may differ in timeliness or basis; please refer to exchange/company disclosures, and do not use this directly as a basis for trading.