A-Share Market Recap | 2026-09-01 (Tuesday)

Scope and basis of this report (judge the strength of the conclusions accordingly) ① Reconciliation is given on two bases: "close chg%" is based on the 8/31 close (the same basis as the pre-market report); "open→close %" is based on today's opening price. The pre-market report was published at 07:00, and the earliest price a reader could actually execute at is the opening price produced by the 09:25 call auction, so "open→close %" is the true return of executing the list as written. This report gives both columns side by side, and the hit rate is disclosed on both bases. ② Index and single-stock quotes are on the Tencent qt.gtimg.cn close-settlement basis, read at 15:30–15:43. The SSE Composite close of 3,979.89 and today's open of 3,979.88 are almost identical, and this report has independently re-verified it: the value stayed unchanged across two reads at 15:30:31 and 15:43:01 while the timestamp advanced, and the day's high of 3,995.18 / low of 3,976.47 bracket it — confirming a genuine full-day arc that flattened out, not a stale snapshot. ③ Limit-up / consecutive limit-up data are on the East Money official limit-up pool 9/1 basis; the number of entries in the pool has been verified to match the official count (both 83). ④ Sector chg%, sector net inflow and advancer/decliner counts are on the THS 90-industry-sector closing snapshot basis, which differs from East Money's "main-force net inflow" basis and cannot be directly subtracted from the pre-market report's East Money figures. The §Fund-flow section of the pre-market report also cited the THS basis, so this report and the pre-market report are comparable. ⑤ The dragon-tiger list (LHB) is left blank today: the exchange only discloses the official data after 18:00, and this report was written at 15:30, so not being able to get it is not an API failure. The data will be filled in tomorrow morning's pre-market report. ⑥ No northbound flow data: since 2024-08-19 the daily net buy figure of the Stock Connect has ceased to be disclosed. This report gives no northbound number, and cites no API value that returns 0.0. ⑦ The limit-down count could not be secondarily confirmed; see the note in §1. This report gives no definitive figure.

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

I. API failures that actually occurred this run

• East Money clist whole-market endpoint completely unavailable: across two rounds, a total of 29 pages × 3 retries were attempted, all failed (first round RemoteDisconnected, second round returned non-JSON). This has now been broken for several consecutive trading days. Advancer/decliner counts were therefore switched to summing the "advancers/decliners" of the THS 90 industry sectors (3,387 / 2,040) — exactly the established fallback channel recorded in memory as "when East Money clist is down, switch to THS", and it worked this time.
• ak.stock_board_industry_name_em() (East Money industry sectors) ConnectionError, same-origin failure. Switched to ak.stock_board_industry_summary_ths(), which returned 90 rows normally.
• The direct push2ex limit-down pool / broken-limit pool call returned rc=206 (rate limited), data=None. But before the rate limit took effect, ak.stock_zt_pool_zbgc_em had already successfully fetched 6 broken-limit entries (complete column structure, not an empty shell), so that reading is usable.
• The limit-down pool ak.stock_zt_pool_dtgc_em returned 0 rows with empty columns. ⚠️ Under the criterion "a non-zero tc with an empty pool is a silent failure", 0 rows + empty columns cannot be distinguished from a genuine "no limit-downs today"; moreover the midday news reported "1 limit-down". The secondary confirmation channel (direct push2ex) was blocked by the 206 rate limit, so this report gives no limit-down count and only notes in §1 that the midday reading was 1. Do not write this up as "zero limit-downs today".
• The dragon-tiger list ak.stock_lhb_detail_em('20260901','20260901') threw TypeError: NoneType — this is a scheduling problem of running at 15:30, earlier than the 18:00 disclosure time, not an API failure, the same phenomenon as in the 8/31 recap. Left blank per established practice with an explanatory note; to be filled in tomorrow morning's pre-market report.

II. Channels that worked

• The limit-up pool push2ex getTopicZTPool (dpt=wz.ztzt, sort=fbt:asc) returned tc=83 / len=83, verification passed. Continue using sort=fbt:asc.
• ak.stock_zt_pool_previous_em(date='20260901') returned 88 rows, consistent with the official 8/31 limit-up count of 88, and the per-stock chg% matches today's quotes (我爱我家 +10.00, 浪潮信息 +0.49). Confirmed once again: the date parameter of this endpoint must be set to "today" to get "how yesterday's limit-up stocks performed today"; passing yesterday returns yesterday's data.
• Quotes come from Tencent qt.gtimg.cn, timestamps 20260901153031–20260901154301, confirmed to be post-close settlement data.

