A-Share Market Recap | 2026-09-10 Thursday

Scope boundaries of this report (please gauge the strength of the conclusions accordingly) ① All quotes, sector and single-stock fund flows are on a 2026-09-10 closing basis, read at 15:30–15:40 (Beijing). ② Limit-up / failed-limit-up / limit-down counts come from the official East Money pools (9/10: limit-up 35, failed limit-up 22, limit-down 11). Limit-up determination is tiered by board: Shanghai/Shenzhen main board ±10%, ChiNext/STAR ±20%, Beijing Stock Exchange ±30%. Every stock in this report is tagged with its board. ③ ⚠️ Today's Dragon-Tiger List was not obtained in this report. The Shanghai/Shenzhen exchanges publish the Dragon-Tiger List after 18:00 following the close, and this report was generated at 15:40 — this is a scheduling matter, not a failure. The Dragon-Tiger section of this report is left blank; yesterday's data is not passed off as today's. ④ ⚠️ No northbound flow data (daily net buying has not been disclosed since 2024-08-19); this report gives no northbound figures. ⑤ Advance/decline counts and the median move are computed in-house from a stock-by-stock snapshot of all 5,912 listed names (5,551 with valid quotes, 361 suspended / no quote), with all pagination exhausted and cross-checked against the interface's declared total; truncated sorted-pagination results are not used. ⑥ "Open→Close %" is based on today's open — this morning's pre-market list was published before the open, so the earliest executable price for a reader is today's open, and this column is the executable basis; "Chg %" is versus yesterday's close. Both columns are given side by side. ⑦ ⚠️ This report marks the attribution of 3 limit-up stocks as "not established", tagged stock by stock in §4, with no "close enough" old news used as filler.

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

I. Channels that failed

• ⚠️ akshare's East Money cluster returned RemoteDisconnected for the ninth consecutive trading day: stock_zh_index_spot_em, stock_board_industry_name_em, index_zh_a_hist all threw ConnectionError. Switched to direct calls to push2delay.eastmoney.com's /api/qt/clist/get and /api/qt/stock/get, which worked on the first try.
• ⚠️ push2delay .../api/qt/stock/kline/get returns klines: [] (this delayed host does not serve K-lines), so the multi-day index turnover comparison was not obtained. Fell back to the share-volume series from Sina's stock_zh_index_daily for a volume-shrinkage comparison (Shanghai Composite today 484,675,114 lots = 48.468 billion shares, versus 51.750 billion shares on 9/9, −6.3%).
• ⚠️ The columns of ak.futures_zh_realtime(symbol='原油') are settlement/presettlement/preclose, not settle — reusing the old field name throws KeyError. symbol='铜' does not exist; the correct form is 沪铜.
• ⚠️ ak.stock_lhb_detail_em(20260910) throws TypeError: 'NoneType' object is not subscriptable — this is what "not yet published today" looks like, not an interface failure (the same call with 20260909 returns 66 rows normally). Third hit for memory [[lhb-publishes-after-1800]].

II. Channels that worked

• All three pools (limit-up / failed limit-up / limit-down) succeeded; stock_zt_pool_previous_em succeeded.
• The full-market stock-by-stock snapshot of 5,912 names was pulled across 30 pages, with total cross-checked as consistent.
• 496 industry sectors and 504 concept sectors, with all three fund-flow bases (today / 5-day / 10-day) obtained.
• Announcements were pulled via np-anotice-stock, with a stock-by-stock trace back over the 14 key limit-up names of today.

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

1. ⚠️⚠️ The basis of stock_zt_pool_previous_em(date='20260910') was verified on the spot by cross-referencing individual stocks: it returns 48 rows = the 9/9 limit-up count, while the price changes are today's (9/10) readings. Cross-validated with 渝三峡Ａ (Chongqing Three Gorges A, 000565) — the stock shows the identical RMB 7.91 / +10.01% in both today's limit-up pool and the previous pool, so the two pools close. The misuse direction flagged by memory [[zt-pool-previous-returns-same-day]] (treating date=D as "how D's limit-up stocks performed on D+1") does not apply here, but it must be verified live every time and never assumed from memory.

2. ⚠️⚠️ The limit-down count went from 12 to 11 within 7 minutes after the close. The first reading at 15:30 was 12 names (including 深中华 A / Shen Zhonghua A, 000017); re-reads at 15:37 and 15:40 both gave 11. This report uses 11, with the read time noted. Had it been read only once and written straight into the body, the sentiment gauge would be systematically one notch too high.

3. ⚠️⚠️ Every announcement from today's limit-up stocks is old news; not one falls inside the news window. Tracing eiTime stock by stock across the 14 key limit-up names: 太极实业 (Taiji Industry) latest announcement 8/26, 金安国纪 (Jin'an Guoji) 8/27, 超声电子 (Ultrasonic Electronics) 8/21, 华银电力 (Huayin Power) 8/31, 闽东电力 (Mindong Electric Power) 9/7, 湖南发展 (Hunan Development) 9/3, 水发燃气 (Shuifa Gas) 9/2, 九鼎新材 (Jiuding New Material) 9/3, 新力金融 (Xinli Finance) 9/8, 津药药业 (Jinyao Pharmaceutical) 9/8… without a single exception. Conclusion: today was a pure liquidity/theme-rotation day, not an announcement-driven day — this judgment rewrote the entire attribution framework of §3. Following the usual habit of "matching each limit-up stock with a piece of good news" would inevitably stitch in old news (memory [[ai-attribution-stitches-old-news]]).

