A-Share Market Recap | 2026-09-03 Thursday

Scope and caliber boundaries of this report (judge the strength of the conclusions accordingly) ① Index and single-stock quotes are Tencent qt.gtimg.cn closing-settlement caliber, read at 20260903153024–20260903153124, all after the 15:00 close. ② Limit-up / broken-limit-up / limit-down pools are the East Money official pools, 9/3 caliber, and all three pools passed the tc == len(pool) check: limit-up tc=44/len=44, broken limit-up tc=33/len=33, limit-down tc=16/len=16. ⚠️ The limit-down pool returned tc=16 but len(pool)=0 on the first request (a silent failure); switching to sort=fund:asc retrieved the full set, and this report uses the copy that passed the check. ③ Limit-up determination follows per-board price limits: main board ±10%, ChiNext / STAR Market ±20%, BSE ±30%. Every stock in this report is tagged with its board. ④ Sector data is the Tonghuashun 90-industry-sector closing snapshot caliber (net inflow is Tonghuashun model output, different from East Money's "main-force funds" model; this report does not mix it with East Money caliber). ⑤ Seal ratio = sealing-order value ÷ same-day turnover, computed in-house; today's median across the whole pool is 0.1122. ⑥ No northbound flow data (daily net buy disclosure was discontinued from 2024-08-19); this report gives no northbound figures. ⑦ ⚠️ This report did not obtain today's Dragon-Tiger List. The Dragon-Tiger List is published around 18:00 after the close, this report was generated at 15:30, and stock_lhb_detail_em(20260903) returned empty. This is a scheduling problem, not a data-source failure; §5 is left blank accordingly, and yesterday's Dragon-Tiger List is not passed off as today's. ⑧ Every reconciliation gives the "today's open → today's close" column. The executable price in A-shares is the open, not the previous close; looking only at the close-to-close change systematically overstates the hit rate — the single most important reversal in this report (Inner Mongolia First Machinery) is visible only in that column: +1.12% on the close, −7.12% measured from the open. ⑨ ⚠️ The "9/3 closing commentary" retrieved was actually the midday commentary. That piece said "the Shanghai Composite rose 0.43% to 3958.19" and "43 limit-ups, 7 limit-downs", whereas the actual close was 3942.09 (+0.02%), with 44 limit-ups and 16 limit-downs — the index gave back all of its gains in the afternoon and the limit-down count more than doubled. Copying it would have turned a day that weakened into the close into a broad-rally day.

⚠️ Data-retrieval and search log (this section is not sent to clients):

I. Channels that worked

• Limit-up pool stock_zt_pool_em(20260903) 44 rows; broken-limit-up pool stock_zt_pool_zbgc_em 33 rows; all three pools were closed-loop checked with tc == len(pool) against the raw push2ex.eastmoney.com endpoint.
• stock_zt_pool_previous_em(date='20260903') returned 52 rows = the size of the 9/2 limit-up pool, and within it SINOPEC Oilfield Equipment −7.92% and Boyun New Materials +10.00% match, digit for digit, the Tencent quotes I pulled independently for today — confirming that this endpoint returns "9/2 limit-up stocks + their 9/3 performance", not the same day. (The memory note "ZTPoolPrevious returns the same day, not the next day" was ruled out this time by stock-by-stock cross-checking.)
• Sectors via stock_board_industry_summary_ths() (90 rows). The East Money clist series was bypassed via the established backup channel.
• Indices and 24 individual stocks via Tencent; all returned, 0 timeouts.
• Announcements via East Money np-anotice-stock, with the timestamp embedded in art_code re-checked; notice_date was not relied upon.