III. One search contamination intercepted this run

• Among the top results for the query "9/1 A-share closing commentary", two were intraday pieces published at 10:28 and at midday today (Jiemian News "Midday limit-up/limit-down analysis: 68 limit-up stocks, 1 limit-down stock", Zhitong Finance "A-share midday"), and another, Sina's "Close: SSE Composite dips then rebounds, up 0.19%, agriculture sector explodes", was actually published on 2026-08-18 — the headline carries no year and the content is highly similar to today's (both about agriculture exploding). ⚠️ Copying it would have taken 8/18's closing commentary as today's, and the direction happens to be the opposite (that day the SSE Composite rose 0.19%, today it fell 0.16%). Discarded. All closing figures in this report are fetched by us; no closing figure from search is used.

IV. A methodological check that was actively done and that changed the conclusion

• The first draft was about to write agriculture up as "a main line entirely unforeseeable from this morning's pre-market". Checking stock_zt_pool_previous_em for the 8/31 limit-up pool overturned that: among the 88 limit-up stocks on 8/31 there were already 3 agriculture names (万向德农 5 consecutive limit-ups / 8 limit-ups in 11 days, 新赛股份 3 consecutive, 福建金森 2 consecutive), and 万向德农 was at that time the highest consecutive-limit-up tier in the entire market. The pre-market report's §internal explicitly stated "the main-line judgement is now computed by ourselves from the industry distribution + first-seal time + seal/turnover ratio of the 88 limit-up stocks" — which means the data was already in hand; agriculture was not invisible, it was overlooked. The conclusion was therefore rewritten from "an unforeseeable surprise" to "a foreseeable miss", and the self-critique in §3 was rewritten accordingly. This is another instance of "the scan window misses multi-day-brewing main lines", and more serious than before — this time even the in-window limit-up pool had already pointed at it.

V. What this report did not do (listed honestly)

• Did not obtain the dragon-tiger list (scheduling reason, see above).
• Did not obtain stock-level main-force net inflow rankings (depends on the already-broken East Money clist); the fund-flow conclusions in §5 only go down to sector level.
• Did not obtain a reliable reading of the limit-down count (see above).

---

0. One-line recap

Today's sentiment temperature: warming up, leaning exuberant — but the heat switched tracks. At the index level it was green: SSE Composite −0.16%, SZSE Component −1.02%, ChiNext Index −1.32%, STAR 50 −2.19%; but at the single-stock level it was red — 3,387 advancers versus 2,040 decliners (62.4% advancing), and 68 of the 90 industry sectors closed up, 22 down, with 83 limit-ups against only 6 broken limits, the broken-limit rate collapsing from yesterday's 25.42% to 6.7%, and the consecutive-limit-up height extending from 2 boards all the way to 7 boards. This is the classic "weak index, strong money-making effect", an exact mirror-image reversal of the "good index, poor relay" call in yesterday's pre-market report.

Today's strongest main line was none of the ones the pre-market report backed — it was "agricultural planting + mass consumption". Planting & forestry +7.51% (30 up, 0 down), the best in the market with not a single decliner; retail +3.30%, insurance +3.16%, diversified financials +3.02%, agricultural product processing +2.90% and baijiu +2.85% followed. Of the 83 limit-ups, 14 are agriculture-related and 11 are retail/consumption-related, and three of the top four consecutive-limit-up heights are agriculture (万向德农 6 boards, 新赛股份 4 boards, 福建金森 3 boards). The driver is overnight external commodity prices: the Bloomberg Agriculture Spot Index rose more than 13% cumulatively in August, the largest monthly gain since July 2012, with attacks on Black Sea ports pushing wheat to a three-year high and sugar and cocoa each up about 20%.

⚠️ But this must be stated plainly: this main line is not one that "only appeared today and was unforeseeable pre-market" — the pre-market report missed it. Among the 88 limit-up stocks on 8/31 there were already 3 agriculture names, and 万向德农 was already on 5 consecutive limit-ups (8 limit-ups in 11 days), the highest consecutive-limit-up tier in the whole market. The pre-market report explicitly stated that its main-line method was "computing the industry distribution of the 88 limit-up stocks ourselves" — the data was in hand, the conclusion never came out. This is the single thing most worth remembering from this report; see §3.