4. ⚠️⚠️ The Jiemian News piece at 13:55, "New-energy generation capacity in the North China grid tops 200 million kilowatts", is not the catalyst for today's power stocks; it is after-the-fact narrative. Verified: the event described took place on 9/3, the article was published 9/10 13:55, whereas 华银电力 (Huayin Power, 600744) was already sealed at 10:09 and 闽东电力 (Mindong Electric Power, 000993) was already sealed at 09:33. An ETF-promotion piece published at 13:55 about a 9/3 event cannot explain seals that happened before 10:09. Discarded.

5. ⚠️ NetEase's "A-share close: Shanghai Composite up 0.28%" and "Close | Shanghai Composite +0.28%, power sector strengthens" match the search intent closely by headline, but neither is from today — the Shanghai Composite is −0.43% today. Memories [[recurring-column-headlines-hide-date]] and [[opening-snapshot-sold-as-close-recap]] both hit. All advance/decline counts, turnover and sector moves in this report are computed in-house, with zero citations from close-review summaries.

6. ⚠️ The contamination of sector fund flows by a new listing has been quantified and stripped out: N 电科思 (Dianke Si, 301689, ChiNext) debuted today at +237.50%, contributing +1.298 billion of net main-force inflow on its own, booked under "General Equipment / Machinery". Machinery shows a headline −3.049 billion, but after stripping it out the true figure is −4.347 billion, a 43% difference in magnitude. Memory [[new-listing-pollutes-sector-fund-flow]].

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

• Today's Dragon-Tiger List was not obtained (published after 18:00); the Dragon-Tiger section of §5 is left blank.
• The seat character of today's top net-inflow stocks cannot be determined (it depends on the Dragon-Tiger List).
• The multi-day index turnover series was not obtained (see above); the volume-shrinkage conclusion rests only on share volume.
• The attribution of three large limit-up names — 太极实业 (Taiji Industry, 600667), 环旭电子 (Universal Scientific, 601231), 再升科技 (Reetec, 603601) — was not established, explicitly tagged in §4, with no old news used as filler.
• No primary 9/9–9/10 price-increase letter was obtained for the current CCL / electronic-grade fiberglass cloth price cycle; only 富乔工业 (Fuqiao Industrial)'s July price notice and August industry reporting were retrieved, so §3 characterizes it as an ongoing price-hike cycle, not a fresh catalyst today.

---

0. One-Sentence Recap

**Sentiment is "day 3 of the ebb" rather than a turning point: the limit-up count stepped down for a 4th straight day (93 → 73 → 48 → 35), the failed-limit-up rate rose to 38.6%, limit-downs went from 7 to 11, the full market was 955 up versus 4,512 down, the median was −1.49%, yet the indices fell only 0.43%–0.93% — this is a defensive rotation of "banks, brokers and utilities holding up the index while individual stocks broadly fall". Of the three branches on this morning's list, only branch 2 (CCL/PCB) was called right, and even there the branch was right while the wrong end of the chain was picked — what actually surged was the upstream electronic-grade fiberglass cloth and CCL (金安国纪 / Jin'an Guoji +10.00%, 超声电子 / Ultrasonic Electronics +10.00%, 逸豪新材 / Yihao New Material +15.29%), not the downstream PCB manufacturing we recommended. Branch 1 (oil shipping, rated S) failed to rise at all even as oil jumped 3.71%, and branch 3 (power grid) was −1.54% on the day with multi-day fund flows already negative. Tone for tomorrow: the upstream materials chain takes the baton, power is on day 2 of build-up, and keep avoiding indiscriminate follow-on buying of yesterday's limit-up stocks.

---

1. Market Overview

1.1 Indices (15:30 closing basis)

• Combined Shanghai + Shenzhen turnover RMB 1,647.1 billion; summing the full market stock by stock (including the BSE) gives RMB 1,662.4 billion. Shanghai volume was 48.468 billion shares versus 51.750 billion shares on 9/9, −6.3%, a modest contraction.
• ⚠️ SSE 50 −0.21% most resilient, BSE 50 −2.81% weakest, CSI 1000 −0.80% — the large-cap/small-cap divergence is unambiguous, and this is the first piece of evidence for characterizing today.

1.2 Advance/Decline and Sentiment Gauges (computed in-house across the full market, n=5,551 valid quotes)

Distribution detail: up more than 9% 47 names, up more than 5% 102 names, up more than 3% 193 names; down more than 3% 889 names, down more than 5% 208 names, down more than 7% 76 names, down more than 9% 33 names.

1.3 Sentiment Temperature: day 3 of the ebb, neither a freeze nor a recovery

Four pieces of evidence point the same way:

1. The limit-up count stepped down for a 4th consecutive trading day: 93 → 73 → 48 → 35, with no sign of the decline converging.
2. The failed-limit-up rate rose from 36.0% to 38.6% and limit-downs from 7 to 11 — selling pressure above and panic below are strengthening at the same time.
3. The follow-on money effect is still negative: the 48 limit-up stocks from 9/9 posted a median −1.04% today with only 22 of them up (45.8%); by comparison, the 73 limit-up stocks from 9/8 posted a median −0.60% with 49.3% up on 9/9. Both metrics are worse than yesterday's.
4. The "yesterday's cohort" concept sectors are all at the bottom: yesterday's failed limit-ups −3.17%, yesterday's touched-limit names −3.30%, yesterday's high-turnover names −3.76% (net main-force outflow 4.489 billion).

⚠️ But it must be said just as clearly that this is not a freeze: index declines are mild (Shanghai Composite −0.43%), volume contracted only slightly, and there is a clear direction in which money is being absorbed (Printed Circuit Boards +3.017 billion, Utilities +2.031 billion, Glass & Fiberglass +1.676 billion, Brokers +1.227 billion). This is a structural rotation, not a broad bleed-out — money moved out of elevated themes into the materials price-hike chain, utilities and low-valuation financials, rather than out of the market.