II. Channels that failed

• ⚠️ stock_lhb_detail_em(20260903) raised TypeError: 'NoneType' object is not subscriptable. This is a scheduling-driven gap caused by publication only at 18:00 while this report is generated at 15:30 — not an endpoint failure. The error signature looks a lot like a dead endpoint; do not misdiagnose it. Suggestion: if the Dragon-Tiger List is to be included in the recap, this slot must be pushed past 18:15, or a supplementary slot must be split off.
• ⚠️ The limit-down pool's first request sort=fbt:asc returned tc=16 / len(pool)=0 — a textbook silent failure (tc non-zero while the pool is empty). The DT pool does not support the fbt (first sealing time) sort field; switching to fund:asc retrieved all 16 rows. This check genuinely blocked an erroneous "no limit-downs today" conclusion this time.
• ak.stock_dt_pool_em does not exist in akshare 1.18.79 (AttributeError); switched to the raw endpoint.

III. Three cross-checks actively performed on methodology that changed the conclusions

1. ⚠️ For the 6 names I passed on that nevertheless surged today, I checked in-window announcements one by one — Boyun New Materials' latest announcement stops at 8/27, Sun Cable at 8/27, Hangzhou Cable at 8/29, confirming there was no new company-level catalyst today and that the moves were purely flow-driven. This check matters: it shows the pre-market judgment "the evidence does not hold" was itself correct; what was wrong was treating "the evidence does not hold" as "it will fall tomorrow". The two are orthogonal, and §2 has been rewritten accordingly.
2. The "9/3 closing commentary" I found was identified as the midday commentary by comparing it with my own closing data (3958.19 vs the actual close of 3942.09; 43 limit-ups / 7 limit-downs vs the actual 44/16). This belongs to the same family as the memory note "searching for the day's closing commentary returns the opening broadcast"; this time it was the midday version. It was not copied, and the discrepancy itself became evidence of the late-day fade in §1.
3. Insurance +2.56% ranked first among the 90 sectors, so I went looking for the driver and found no public catalyst dated 9/3 — searches returned only old pieces from January and July. Separately, the ChinaBond 10Y yield measured 1.6861% on 9/2 and has been range-bound at 1.679–1.699 for the past two weeks, which does not support the common attribution "rising long-end rates drive insurers". Following "no news found ≠ no news", §3 is written as "no confirmable catalyst obtained", not "no catalyst", and it is listed as a must-check item for tomorrow.

IV. What this report did not obtain / did not do (listed honestly)

• Today's Dragon-Tiger List (see above). §5 has only sector fund flows, no seat data, so this report draws no conclusion about the nature of the buyers (institutional vs hot money) for any individual stock.
• Intraday tick data was not obtained; the "late-day plunge" is inferred from two time points — "midday commentary 3958.19 → close 3942.09" and "7 limit-downs at midday → 16 at the close" — and the exact moment of the plunge cannot be given.
• No confirmable catalyst was obtained for either of today's two strong branches, insurance and precious metals.
• The fundamentals of the 44 limit-up stocks were not checked one by one.
• No sub-agents were used (this report is pure quote reconciliation and requires no fundamentals deep-dive).

---

0. One-line recap

Sentiment: flat on the surface, cooling underneath. All four headline indices were essentially flat (Shanghai Composite +0.02%, Shenzhen Component +0.10%, ChiNext +0.01%, STAR 50 −0.40%), but the broken-limit-up rate deteriorated from 22.4% to 42.9%, limit-downs doubled from 8 to 16, and the Shanghai Composite was still +0.43% at midday but only +0.02% at the close — this was a day that faded into the close with a shrinking money-making effect, not a flat day.

Pre-market got the direction of two main lines right and got half of the Pass list wrong — and it was wrong in a very regular way. Defense equipment fell from first place among 90 sectors yesterday (+2.06%, 56 up / 24 down) to −0.59%, 28 up / 53 down today; oil & gas extraction and services −0.97%, 5 up / 14 down. The pre-market "watch only" directional call on these two lines holds. But the 6 names on the Pass list judged to have "no valid catalyst" averaged +6.58% today, with 博云新材 (Boyun New Materials, 002297) and 太阳电缆 (Sun Cable, 002300) both limit-up.

The most important finding in this report is the dividing line that appears once the Pass list is split by reason:

"The company itself said the opposite" is a directional criterion; "I could not find a catalyst" is not. The former went 3 for 3 today, the latter 1 for 6. Lumping these two categories under the same Pass label in the pre-market report is the single thing most in need of fixing today.