Pre-market list reconciliation: the 5 recommended names were 3 up / 2 down on the close basis, averaging +1.60%, but only +0.48% if executed at the open — the two bases differ by 1.12pct. Over the same period the CSI 300 was −0.30% and the ChiNext Index −1.32%, so the recommended group still beat the indices; it just lost to its own list: all three Pass categories beat the recommended group (the third, "poor financial quality", averaged +4.05%, the strongest group on the whole list). This is now the second occurrence of the same error, and the self-critique at the end of §2 does not dodge it.

Tomorrow's tone: the main lines are agriculture and consumption, but agriculture is already in its 3rd week of pricing and the consecutive-limit-up height has reached 7 boards — this is a "participate with a reduced position" spot, not a "just started, chase it" spot. The tech chain (semiconductors −2.64%, net outflow of RMB 17.448 billion, the largest in the market) completed a sizeable capital withdrawal today; do not bottom-fish tomorrow, let it stop bleeding on its own.

---

1. Market overview

Combined turnover of the two exchanges was approximately RMB 2,033.4 billion (the THS 90-industry-sector basis totals RMB 2,051.9 billion; the difference comes from stocks not covered by the sector classification and is normal).

The SSE Composite traced a complete arc through the day: open 3,979.88 → high 3,995.18 (briefly turning positive at midday to 3,987.56) → low 3,976.47 → close 3,979.89, with the close 0.01 points above the open. The late-session weakness came from ChiNext and the STAR Market, not the main board.

Sentiment indicators:

⚠️ This report gives no limit-down count. The close-basis limit-down pool endpoint returned 0 rows with empty columns, which cannot be distinguished from a genuine "zero limit-downs", and the secondary confirmation channel was blocked by rate limiting. The only confirmable figure is an intraday reading: 1 limit-down at midday (Jiemian News 9/1 midday¹). Please do not read this report as "zero limit-downs today".

Sentiment temperature call: warming up, leaning exuberant. Three mutually independent grounds: ① a broken-limit rate of 6.7%, meaning sealed limits were rarely pried open — strong absorption; ② the consecutive-limit-up height rising from 3 boards to 7 boards, showing money is willing to pay a premium for the high-tier names, something entirely absent yesterday; ③ 62.4% of stocks advancing on a day when the indices broadly fell, showing the selling pressure was highly concentrated in heavyweight tech names rather than market-wide.

But one counter-indicator must be written up alongside: turnover did not expand to confirm it. Yesterday's pre-market report recorded that on 8/31 the semiconductor sector alone did RMB 260.856 billion of turnover; today's whole-market RMB 2,033.4 billion is a normal level. The improvement in sentiment comes from a structural switch — "abandon the losing sectors, concentrate holdings in the winning ones" — not from incremental money entering. That makes tomorrow's character rotation, not a broad rally.

---

2. Pre-market list reconciliation

2.1 The five recommended names (pre-market report §9.1)

Hit rate (all three bases given, because they give different answers):

The two bases differ by 1.12pct, in the same direction as yesterday (yesterday's gap was 2.34pct). Reconciling on close chg% alone systematically overstates the hit rate, so this report keeps both columns. The recommended group did beat the indices (+0.48% vs −0.30%) and that deserves credit; but the next section shows it lost to its own Pass list.

The performance of the verification points themselves deserves its own note: all 5 verification points called the direction correctly. The two that were met (昆仑万维, 中际旭创) both rose today; of the three not met, two fell outright and one (瑞芯微) closed up but lost money if executed at the open. In particular 恺英网络: the pre-market report wrote "if there is no volume, the market does not buy the 'earnings + low valuation' line" — today's turnover rate of 3.35% did not even reach yesterday's level, and the market indeed did not buy it. The discriminating power of the verification points is clearly higher than that of the score (73-pt 恺英网络 fell, 63-pt 中科创达 also fell, but the score ordering did not predict it).

2.2 Pass-list reconciliation — the ugliest section of this report

The pre-market report split the Pass list into three categories by reason. Back-tested on the same classification, all three categories beat the recommended group:

Name-by-name detail:

Also: the other two names marked "watch only" in the overall ranking of §3 of the pre-market report — 新易盛 (Eoptolink, 300502) +0.23% (called right), 中国银行 (Bank of China, 601988) +2.46% (called wrong; banks were the top net-inflow sector today), 鹏鼎控股 (Avary Holding, 002938) −5.44% (called right, among the largest declines in the market).