---

2. Reconciliation of the Pre-Market List

Basis: this morning's list was published before the open, so the earliest executable price for a reader is today's open; therefore "Open→Close %" is the executable return, while "Chg %" is the paper return versus yesterday's close. Both columns are given, with Open→Close as the basis for judging delivery.

2.1 Recommended Slots (4 names; the 5th slot was publicly left blank this morning)

Recommended-group statistics

⚠️ This result must be read from both sides; picking only one side will not do. The good side: on a day with 4,512 decliners and a −1.49% median, the recommended group was −0.63% and 3 of 4 beat the median, so the list did provide relative protection. The bad side: on an executable basis (Open→Close) it was −0.92%, with 3 of the 4 losing money. "Beating the median" is not the same as "this trade made money". The first duty of a pre-market list is to give a direction that makes money, not a direction that falls less than others. Today it failed to do that.

2.2 The "Watch Only / Observe" Group — this group was judged better than the recommended group today

2.3 Pass-List Reconciliation — this is the most successful part of today

Pass-group statistics: of the 31 individually verifiable names above, 24 were called correctly, 5 were called wrong and 2 were neutral, an accuracy of roughly 77%; among them two went straight to limit-down and one gapped up 8.8% then posted an Open→Close of −10.89%. The value of the Pass list today was markedly higher than that of the recommended list.

2.4 ⚠️ Hit Rate and Self-Criticism — three wrong calls, each wrong in a different place

Overall hit rate: recommended slots 1/4 delivered (25%); Pass group 24/31 correct (77%); watch-only group 11/14 correct (79%). One-sentence self-criticism: today's list "excluded very well and selected very badly". The three wrong calls each have their own pathology and must not be lumped together.

---

① The branch was right, the end of the chain was wrong — this is the most valuable item

CCL/PCB was branch 2 on this morning's list, and today it was fully confirmed: Printed Circuit Boards +1.95%, net main-force inflow 3.017 billion (1st among all industries), 33 up 15 down, 5-day cumulative +18.283 billion (1st in the whole market), 10-day +6.527 billion (1st in the whole market). There is nothing wrong with the branch judgment.

The problem is: the money went up the upstream of this chain, while both of our recommended slots sat downstream.

⚠️ The lesson must be stated more precisely than "picked the wrong stock":

• The reason 崇达技术 (Chongda Technology) was given a recommended slot this morning was "PB percentile 42.7% + net cash 1.39 billion + buyback price cap RMB 19.27" — all balance-sheet variables. This morning's own text said "the reason to buy it is not AI". Not one of those arguments has been falsified (it was indeed the only recommended slot to close positive today), but what they answer is "will this stock break down", not "which stock will rise today". Memories [[financial-quality-is-not-a-short-term-ranker]] and [[position-metrics-size-not-direction]] both hit: chip-distribution and asset-quality metrics set position size and the floor, not the ranking for the day.
• The real ranking variable today was "how close you are to the price-increase letter". The current driver is the price-hike cycle running from electronic cloth to CCL (7628 electronic cloth rose from RMB 3.8/meter in September 2025 to about RMB 10/meter in August, more than 165% year on year; 富乔工业 (Fuqiao Industrial) raised standard E-glass by 30% and LowDK2 by 15% effective 7/1), and the order of price pass-through runs from upstream to downstream — PCB manufacturers on this chain are the ones bearing the cost, not the beneficiaries. We drew the branch boundary at "PCB", while the money drew its boundary at "the price-increase letter".
• ⚠️ The most glaring item: the 华正新材 (Huazheng New Material) we called "watch only" (+2.80%, net inflow 419 million) beat the 崇达技术 (Chongda Technology) we recommended (+1.05%, net outflow 36 million). The reason for excluding it this morning was "PB 13.50 at the 98.4th percentile + 0 institutional seats on the Dragon-Tiger List" — position and the character of the buying. Both are invalid rankers during a theme's launch/acceleration phase, which is exactly what memory [[position-metrics-size-not-direction]] says, and this morning it was used as a binary switch.
• ⚠️ But the counter-evidence must be recorded at the same time, or tomorrow will over-correct: the leader 生益科技 (Shengyi Technology) was −0.81% with net main-force outflow of 499 million, and 东山精密 (Dongshan Precision) was −1.18% with net outflow of 363 million. All the leaders of the advance were second-tier (金安国纪 / Jin'an Guoji, 超声电子 / Ultrasonic Electronics, 四会富仕 / Sihui Fushi +9.69%, 明阳电路 / Mingyang Circuit +8.68%, 博敏电子 / Bomin Electronics +4.52%). "Leaders not following while second-tier names rise broadly" is a signature of capital rotation, not confirmation of industry prosperity — this feeds directly into §6's judgment for tomorrow.

---

② The verification point passed with a huge margin and the trade still lost money — because it was not testing the variable that loses money

This morning's §8 verification point 1 read: "If WTI falls back below 93.03, or Shanghai crude SC2610 breaks below 741.5 (9/9 close) in the day session, then the overnight premium has been erased — do not chase."

Actual readings today: SC2610 closed at 769.0, up +3.71% from the 9/9 close of 741.5; the full-day low of 752.4 was also above the threshold; the intraday high was 791.6 (+6.76%); Brent was around USD 101.29 per barrel. The commodity leg passed with a margin of +3.71% and never broke below for even a second.