The main line missed in the pre-market report was sitting in the pre-market report's own table. §1 item 3b noted "European container freight up more than 9% on 9/2", yet the whole report mapped the geopolitical conflict onto only two legs, defense and oil & gas — it was never mapped to shipping. Today ports & shipping +1.97%, 28 up / 5 down, net inflow RMB 476 million, and 海通发展 (Haitong Development, 603162) limit-up — one of the best branches in the market by breadth.

Tomorrow's tone: main-line vacuum + cooling sentiment; lower expectations rather than switch direction. The limit-up pool's industry distribution is extremely scattered (the largest, auto parts, has only 4 names), and the highest consecutive-limit-up name is 国芳集团 (Guofang Group, 601086, 5 in a row, already having issued three abnormal-movement / risk-warning announcements) in general retail, which belongs to no branch with any industrial logic.

---

1. Market overview

Sentiment indicators (versus 9/2):

⚠️ Sentiment temperature call: the recovery is an illusion, the cooling is real. Three sets of numbers contradict each other and it must be spelled out which is more credible:

1. Advancers/decliners and fund flows really were better than yesterday (1846 up vs 1541 up; net outflow narrowing from RMB 95.964 billion to RMB 42.317 billion).
2. But the odds on continuation deteriorated markedly: a broken-limit-up rate of 42.9% means for every 10 stocks that touch the limit-up price, 4.3 fail to hold it; limit-downs doubled.
3. And the intraday path was downward: at the midday commentary the Shanghai Composite was +0.43% with 43 limit-ups / 7 limit-downs; at the close it was +0.02% with 44 limit-ups / 16 limit-downs. The index gave back all of its gains in the afternoon and the limit-down count rose by 9.

Conclusion: today was a structural day of "index pinned, single-stock divergence widening, late-day fade", and the sentiment temperature is set to "recovery stalled" rather than "recovery". BSE 50 at −1.61% was the weakest index in the market; small-cap high-volatility names are being trimmed.

Yesterday's limit-up stocks (52 names) today — the direct read on the continuation effect:

• 25 up / 27 down, median −0.96%, mean +0.19%
• 9 hit limit-up again today; 4 hit limit-down today
• ⚠️ The mean is positive while the median is negative, meaning a handful of big gainers (博云新材 Boyun New Materials +10.00%, 太阳电缆 Sun Cable +9.96%, 光洋股份 Guangyang Bearing +10.03%, 信达地产 Cinda Real Estate +10.16%) pulled the mean up, while more than half of yesterday's limit-up stocks lost money today. Using the mean to judge continuation odds gets it wrong.

---

2. Reconciliation of the pre-market list

① The 5-name recommendation list (pre-market §9①)

The 5-name recommendation list: average −0.04% on the close, average −1.82% if executed at the open.

⚠️ The gap between these two numbers (1.78 percentage points) is itself the conclusion: looking only at the close-to-close change gives the impression that "the list was roughly flat"; measured at the executable opening price it is a clear loss, and almost the entire difference comes from Inner Mongolia First Machinery alone (+1.12% on the close / −7.12% from the open). Caliber boundary ⑧ requires this column, and today it rescued a self-assessment from distortion.

On direction, all 5 were called correctly: the one "watch closely" name closed green (+1.15%), and of the four "watch only" names three underperformed the market or fell, while the only one that closed green, Offshore Oil Engineering (+0.99%), indeed had no volume and was not worth buying. The greatest value of this list is not that it picked winners, but that it did not push anyone into Inner Mongolia First Machinery at the moment it gapped up 8.87%.

② Pass / do-not-chase list (pre-market §9③)

Sector-level Pass reconciliation:

③ Hit rate and self-critique

Overall hit rate: recommendation list direction 5/5 (100%); Pass list 9/14 (64.3%); combined 14/19 = 73.7%.