23 ⚠️ 芒果超媒300413the downgraded name hit the limit, but this account has to be settled precisely

The pre-market report spent an entire section (§5.11) downgrading 芒果超媒 from a recommended slot to "watch only, 53 pts", on three independent pieces of evidence: operating profit of only RMB 30.2 mn, an ex-non-recurring loss in Q2 alone, and RMB 168 mn of the RMB 202 mn net profit attributable to parent coming from the recognition of deferred tax assets; membership revenue −22.6%; and on its limit-up day, foreign capital + institutions net-sold RMB 134 mn on the dragon-tiger list, absorbed by hot money.

Today it was +20.02% limit-up, the best performer among all stocks within the recommendation universe. On the surface this is the biggest miscall in this report. But on the executable basis, the conclusion must be corrected:

• The opening price of RMB 20.38 = the limit price, i.e. a one-line open at the limit. It could not be bought at the open.
• It broke the limit once intraday, with a low of RMB 19.66 (+15.78% versus yesterday's close), then re-sealed into the close.
• In other words: the only buying window of the day was at +15.78% or higher, while the return from executing at the open is 0.00%.

The honest conclusion is: the "do not buy" decision did not lose us money, but it did make us miss an executable intraday entry (in at +15.78%, out at +20.02%, roughly 3.7pct of room). We cannot say the call "was actually right" just because it opened at the limit on one line — there was 1 broken limit intraday, so the window genuinely existed.

More importantly, the downgrade reasoning itself was neither falsified nor confirmed today. 芒果超媒's seal/turnover ratio today was only 0.061 (yesterday 0.357) and its turnover rate 13.45% (yesterday 6.41%), a marked deterioration in lock-up quality versus yesterday — consistent in direction with the pre-market report's read that "institutions are distributing, hot money is absorbing". The fundamental grounds for the downgrade (deferred tax, declining membership revenue) are quarterly-level facts and cannot be falsified by price in a single day.

2.4 Self-critique: this is now the second occurrence of the same error

The types of wrong calls are highly concentrated, and the pattern is the same as last time:

1. "Weak fundamentals" was used as a short-term ranker. Category III Pass (poor financial quality) averaged +4.05%, the strongest group on the whole list, 2.45pct above the recommended group. The pre-market report itself wrote, in §2.1, the completely correct sentence "this branch can be participated in today, but the reason for participating must be 'event + money flow', not 'earnings delivery'" — and then, in the scoring model, raised the fundamentals weight to 15 pts and on that basis deleted 芒果超媒 from the recommended slots and removed 欢瑞世纪 from the list. The judgement was written right, the ranking was done backwards.

2. "Position too extended" is not a reason to avoid; it is only a reason to cut size. Category II Pass averaged +2.31%, within which 我爱我家 went from 2 boards to 3 boards, +10.00%. The pre-market report's wording — "already 2 consecutive limit-ups, yesterday's seal ratio of 2.47 was already an extreme, today you are taking the odds of the 3rd board" — describes the odds correctly, but the conclusion jumped from "the odds got worse" to "do not participate", and that step had no basis.

3. But a distinction is needed: within Category I Pass (narrative refuted by primary-source evidence), the names genuinely refuted by primary evidence were called right — 荣信文化 −2.59%, 香江控股 −1.01%, 光线传媒 −0.17%, 中来股份 (execution basis) −0.95%. The ones called wrong (欢瑞世纪, 果麦文化) were refuted on the ground of "an interim-report loss", which falls right back into problem no. 1. So "using primary-source evidence to refute a specific attribution" works, while "using weak financials to refute a theme" does not — and these two things were mixed into the same category in the pre-market report.

Concrete fixes for tomorrow morning's pre-market report (three, directly actionable):

• Split the scoring model into two tables: a "direction/theme table" that uses only events + money flow + the limit-up ladder to define the stock set, and a "position table" that uses fundamentals and price position to set each name's weight. Fundamentals no longer enter the ranking of the stock set, only the position sizing.
• The Pass list must be listed by reason in separate columns and back-tested separately (done in this report; recommend making it permanent): "attribution refuted by primary-source evidence" may be removed from the list; "poor financial quality" and "position too extended" must uniformly become position cuts rather than removals, and their chg% must be kept in the list for next-day reconciliation.
• Verification points first: all 5 verification points in this report called the direction right, with higher discriminating power than the score. Recommend that the pre-market report write the verification points first and derive the recommendation conclusions backwards from them, rather than ranking first and adding verification points afterwards.