And yet oil-shipping stocks did not rise at all: 中远海能 (COSCO Shipping Energy) −0.61%, 招商轮船 (China Merchants Energy Shipping) −1.95%, 招商南油 (China Merchants Nanjing Tanker) +0.42% (Open→Close −1.02%), 海通发展 (Haitong Development) +0.58% (Open→Close −4.96%), 南京港 (Nanjing Port) +2.81% (Open→Close −6.56%). Of the 35 names in the Shipping & Ports sector, 14 rose and 20 fell.

⚠️ Verification point 2 in the same report, however, was right: "How many of 9/9's 4 first-time limit-ups reach a 2nd consecutive limit-up today? If there are 0 consecutive limit-ups before 11:30, judge it an event-driven single-day pulse." The oil-shipping branch had 0 limit-ups today, and 南京港 (Nanjing Port) and 海通发展 (Haitong Development) both touched the limit intraday and then broke it, both appearing in the official failed-limit-up pool. It predicted the failure precisely, and it was the only one that predicted correctly.

The correct statement of this lesson is not "verification points are useless", but:

The commodity price is a necessary condition for the branch to hold, not a sufficient one; and it is the leg that "was already priced yesterday". When 4 oil-shipping names hit the limit on 9/9, oil had already risen; today oil rose another 3.71%, and the market did not pay a second time for the same piece of information. The variable that was genuinely unpriced, and that genuinely determined the P&L, was whether the branch could carry itself forward (verification point 2). Memories [[verification-point-must-test-the-loss-driver]] and [[repriced-is-not-realized]] both hit on the same stock.

The actionable fix: the seven verification points should carry weights and priorities. This morning all seven were listed side by side, leaving the reader no way to know which one to follow when "the commodity price passes" conflicts with "0 consecutive limit-ups in the branch". The next pre-market report must state explicitly: structural verification points (consecutive limit-ups, anchor leaders, multi-day fund flows) take precedence over exogenous price-type verification points.

---

③ The scoring discipline I wrote down myself was applied only to the branches I did not want to recommend, not to the one I most wanted to recommend

Item 2 of this morning's §4 "four scoring disciplines specific to this report" states in black and white: "Single-day sector fund flow may not be extrapolated to the next day's direction; any extrapolation must use a 3–5 day cumulative basis. 'Theme persistence' is therefore scored on 5-day/10-day cumulative flows." And this discipline was indeed executed twice — it pushed optical fiber from 2nd down to 4th and lifted PCB from 4th to 2nd, and both calls were right (optical fiber fell across the board today, and PCB was the top fund destination in the whole market today).

But it was not applied to branch 3 (Power Grid Equipment).

⚠️ Power Grid Equipment's 5-day and 10-day cumulative fund flows are both negative, and yet this morning it was given branch 3 and a recommended slot (汉缆股份 / Hanhe Cable). What supported it was four tender-win announcements — that is single-day, event-type evidence, precisely the class that discipline item 2 requires to be discounted. Had the same yardstick been applied to the power grid, 汉缆股份 (Hanhe Cable) should not have received a recommended slot. Today it was −3.39%, the worst of the recommended group.

This is a verbatim replay of memory [[checklist-skipped-on-favored-branch]]: the criterion gets silently skipped on the branch one most wants to recommend. The fix is concrete: make "5-day/10-day cumulative fund flow" a precondition for admission to a recommended slot, write it into the template and fill in the table branch by branch, and bar any branch that is negative on both from occupying a recommended slot — make it a cell that must be filled in, not a principle used "when one happens to remember it".

---

④ The announcement was read correctly, the reaction function was not — the wrong Pass call on 深科技 (Shenzhen Kaifa)

This morning's announcement check on 深科技 (Shenzhen Kaifa, 000021, Shenzhen main board) was correct item by item: of the 1.85 billion, 1.22 billion is construction/fit-out/power/land with zero equipment, the 630 million R&D line is described by the company itself as "not generating direct economic benefit", both are due for completion in December 2028, the payback period is 12.01 years, and of the previous two rounds totaling 2.21 billion only 26 million was in construction-in-progress as of 6/30. Not one of these facts was overturned today.

But it rose +4.32% today (Open→Close +2.44%), with net main-force inflow of 193 million and a volume ratio of 2.20. Its memory-packaging peer 太极实业 (Taiji Industry, 600667, Shanghai main board) hit the limit at +10.02% and was the largest net main-force inflow in the whole market (1.891 billion, net share 50.52%); the memory-chip concept saw net inflow of 2.022 billion today.

⚠️ Lesson: "the announcement does not survive a reading of the original text" answers how much this piece of information is worth, not how much the market is willing to pay for it today. Memory [[direction-and-holding-value-are-orthogonal]]: directional judgment and holding value are two orthogonal conclusions. The Pass label is currently carrying two meanings at once — "insufficient evidence" and "expected to fall" — and today those two came apart. Suggestion: add a "Pass type" column to the Pass list, distinguishing "the logic does not hold" from "position/valuation too high"; the former only means we do not endorse it, not that we are bearish. The statistics in memory [[position-metrics-size-not-direction]] (the 7 "logic does not hold" names at −0.21%, the 7 "position too high" names at +6.18% with 7/7 outperforming) have already quantified this distinction.

---

⑤ The main line that was missed: power/green power, 5 limit-ups, and the clue was in the data I pulled myself this morning

See §3.4. This morning's list had zero coverage of it — not one power stock appeared in the branch table, the 21-name master list or the Pass list.

---

3. Verification of Today's Main Lines

3.1 Master Reconciliation Table by Branch

3.2 ⚠️ Did the Pre-Market Report Call the Strongest Main Line? — the branch was right, the end was wrong

Answer: yes and no.