⚠️ But the overall hit rate is the least useful number in this reconciliation. The real information appears once the Pass list is split by reason:

Which category was called right:

• Category A went 3 for 3, average −5.56%. SINOPEC Oilfield Equipment wrote "Hormuz disruption" into its own interim report as a drag → −7.92% today; Great Wall Military Industry's military revenue halved by 50.65% and the company itself flagged "elevated valuation" in July → −5.27% today; Zhongman Petroleum had 9 Iraqi drilling crews suspended by that same conflict → −3.49% today. When a stock is being bought for reasons that run opposite to the company's own written disclosure, that can be used as a bearish directional judgment.
• Category C went 5 for 5, average −4.30%. Shenzhen China Bicycle went straight to limit-down and Shengnan Technology fell −7.25%. Regulatory risk and price-level risk are reliable directional criteria.
• On the recommendation side, the most valuable correct call was Inner Mongolia First Machinery. It had the highest seal ratio in the market on 9/2 at 0.326 with 0 broken seals — objectively the best trading structure — and pre-market resisted that temptation and tagged it "watch only"; today it topped at the open, went −7.12% open-to-close, and distributed on 3.01× volume. "Good flow behavior" and "worth buying" really are two different judgments.

Which category was called wrong (this is today's main lesson):

• Category B went 1 for 6, average +6.58%, with two limit-ups. The pre-market checks on these 6 names were not themselves wrong — today I re-checked their announcements one by one: Boyun New Materials' latest announcement stops at 8/27, Sun Cable at 8/27, Hangzhou Cable at 8/29, confirming there was no new company-level catalyst today and that they rose purely on flows. In other words, the pre-market factual judgment that "the catalyst does not hold" was confirmed again today, while the directional judgment built on it was falsified.
• Where the error occurred is perfectly clear: "strength of evidence" was treated as "directional forecast". "I cannot find an in-window catalyst" answers the question "does this move have fundamental support"; it does not answer "will it rise again tomorrow". A theme with no catalyst is perfectly capable of rising for another two days without a catalyst — the cable branch was on day two today, Sun Cable posted 2 consecutive limit-ups, and with a seal ratio of 0.2472 and 0 broken seals it was the second-hardest seal in today's whole pool (median only 0.1122), a far better trading quality than the pre-market impression of "the flimsiest seal".
• Pre-market §8 stated in black and white that "the cable branch will most likely retreat today, the catalyst has been falsified" — that sentence joined "the catalyst is falsified" and "will most likely retreat" with a comma, and the logical leap in between is the entire source of today's loss.

Exactly how to change the pre-market report next time (three items, directly executable):

1. The Pass list must be tagged in two separate columns by reason; they cannot share one label.
   • "Direction falsified" column (company's own account / primary disclosure contradicts the buy thesis, regulatory risk, price level too extended) → retains bearish meaning and can serve as a directional criterion.
   • "Insufficient evidence" column (no in-window catalyst, catalyst fails on back-calculation, label mismatch) → only lowers the recommendation weight; gives no directional judgment. Change the wording from "will most likely retreat" to "the move lacks fundamental support; not participating, but no direction assumed".
2. For branches that are "rising without a catalyst", add a trading-quality dimension before concluding. Today Sun Cable's seal ratio of 0.2472 with 0 broken seals is nothing like Hangzhou Cable's 0.038 on 9/2, and pre-market used Hangzhou Cable's seal quality to stand for the entire cable branch — that is generalizing from a part. Within a branch, the seal ratio must be given stock by stock; the worst single name cannot characterize the whole group.
3. All four transmission channels of a geopolitical event must be worked through; only two cannot be mapped (see the missed-detection part of §3).

---

3. Verification of today's main lines

⚠️ Did pre-market identify the strongest main line? No. Not one of today's three strongest branches was written up pre-market.

⚠️ The missed detection most worth remembering: shipping. The clue was written in the pre-market report's own table.

Pre-market §1 news table, item 3b, verbatim: "Domestic INE crude futures hit an intraday high of +11.35% (closing +8.70%); European container freight up more than 9%, fuel oil up more than 6%, ethylene glycol limit-up".