---

3. Verification of today's main lines

3.1 Branch reconciliation table

Branch hit rate: of 8 branches, 2 called right (AIGC film & TV, banks), 1 diverging (optical modules), 5 called wrong.

3.2 Did the pre-market report call the strongest main line? — half right

The pre-market report ranked "AIGC film & TV / micro-dramas" first, and this line did keep strengthening today, with its self-set verification point cleanly delivered. The pre-market report wrote: "no 2nd boards tomorrow = one-day-trip sentiment; only with 2nd boards does it qualify as a main line".

Today's result: the media complex produced 7 names on 2 boards or more — 时代出版 (Time Publishing, 600551) 3 boards, 欢瑞世纪 (000892) 2 boards, 中广天择 (Zhongguang Tianze, 603721) 2 boards, 大晟文化 (Dasheng Culture, 600892) 2 boards, 芒果超媒 (300413) 2 boards, 掌阅科技 (iReader Technology, 603533) 2 boards, 龙版传媒 (Longban Media, 605577) 2 boards, 博纳影业 (Bona Film, 001330) 2 boards. Yesterday's "flat ladder" of 12 first boards out of 13 limit-ups has grown structure today. By the pre-market report's own criterion: the main line is established.

⚠️ But one piece of counter-evidence must be written up alongside: the cultural media sector rose 2.74% while money net-outflowed RMB 1.476 bn. This is exactly the same pattern as yesterday's communications equipment "up 1.56% yet net outflow of RMB 3.426 bn" — and communications equipment fell 0.94% today. A gain propped up by a handful of sealed limit-up names while the sector as a whole is being sold is a classic late-main-advance feature, not a healthy main advance.

A second downgrade signal is the extreme polarisation in seal quality: 欢瑞世纪 seal/turnover 5.314, 时代出版 only 0.003 (and 9 broken limits), 掌阅科技 0.004, 龙版传媒 0.004 (4 broken limits). Within the same main line, the hardest and the softest differ by three orders of magnitude. 时代出版, as a 3-board high-tier name, only barely sealed after 9 broken limits, which shows the disagreement at the highs is already extremely violent.

3.3 ⚠️ The surprise main line missed pre-market: agricultural planting — except it was not a "surprise", it was a miss

The strongest sector in the market today was planting & forestry +7.51% (30 up, 0 down), and there is not one word about it in the pre-market report. Related sectors exploded in step: agricultural product processing +2.90%, agrochemicals +2.39%. Of the 83 limit-ups, 14 are agriculture-related:

The driver is external commodity prices, not domestic policy: the Bloomberg Agriculture Spot Index, which tracks 10 major agricultural products, rose more than 13% cumulatively in August, the largest monthly gain since July 2012; attacks on Black Sea ports cut exports from that key producing region, and wheat prices hit a three-year high; sugar and cocoa each rose about 20%, with a strengthening El Niño aggravating weather concerns (Jiemian News 9/1 10:28² / UkrAgroConsult³).

⚠️ The point of this section is not "agriculture rose today" but "why it was not written pre-market" — and the answer is not that it was unforeseeable.

Among the 88 limit-up stocks on 8/31 there were already 3 agriculture names: 万向德农 (5 consecutive limit-ups at the time, 8 limit-ups in 11 days), 新赛股份 (3 consecutive at the time), 福建金森 (2 consecutive at the time). And 万向德农 was at that time the highest consecutive-limit-up tier in the whole market — the pre-market report's §2.1 explicitly wrote "12 of the 13 media limit-ups are first boards, the ladder is extremely flat", yet failed to notice that within the very same limit-up pool, agriculture had already produced a 5-consecutive-limit-up high-tier name.

The pre-market report's §internal states in black and white that its main-line method is "now computed by ourselves from the industry distribution + first-seal time + seal/turnover ratio of the 88 limit-up stocks". The data was in hand and the method was right, but the industry-distribution dimension was only used to count how many media names there were — it was never used to find anything outside media.

This is a repeat of "the scan window misses multi-day-brewing main lines", and worse than before: previously the catalyst of the main line fell outside the scan window (the JPMorgan grain report), whereas this time both the catalyst and the ladder had already shown up inside the window, in the limit-up pool.