Yes: this morning listed CCL/PCB/Components as branch 2, and today it was the strongest fund destination in the whole market (Printed Circuit Boards net inflow 3.017 billion, 1st among industries; 5-day 18.283 billion and 10-day 6.527 billion, both 1st in the whole market). And the reasoning given this morning was honest — it explicitly wrote that "what supports ranking this branch 2nd is not the AI narrative but the persistence of fund flows and the valuation percentile", with no fabricated catalyst.

No: the leading end of this chain today was upstream, while the list landed downstream (tabulated in §2.4①). Putting fiberglass/electronic cloth (⑦) alongside CCL, 3 of the top 5 industry gainers today belong to the same value chain: Fiberglass Manufacturing +3.77%, Glass & Fiberglass +2.53%, Printed Circuit Boards +1.95%. These are not three main lines but one — the electronic cloth → CCL → PCB price-transmission chain, which today drove the first two links.

⚠️ The characterization of the driver must be stated clearly and not inflated: this is an ongoing price-hike cycle, not a fresh catalyst today. The primary time stamps verified are 富乔工业 (Fuqiao Industrial)'s price-increase notice issued 6/30 and effective 7/1 (standard E-glass +30%, LowDK2 +15%), plus 7628 electronic cloth rising from RMB 3.8/meter in September 2025 to about RMB 10/meter in August 2026 (more than 165% year on year). This report obtained no new price-increase letter for 9/9–9/10, so one cannot say "a price-increase letter came out today, hence the rally". ⚠️ One counter-fact must also be logged: 金安国纪 (Jin'an Guoji), 超声电子 (Ultrasonic Electronics) and others have recently issued a string of clarifications that they do not supply NVIDIA and that high-end products are still in development — the high-end narrative on this chain has a ceiling set by the companies themselves, the same in kind as this morning's caveats on 崇达 (Chongda) and 华正 (Huazheng).

3.3 Were There Unexpected Main Lines Missed Pre-Market? — two, and the second was findable

① Fiberglass/electronic cloth: 1st industry gainer today. This falls under "the branch boundary was drawn too small", not "did not see it at all" — it is the same chain as the branch 2 already listed.

② Power/green power/gas: this one is a genuine miss; see §3.4.

3.4 ⚠️ Full Post-Mortem of the Missed Power Main Line — the clue was in the data I pulled myself this morning

Today's facts:

Catalysts (all outside this morning's news window):

• 9/1 "New Power System Construction 15th Five-Year Plan" — it specifies grid fixed-asset investment of more than RMB 5 trillion during the 15th Five-Year period, up more than 30% from the 14th Five-Year period. 7 trading days ago.
• Agency power-purchase tariffs rose month on month in 14 regions nationwide in September; Jiangsu's September monthly trading tariff was +7.8% year on year, the fourth consecutive month of recovery.
• National power load set four consecutive record highs during the summer peak season (reaching 1.553 billion kW on 7/30 and rising to 1.557 billion kW on 8/7).

⚠️ Why it was missed, and why "outside the window" cannot be used to explain it away this time:

The conclusion of §1 of this morning's list was "no new policy-type positive appeared inside the window… there is no policy-driven branch today". That sentence is literally correct (there was indeed no new policy between 9/9 15:00 and 9/10 07:30), but it was used as "there is no policy-driven main line today", and those are two different things.

Worse still: the clue was already in the data I pulled this morning —

• 闽东电力 (Mindong Electric Power, 000993) is on its 2nd consecutive limit-up today, meaning it hit the limit on 9/9, and it was right there among the 48 names in the official 9/9 limit-up pool cited on this morning's list (2 power/gas limit-ups on 9/9: 闽东电力 / Mindong Electric Power and 新中港 / Xinzhonggang).
• Earlier still: 贵州燃气 (Guizhou Gas) hit the limit on 9/8. From 9/7 → 9/8 → 9/9 → 9/10, the power-and-gas limit-up count went 0 → 1 → 2 → 7, monotonically increasing.
• This morning's list did run an industry-distribution tally on the 9/9 limit-up pool (the internal log explicitly recorded that the mandatory "limit-up pool industry distribution" item was completed and identified the catalyst behind the 5-consecutive-limit-up general retail name), but it was used only to count names inside the target branch, never to ask "which industry's limit-up count is rising continuously".

This is a compounded replay of memories [[scan-window-misses-multiday-themes]] and [[ztpool-industry-mix-only-counts-target-branch]]. The fix must land in the template, not be written as an admonition:

Add a mandatory pre-market table: "Industry distribution of the limit-up pool across the last 3 trading days", listing each industry's limit-up count at T−3 / T−2 / T−1. Any industry whose count rises on two consecutive days automatically enters the candidate pool and must receive a conclusion (either included, or with an explicit reason for exclusion).

This table uses only data already being pulled, at zero extra cost; it adds a second dimension, "pricing progress", to the one-dimensional scanner that is the "news window". Today it would have flagged power at the 9/9 close — exactly when power went from 1 name to 2.

---

4. The Limit-Up Ladder and the High-Rung Names

4.1 Ladder Structure: 1 lone survivor at 4 consecutive limit-ups + 2 at 3 + 6 at 2 + 26 first-time

⚠️ The ladder breaks at the 3-consecutive rung: 1 name at 4, 2 at 3, 6 at 2 — the same shape as 9/9 (a lone 5-consecutive survivor, 4 names at 3), a second straight day with no viable promotion path. The highest rung fell from 5 to 4 consecutive limit-ups and changed hands (9/9's 5-consecutive name was 百大集团 / Baida Group, today −3.04% with an Open→Close of −10.89%) — the high rung is not a continuation; it broke and restarted.