The figure "European container freight up more than 9%" made it into the report, but it was used only to corroborate "that jump in oil prices" and was never mapped to a single shipping stock. The whole pre-market report listed only two transmission channels for the US–Iran conflict — defense (sentiment) and oil & gas (commodity). Today's outcome: both of those legs ebbed, while the unmapped third leg was one of the best branches in the market by breadth.

A geopolitical conflict transmits into A-shares through at least four channels, and pre-market worked through only two:

Why shipping was the leg that most should have been caught: it has the shortest transmission lag. Pre-market spent extensive space arguing that transmission into oilfield-services equipment takes 4–8 quarters and that the first gate is the human decision of a capex budget — that analysis was correct, and it pointed precisely in the right direction: go find the leg with the shortest transmission lag. Freight rates are priced spot; a conflict can be reflected in shipowners' revenue the same day, without waiting for anyone's budget decision. Pre-market got 90% of the reasoning right and stopped one step short of the end.

⚠️ But a constraint must be attached to shipping too; it cannot be taken wholesale tomorrow: today 中远海控 (COSCO Shipping Holdings, 601919) was only +0.73% and −0.12% open-to-close, and 上港集团 (Shanghai International Port Group, 600018) was +0.18% — the core large caps did not move. What rose were high-beta names like 海通发展 (Haitong Development, 603162, the private-sector dry-bulk leader, seal ratio only 0.0384, 5 broken seals, turnover rate 18.36%). This is hot money trading freight-rate elasticity, not institutions allocating to the shipping sector. See §6② for details.

⚠️ Insurance: today's strongest sector, but no confirmable catalyst obtained

Insurance +2.56%, first of the 90 sectors, 5 up / 0 down with every name advancing, net inflow RMB 2.062 billion, second in the market. 中国太保 China Pacific Insurance +3.69%, 中国人寿 China Life +2.94%, 中国平安 Ping An +2.42%, 中国人保 PICC +1.52%.

I looked for the driver and found no public catalyst dated today — the searches returned only old pieces from January and July 2026 (risk-factor reductions, the draft "Measures for Asset-Liability Management of Insurance Companies" for comment, etc.), not one of them dated 9/3. Moreover, the common attribution "rising long-end rates ease the negative interest spread" does not hold today: the ChinaBond 10-year government bond yield measured 1.6861% on 9/2 and has been range-bound at 1.679–1.699 for the past two weeks, with no upward move.

Handled per "no news found ≠ no news": this report says "no confirmable catalyst obtained", not "no catalyst". Every name advancing + net inflow of RMB 2.062 billion + zero decliners — that kind of breadth does not look like random noise, and there very likely is institutional behavior or policy expectation that I failed to retrieve; it must be a priority follow-up check tomorrow.

---

4. Limit-up cohort / highest rungs

44 limit-ups today, with a highly imbalanced board distribution: 42 on the main board, 2 on ChiNext, 0 on the STAR Market, 0 on the BSE.

⚠️ On ChiNext there were only 金现代 (Jinxiandai Information, 300830, +19.98%) and 思泉新材 (Siquan New Materials, 301489, +20.00%), while the STAR Market and the BSE posted zero. Combined with BSE 50 −1.61% and STAR 50 −0.40%, the high-volatility boards took no part at all in today's limit-up wave, consistent with the pre-market judgment that "macro headwinds suppress growth stocks".

Consecutive limit-up ladder

⚠️ Industry distribution of the limit-up pool — this is the hardest evidence today for judging whether a main line exists:

The most concentrated industry has only 4 limit-ups, and it is "auto parts", a broad classification with no thematic identity. Not a single industry in the market formed a cluster of 5 or more limit-ups — this is a textbook main-line vacuum.

⚠️ And the highest consecutive limit-up name in the market (国芳集团 Guofang Group, 601086, 5 in a row) belongs to "general retail" — not a branch with any industrial logic, and moreover:

• it has already issued three announcements in a row on 9/1, 9/2 and 9/3 — "Announcement of Abnormal Share Price Movement" / "Share Trading Risk Warning Announcement";
• 9/2 was a one-line limit-up (volume of only 58,000 lots), while 9/3 saw volume expand to 584,200 lots (10.1×) with 2 broken seals;
• the general retail sector itself was only 22 up / 45 down today (retail sector +0.17%), with no followers — 百大集团 (Baida Group, 600865) and 东百集团 (Dongbai Group, 600693) also hit limit-up, but the sector as a whole fell.

This is the "lone high rung + no sector follow-through" pattern, typically the late stage of crowded flows.

Core large-cap identification (large turnover + advancing + recognizable)

⚠️ Broken streaks / broken seals:

• 33 broken limit-ups, a broken-limit-up rate of 42.9% (22.4% on 9/2), close to a doubling in deterioration.
• Of the 13 names with ≥2 consecutive limit-ups yesterday, 8 broke the streak today: 英力特 (Yinglite Chemicals, 000635) −6.42%, 欢瑞世纪 (Huanrui Century, 000892) −2.31%, 恒宝股份 (Hengbao, 002104) −3.40%, 小方制药 (Xiaofang Pharmaceutical, 603207) −7.12%, 香溢融通 (Xiangyi Financing, 600830) +0.10%, 九牧王 (Joeone, 601566) +3.09%, 竞业达 (Jingyeda, 003005) +2.65%, 茂业商业 (Maoye Commercial, 600828) +6.03%. Only 集泰股份 Jointas Chemical, 国芳集团 Guofang Group and 龙版传媒 Longban Media, 3 names, advanced a rung.
• The most extreme name: 光洋股份 (Guangyang Bearing, 002708) broke its seal 20 times, finally scraping a limit-up close on a seal ratio of 0.0042 (only RMB 6.04 million of sealing orders against RMB 1.444 billion of turnover) — a seal like that will almost certainly open below the limit tomorrow.

---

5. Fund flows

① Northbound flows

No data. Since 2024-08-19, Stock Connect no longer discloses daily net buy amounts, and this report gives no northbound figures (the endpoint silently returns 0.0, which cannot be relied upon).

② Dragon-Tiger List

⚠️ Not obtained. The Dragon-Tiger List is published around 18:00, this report was generated at 15:30, and stock_lhb_detail_em(20260903) returned empty. This is a scheduling problem, not an endpoint failure.

This report did not obtain today's Dragon-Tiger List data (the exchanges publish it several hours after the close, later than this report's generation time).

⚠️ Therefore this report draws no conclusion whatsoever about the nature of the buyers (institutional seats vs hot-money seats) for any individual stock. Yesterday's (9/2) seat data cannot be used to describe today, and statements in §4 such as "hot money is trading freight-rate elasticity" are all inferences from trading structure — turnover rate, seal ratio, number of broken seals — not seat evidence; discount them accordingly.

③ Main-force fund flows (Tonghuashun 90-industry-sector caliber)

Market-wide aggregate net outflow across the 90 sectors was RMB 42.317 billion (−RMB 95.964 billion on 9/2, an outflow narrowing of 56%), with 18 sectors posting positive net inflow (only 6 on 9/2).

Top 8 by net inflow:

Top 8 by net outflow:

⚠️ Reading the flow structure — one very clear style rotation:

1. The inflow side is entirely "financials + pro-cyclical resources": industrial metals, insurance, securities and precious metals occupy the top four net-inflow slots, totalling +RMB 7.414 billion. What these four share is low valuation + pro-cyclical + geopolitically favoured.
2. The outflow side is entirely "TMT growth": communications equipment, IT services, semiconductors, optics & optoelectronics and software development had a combined net outflow of −RMB 15.362 billion. And the breadth is dreadful — IT services 25 up / 103 down, software development 30 up / 108 down.
3. This is directionally consistent with the pre-market macro call (the 10Y US Treasury yield at a three-year high, CME-implied odds of a September Fed hike at 68% → headwind for growth stocks), and the pre-market §9③⑨ call to avoid "extended tech growth" holds today, with STAR 50 −0.40% the weakest major index.
4. ⚠️ Real estate is the only sector that fell yet saw net inflow (−0.57%, 25 up / 58 down, net inflow +RMB 769 million), with 信达地产 (Cinda Real Estate, 600657, 2 in a row) in the limit-up pool. This kind of divergence is usually bottom-fishing absorption, not a main line; recorded for reference.
5. ⚠️ Within precious metals it is "silver strong, gold weak"; one cannot speak loosely of gold: 白银有色 (Baiyin Nonferrous, 601212) +10.06% limit-up, 湖南白银 (Hunan Silver, 002716) +7.81%, 湖南黄金 (Hunan Gold, 002155) +5.07%, while 山东黄金 (Shandong Gold, 600547) was −1.03% and −2.47% open-to-close, and 紫金矿业 (Zijin Mining, 601899) only +1.09%. The leaders are silver names and names with industrial-metal attributes, not pure gold names.