4.2 Seal Quality: weak seals are nearly 30%, and 63% broke the limit before resealing

• "Weak seals" with sealing capital/turnover < 0.05 number 10 out of 35 (28.6%): 美能能源 (Meineng Energy, 0.020), 超声电子 (Ultrasonic Electronics, 0.024), 再升科技 (Reetec, 0.024), 京基智农 (Kingkey Smart Agriculture, 0.036), 津药药业 (Jinyao Pharmaceutical, 0.036), 水发燃气 (Shuifa Gas, 0.038), 云煤能源 (Yunmei Energy, 0.040), 中新赛克 (Sinovatio, 0.042), 新力金融 (Xinli Finance, 0.048).
• 22 of the 35 limit-up stocks (63%) broke the limit at some point intraday, of which 6 broke it 3 or more times: 津药药业 (Jinyao Pharmaceutical) 7 times, 新力金融 (Xinli Finance) 6 times, 美能能源 (Meineng Energy) 5 times, 克明食品 (Keming Foods) 4 times, 中新赛克 (Sinovatio) 4 times, 长城电工 (Great Wall Electric) 3 times.
• 24 names (69%) sealed for the first time before 10:00 — concentrated early-session seals loosening through the afternoon is the classic shape of an ebb phase.
• The three most solid seals: 瑞尔特 (Rieter Sanitary) (seal-to-turnover ratio 5.671, locked at the open, impossible to buy), 国创高新 (Guochuang Hi-Tech) (0.492), 大连重工 (Dalian Heavy Industry) (0.250).
• Largest turnover among today's limit-up stocks: 金安国纪 (Jin'an Guoji, 002636) RMB 4.791 billion, 太极实业 (Taiji Industry, 600667) RMB 3.744 billion, 超声电子 (Ultrasonic Electronics, 000823) RMB 2.770 billion, 再升科技 (Reetec, 603601) RMB 1.966 billion — the top three all belong to the two main lines of the day, "Components + semiconductors/engineering", so volume and main-line direction agree.

4.3 Anchor-Leader Identification (recognizable high-turnover gainers)

4.4 ⚠️ Attribution of Three Large Limit-Up Names: not established

太极实业 (Taiji Industry, 600667, Shanghai main board), 环旭电子 (Universal Scientific, 601231, Shanghai main board) and 再升科技 (Reetec, 603601, Shanghai main board) all sealed limit-up today with turnover among the highest (RMB 3.744 / 1.025 / 1.966 billion), and this report was unable to establish their attribution for today.

Exclusion work already done: the eiTime of the latest announcements for the three are 8/26, 9/9 16:27 (legal opinion on the employee share-ownership plan) and 9/8 15:33 (notice of a results briefing) respectively, none of which carries substantive new information; the article found attributing 太极实业 (Taiji Industry)'s limit-up (memory-chip sector +10.02%) was published on 08/11, not today, and has been discarded.

The honest answer of this report is: attribution unknown. The facts that can be logged are: 太极实业 (Taiji Industry) is simultaneously today's leading gainer in four concept sectors — green power, memory chips, engineering construction and architectural decoration — and its net main-force inflow accounted for a net share of 50.52% (i.e. the net inflow is about half its turnover), an extreme one-sided buying pattern. "No news found" is not the same as "there was no news" (memory [[no-news-found-is-not-no-news]]), and tomorrow's pre-market work must re-run a line-by-line trace of announcements and investor-interaction platforms for these three.

---

5. Fund Flows

5.1 ⚠️ Northbound Funds: no data

Since 2024-08-19, Stock Connect no longer discloses daily net buying. This report gives no northbound figures and does not pass off quota balances as inflows.

5.2 ⚠️ Dragon-Tiger List: not obtained in this report

The exchanges publish the Dragon-Tiger List after 18:00 following the close, and this report was generated at 15:40, so today's Dragon-Tiger List and seat preferences are left blank here. This is scheduling, not an interface failure (the same interface returns 66 rows normally for 9/9 data). Tomorrow's pre-market work must add a seat-level breakdown of today's Dragon-Tiger List, noting that overlapping seats between the buy list and the sell list must be de-duplicated stock by stock and reconciled against the official net buy amount — adding them directly systematically overstates net buying.

5.3 Industry Net Main-Force Inflow / Outflow

Top 10 net inflows (⚠️ new-listing contamination already stripped out)

Top 8 net outflows

⚠️ Machinery shows a headline −3.049 billion; after removing the +1.298 billion contributed by today's newly listed N 电科思 (Dianke Si, 301689, ChiNext), the true figure is −4.347 billion. A new listing's debut day contaminates the fund-flow reading of its sector, and tomorrow's reading for that sector will still be contaminated, so do not compare it directly.

5.4 Leading Single-Stock Net Main-Force Inflows / Outflows

Top 8 net inflows: 太极实业 (Taiji Industry, 600667) +1.891 billion, N 电科思 (Dianke Si, 301689, debut day, not comparable) +1.298 billion, 天孚通信 (T&S Communications, 300394) +1.261 billion, 金安国纪 (Jin'an Guoji, 002636) +1.178 billion, 宁德时代 (CATL, 300750) +733 million, 国际复材 (International Composites, 301526) +697 million, 剑桥科技 (Cambridge Technology, 603083) +578 million, 再升科技 (Reetec, 603601) +547 million.