---

6. Outlook for tomorrow

① Judgment on main-line continuation tomorrow

Core judgment: today was a gap day of "old main lines confirmed to be ebbing + new main line not yet formed"; tomorrow will most likely still have no strong main line, and expectations should be lowered rather than direction hastily switched.

Three grounds:

1. The old main lines (defense, oil & gas) are confirmed to be ebbing, and they ebbed cleanly. Defense equipment fell from 1st of the 90 sectors (+2.06%) to −0.59%, 28 up / 53 down; oil & gas extraction and services −0.97%, 5 up / 14 down, 0 limit-ups. Do not catch a bounce in these two lines tomorrow.
2. The new directions (insurance, precious metals, shipping, industrial metals) have only one day of data and all have obvious defects: insurance has no confirmable catalyst obtained; precious metals has an incomplete "silver strong, gold weak" structure; in shipping the core large caps did not move (COSCO Shipping Holdings +0.73%); and industrial metals' leader Baiyin Nonferrous had a seal ratio of only 0.0220. None of the four yet qualifies as a main line.
3. The sentiment indicators are deteriorating, not improving: broken-limit-up rate 42.9% (22.4% yesterday), 16 limit-downs (8 yesterday), yesterday's limit-up stocks at a median of −0.96% today, the most concentrated industry in the limit-up pool at only 4 names, and the highest consecutive limit-up name a lone general-retail name with no industrial logic. The odds on continuation have clearly worsened.

② Key focus for tomorrow

③ Avoid tomorrow

1. ⚠️ 国芳集团 (Guofang Group, 601086, SSE main board, 5 in a row) — a lone high rung. It has already issued three abnormal-movement / risk-warning announcements in a row (9/1, 9/2, 9/3); on 9/3 it went from a one-line limit-up to 10.1× volume with 2 broken seals; the general retail sector itself was 22 up / 45 down, with no follow-through at all. A textbook late-stage crowding pattern.
2. ⚠️ 博云新材 (Boyun New Materials, 002297, SZSE main board, 2 in a row) — although it hit limit-up today, its seal ratio of 0.0284 (1/4 of the 0.1122 median), 5 broken seals and 31.38% turnover rate make it the worst seal quality in the 2-in-a-row rung. The pre-market fundamental judgment on it — "defense label mismatch, Q2 QoQ −71.2%, low PE + high PB signalling a cycle top" — was not overturned today; it simply was not adopted by today's price.
3. ⚠️ 光洋股份 (Guangyang Bearing, 002708, SZSE main board, 2 in a row) — 20 broken seals, seal ratio 0.0042 (only RMB 6.04 million of sealing orders against RMB 1.444 billion of turnover). A seal like that will almost certainly open below the limit tomorrow.
4. Cable (太阳电缆 Sun Cable 002300 / 远东股份 Far East Holding 600869 / 杭电股份 Hangzhou Cable 603618) — this is not a bearish call, it is non-participation. Today was already day 2 of fermentation, Far East Holding's turnover rate was 16.72% and Hangzhou Cable's 24.33%, extremely heavy speculation; and I re-checked all three companies' announcements one by one today — none of them has any in-window catalyst. ⚠️ But per today's lesson, "no catalyst" only lowers willingness to participate; it does not constitute a bearish direction — Sun Cable's seal ratio of 0.2472 with 0 broken seals is objectively a good trading structure. If they keep rising tomorrow, do not repeat today's mistaken phrasing of "will most likely retreat".
5. Defense and the whole oil & gas chain — confirmed to be ebbing, with 1 limit-up (and that is the label-mismatched Boyun New Materials) and 0 respectively. Do not catch the bounce.
6. TMT growth (communications equipment / IT services / semiconductors / software development / optics & optoelectronics) — these five sectors had a combined net outflow of RMB 15.362 billion today with dreadful breadth (IT services 25 up / 103 down, software development 30 up / 108 down), compounded by expectations of a September Fed hike. The pre-market macro-headwind call continues to hold.
7. Continuation across yesterday's limit-up stocks as a group — of the 52 names, 25 up / 27 down today, median −0.96%, 8 broke their streaks and only 3 advanced a rung, with a broken-limit-up rate of 42.9%. Continuation odds are poor; do not chase limit-ups indiscriminately tomorrow.