Top 8 net outflows: 亨通光电 (Hengtong Optic-Electric, 600487) −1.366 billion, 中际旭创 (Innolight, 300308) −1.326 billion, 新易盛 (Eoptolink, 300502) −882 million, 比亚迪 (BYD, 002594) −843 million, 远东股份 (Far East Holding, 600869) −812 million, 恒瑞医药 (Hengrui Pharma, 600276) −802 million, 华工科技 (HGTECH, 000988) −778 million, 立讯精密 (Luxshare Precision, 002475) −677 million.

⚠️ A clear split has emerged inside the optical-module sector, and this is the single most noteworthy item on the fund-flow side today: 天孚通信 (T&S Communications) +1.261 billion of net inflow / 剑桥科技 (Cambridge Technology) +578 million, while 中际旭创 (Innolight) is −1.326 billion / 新易盛 (Eoptolink) −882 million. This morning Passed the seven optical-module names as a single bloc; today proves it is no longer a single bloc.

5.5 Margin Positions and Broad-Based Baskets: systematic de-risking

The margin-trading concept was −29.128 billion (net share −2.05%), FTSE Russell −19.477 billion, S&P −17.160 billion, large caps −15.938 billion, Shenzhen Connect −15.886 billion, MSCI China −15.538 billion, Shanghai Connect −14.363 billion, the robotics concept −11.249 billion, and liquid-cooled servers −9.405 billion (net share −6.67%).

⚠️ This set of readings does not contradict "the index fell only 0.43%"; the two explain each other: broad-based and leveraged money is cutting positions while the index is held up by banks (city commercial banks +1.93%, large state-owned banks +1.51%) and brokers (+0.81%) — SSE 50 −0.21% the most resilient and BSE 50 −2.81% the weakest are precisely the two ends of this structure.

---

6. Outlook for Tomorrow

① Judgment on the Persistence of Tomorrow's Main Lines

② Key Focus for Tomorrow (branches + stocks + verification points)

⚠️ The verification points in this report come in two tiers. Structural verification points (consecutive limit-ups, anchor leaders, multi-day fund flows) rank above exogenous price-type ones — this is the direct lesson from today, where verification point 1 passed with a +3.71% margin and the trade still lost. When they conflict, follow the structural one.

③ What to Avoid Tomorrow

1. ⚠️⚠️ Indiscriminate follow-on buying of yesterday's limit-up stocks — for the third straight day this is the surest way to lose money. The 48 limit-up stocks from 9/9 posted a median −1.04% today with only 45.8% up; the yesterday's-failed-limit-up concept was −3.17%, yesterday's touched-limit names −3.30%, and yesterday's high-turnover names −3.76% (net outflow 4.489 billion).
2. ⚠️ Chasing gap-ups. Today's counter-examples are extremely dense: 百大集团 (Baida Group) gapped up +8.80% → Open→Close −10.89%; 南京港 (Nanjing Port) gapped up +10.02% (limit-up at the open) → Open→Close −6.56%; 海通发展 (Haitong Development) gapped up +5.83% → Open→Close −4.96%; 招商轮船 (China Merchants Energy Shipping) gapped up +2.00% → Open→Close −3.88%. Meanwhile the best-quality seal of the day, 瑞尔特 (Rieter Sanitary, 002790), was locked at the open and impossible to buy.
3. ⚠️ The four confirmed-ebbing lines: oil shipping, power grid equipment, optical fiber & cable, and agriculture/fertilizer. In particular, do not re-enter oil shipping just because oil rises again — today already ran that experiment with a +3.71% move in oil.
4. ⚠️ The lone high-rung survivor: 桂林旅游 (Guilin Tourism, 000978, Shenzhen main board) at 4 consecutive limit-ups. It belongs to none of today's fund main lines, and 9/9's lone 5-consecutive survivor 百大集团 (Baida Group) posted an Open→Close of −10.89% today — the fate of the previous lone survivor is the best warning available.
5. ⚠️ Weak-seal limit-up stocks (sealing capital/turnover <0.05): 美能能源 (Meineng Energy, 001299), 超声电子 (Ultrasonic Electronics, 000823), 再升科技 (Reetec, 603601), 京基智农 (Kingkey Smart Agriculture, 000048), 津药药业 (Jinyao Pharmaceutical, 600488, broke the limit 7 times), 水发燃气 (Shuifa Gas, 603318), 云煤能源 (Yunmei Energy, 600792), 中新赛克 (Sinovatio, 002912, broke the limit 4 times), 新力金融 (Xinli Finance, 600318, broke the limit 6 times).
6. ⚠️ The fund-flow readings for the Machinery/General Equipment sector will still be contaminated tomorrow by the listing of N 电科思 (Dianke Si, 301689), so they cannot be used directly for cross-sector comparison.
7. ⚠️ BSE 50 was −2.81% today, the weakest in the market; BSE names (±30% limit) have markedly amplified volatility, and carry suitability thresholds such as RMB 500,000 in assets.

④ Input Notes for Tomorrow Morning's Pre-Market List (6 directly executable items)