④ Input notes for tomorrow morning's pre-market list

This is today's most important output, and all three items are directly executable:

1. ⚠️ The Pass list must be split into two columns by reason; they cannot share one label. Today's results after splitting are decisive: "company's own disclosure contradicts the buy thesis" 3/3 all correct, average −5.56%; "regulatory / extended level / macro" 5/5 all correct, average −4.30%; while "no in-window catalyst / catalyst fails on back-calculation" was right on only 1 of 6, average +6.58%, with two limit-ups.

   • "Direction falsified" column → retains bearish meaning and can serve as a directional criterion.
   • "Insufficient evidence" column → only lowers the recommendation weight; gives no directional judgment. Change the wording from "will most likely retreat" to "the move lacks fundamental support; not participating, but no direction assumed". The criterion is simple: "this stock is being bought for reasons opposite to what it wrote itself" is directional information; "I could not find why it rose" is only a gap in my information, not a weakness in the stock.

2. ⚠️ Geopolitical / macro events must have their transmission channels listed in full and then scored one by one; mapping only the two most conspicuous is not enough. Today defense (sentiment) and oil & gas (commodity) both ebbed, while shipping (freight rates, shortest lag) and precious metals (safe haven) were the two that actually delivered — and the clue "European container freight up more than 9%" was written in item 3b of the pre-market report's own news table, it just was never mapped to a stock. Add a mandatory question to tomorrow morning's template: "How many transmission channels does this event have? What is the lag on each? Which stocks correspond to the shortest-lag channel?" Pre-market had already argued that oilfield-services equipment has a lag of 4–8 quarters; the next step should have been to ask "then who has the shortest lag" — that step was not completed today.

3. ⚠️ The industry distribution of the limit-up pool must be treated as an independent "is there a main line" test, not merely used to count names within the target branch. Today's 44 limit-ups were scattered across 27 industries, with the most concentrated holding only 4; the market's highest consecutive limit-up name (国芳集团 Guofang Group, 5 in a row) belongs to "general retail" and is in none of the branches under discussion. Mandatory question tomorrow morning: "Which industry does the highest consecutive limit-up name in yesterday's limit-up pool belong to? What were that industry's sector move and advancer/decliner counts yesterday?" If the sector to which the highest rung belongs is itself falling (today general retail was 22 up / 45 down), that high rung should be flagged outright as a lone name to avoid, however hard its seal.

4. One additional scheduling suggestion: if the recap is to include Dragon-Tiger List seat data, this slot must be pushed past 18:15; the current 15:30 generation time cannot possibly retrieve it, and §5② will keep coming up blank.

---

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

Data sources: East Money official limit-up / broken-limit-up / limit-down pools (push2ex.eastmoney.com, all checked with tc == len(pool)), Tonghuashun 90-industry-sector closing snapshot, Tencent Finance qt.gtimg.cn closing quotes (read at 15:30–15:31), the East Money company announcement endpoint (np-anotice-stock, with art_code timestamps re-checked), and the ChinaBond government bond yield curve. Today's Dragon-Tiger List is not included because its publication time is later than this report's generation time.