1. ⚠️⚠️ Add a mandatory table: "Industry distribution of the limit-up pool across the last 3 trading days". List each industry's limit-up count at T−3 / T−2 / T−1, and any industry whose count rises on two consecutive days enters the candidate pool and must receive an explicit conclusion, either inclusion or exclusion. This table costs nothing extra to build, and it would have flagged power at the 9/9 close. (The direct patch for today's missed power line.)
2. ⚠️⚠️ Upgrade "5-day / 10-day cumulative net main-force inflow" into a precondition for admission to a recommended slot, filled in branch by branch, with any branch negative on both barred from a recommended slot. This morning the discipline was written down and correctly applied twice, to PCB and optical fiber, but was not applied to the power-grid branch that was most wanted — and that branch produced the worst recommended slot of the day.
3. ⚠️ Verification points must be tagged with priority: structural (consecutive limit-ups / anchor leaders / multi-day fund flows) above exogenous price-type ones (commodities / FX / overseas markets). And the table must state whom to follow when they conflict. Today verification point 1 passed with a +3.71% margin while verification point 2 signaled failure, and the outcome followed verification point 2.
4. ⚠️ Add a "Pass type" column to the Pass list, distinguishing "the logic does not hold" from "position/valuation too high". 深科技 (Shenzhen Kaifa) belonged to the former today yet rose 4.32%, so "we do not endorse it" and "we are bearish on it" must be separated at the label level.
5. ⚠️ A recommended stock requires first determining "which end of this chain is rising today", then ranking within that end. Today's order of gains along the CCL chain was upstream copper foil/fiberglass (+6% to +15%) > CCL (+2.8% to +10%) > PCB manufacturing (−1% to +1%), while both of our slots were at the very bottom of the chain. Balance-sheet quality sets position size and the floor, not the ranking for the day.
6. ⚠️ The three attribution gaps must be closed: 太极实业 (Taiji Industry, 600667), 环旭电子 (Universal Scientific, 601231), 再升科技 (Reetec, 603601). No recommended slot before attribution is established; and if no substantive catalyst is found after the check, write "attribution unknown" honestly — "attribution unknown" is allowed as a conclusion; padding with old news is not.

⑤ Calendar Events Tomorrow and Over the Next 5 Days

---

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

Supplementary notes (please read together):

1. The labels "delivered / did not deliver / called correctly / called wrong / Pass" in the text are only after-the-fact reconciliation tags against our own list from this morning, not investment ratings, and they carry no implication of position size, entry/exit timing or target price.
2. Data read times: all quotes, sector and single-stock fund flows and the limit-up/failed-limit-up/limit-down pools are on a 2026-09-10 closing basis, read at 15:30–15:40 (Beijing); the limit-down count first read 12 names at 15:30 and 11 on the re-reads at 15:37 and 15:40, and this report uses 11.
3. ⚠️ This report has four known evidence gaps; please discount the strength of the related conclusions accordingly: ① today's Dragon-Tiger List was not obtained (published after 18:00), the Dragon-Tiger section of §5 is left blank, and the seat character of today's top net-inflow stocks cannot be determined; ② northbound funds have not disclosed daily net buying since 2024-08-19, so this report has no northbound data; ③ the attribution of the three limit-up stocks 太极实业 (Taiji Industry, 600667), 环旭电子 (Universal Scientific, 601231) and 再升科技 (Reetec, 603601) was not established, as stated explicitly in §4.4, with no old news used as filler; ④ no primary 9/9–9/10 price-increase document was obtained for the current CCL / electronic cloth price hikes, so §3.2 characterizes it as an ongoing price-hike cycle, not a fresh catalyst today.
4. ⚠️ "No news found" does not equal "there was no news". The three attribution gaps in §4.4 are missing information, not negative evidence; please do not read them backwards as "no catalyst, therefore it should fall".
5. Incomplete-coverage statement: this report took a stock-by-stock snapshot of all 5,912 listed names and computed advance/decline counts in-house, but stock-level deep verification covered only the 46 names on this morning's list and the 68 names in today's limit-up/failed-limit-up/limit-down pools, so the vast majority of names were not individually researched; not being mentioned does not mean there is no opportunity, and being mentioned does not mean it has been sufficiently researched.
6. Price limits are tiered by board: Shanghai/Shenzhen main board ±10%, ChiNext/STAR ±20%, BSE ±30%. Do not read ±20% / ±30% names with ±10% intuitions — for example, 逸豪新材 (Yihao New Material, 301176)'s +15.29% is not a limit-up, while 百迈科 (Baimaike, 920268)'s +29.99% is a BSE limit-up. The STAR Market and the BSE also carry investor-suitability thresholds such as RMB 500,000 in assets, and some readers are not eligible to trade them.
7. Special note on new and recently listed shares: N 电科思 (Dianke Si, 301689) debuted today at +237.50%, and a new listing contaminates both the fund-flow and price-change readings of its sector on debut day; this report has quantified and stripped it out in §5.3 (Machinery headline −3.049 billion → true −4.347 billion). Tomorrow's listing of Enflame Technology will produce the same kind of contamination in related STAR Market sectors.
8. Stock-specific risks: the stocks in this report generally carry risks including results falling short of expectations, shareholder reductions, unlocking of restricted shares, regulatory inquiries, suspension for verification, abnormal share-price movements and insufficient liquidity. 湖南黄金 (Hunan Gold, 002155)'s major asset restructuring still requires approval at the 9/11 shareholders' meeting (it was voted down on 7/29), review by the Shenzhen Stock Exchange and registration with the CSRC, and both approval and timing carry significant uncertainty.
9. This report was generated by an automated process and may contain deviations from information lag, incorrect value-chain mapping or data-interface failures, and cannot be used directly as a basis for trading. Please defer to formal exchange and company disclosures.
10. Position statement of the producer: the producer of this report holds none of the securities described herein and has no interest relationship with the companies described.

---

Data sources: East Money quote and fund-flow interfaces (push2delay.eastmoney.com, stock-by-stock snapshots of all 5,912 names, 496 industry sectors, 504 concept sectors), East Money official limit-up/failed-limit-up/limit-down pools, East Money listed-company announcements (eiTime basis), Shanghai International Energy Exchange crude oil futures (akshare), Sina index historical volume. News and industry information sources: Jiemian News¹, Jiemian News flash², China News Service³, Cailianshe⁴, JRJ⁵, 10jqka Finance⁶.