A hard requirement for tomorrow morning's pre-market report: the industry distribution of the limit-up pool must be listed in full and sorted by consecutive-limit-up height, not just counted for the target branch. "Which industry does the market's highest consecutive-limit-up stock belong to" should be a fixed, must-answer item.

3.4 Today's real money main line: a wholesale switch from tech → agriculture + consumption + financials

This is not the rotation of a single sector, it is a style switch. Put the top and bottom of net inflows side by side and it is obvious:

Everything flowing in is low-valuation / pro-cyclical / domestic demand (banks, insurance, brokers, power, retail, pharma); everything flowing out is the electronics-tech chain (semiconductors, components, consumer electronics, electronic chemicals). The 3.5pct divergence between STAR 50 −2.19% and BSE 50 +1.34% is the same thing projected onto the indices.

---

4. Limit-up ladder / high-tier boards

83 limit-ups today; the consecutive-board structure:

Ladder assessment: the structure is complete, which yesterday's was not. Of yesterday's 88 limit-ups the highest was only 3 boards and 79.5% were first boards; today every tier of 7/6/4/3/2 boards is filled, with no break in the promotion chain. The first-board share fell to 78.3%, slightly better but still high.

⚠️ The quality of the high-tier names varies enormously and must be examined one by one:

• 海鸥住工 (002084), 7 boards — sealed on a one-line open at 09:25, 0 broken limits, seal/turnover 0.275, turnover rate 13.81%. The highest-quality high-tier board today, without dispute.
• 万向德农 (600371), 6 boards — 2 broken limits, seal/turnover of only 0.012. ⚠️ 8 limit-ups in 11 days means it has already run for two and a half weeks, and the seal orders are so thin they are almost no defence at all — it is the high-tier name most likely to break tomorrow.
• 捷荣技术 (002855), 6 boards — 5 broken limits, opened repeatedly all day, seal/turnover 0.183. Extremely violent disagreement.
• 时代出版 (600551), 3 boards — 9 broken limits, seal/turnover 0.003 (the lowest tier in the market), turnover rate 16.00%. ⚠️ This is today's most fragile consecutive-limit-up name; 9 broken limits mean every seal was immediately smashed open. As the highest-tier name of the media main line, its fragility directly determines whether media can continue tomorrow.
• 我爱我家 (000560), 3 boards — 5 broken limits, seal/turnover 0.017, turnover rate 37.08%, turnover of RMB 2.722 bn. High turnover rate + low seal orders: a fully-churned relay board rather than a locked-up board.

The "main force" (large turnover + recognisable advancing names):

Broken streaks: of the 88 limit-up stocks of 8/31, today 43 rose and 45 fell, averaging +1.46% with a median of −0.19%, and 18 hit the limit again (relay rate 20.5%). The worst streak breaks were concentrated in the property and decoration chain: 深物业 A (Shenzhen Properties A, 000011) −9.04% (streak broken after 2 consecutive limit-ups), 金螳螂 (Gold Mantis, 002081) −7.63%, 名雕股份 (Mingdiao Decoration, 002830) −5.77%, 天保基建 (Tianbao Infrastructure, 000965) −4.97%, 万通发展 (Wantong Development, 600246) −4.95%. Yesterday's property chain was a pure one-day trip.

---

5. Fund flows

5.1 Northbound flows

No data. Since 2024-08-19, the daily net buy figure for the Stock Connect has ceased to be disclosed. This report gives no northbound number, and cites no API value that returns 0.0. Any claim that northbound money net-bought or net-sold a given amount today does not hold up.

5.2 Dragon-tiger list

Left blank today. The exchange only discloses official dragon-tiger list data after 18:00, and this report was written at 15:30, so not being able to get it is a scheduling issue, not missing data. The data will be filled in tomorrow morning's pre-market report (the 8/31 dragon-tiger list was already completed in this morning's pre-market report, 84 entries).

5.3 Sector main-force net inflow (THS basis, closing snapshot)

Top 10 net inflows (RMB bn):

Top 10 net outflows (RMB bn):

5.4 ⚠️ One money-flow lesson that must be remembered: yesterday's extremes all reversed today

Put the 8/31 figures cited by the pre-market report next to today's, and the reversal is systematic:

The pre-market report wrote: "yesterday's volume and money are hard evidence", and on that basis ranked domestic semiconductors 5th at A−. Today that sector net-outflowed RMB 17.448 bn with 21 up and 165 down, the largest outflow in the market.

This lesson has been recorded before — single-day main-force fund-flow extremes mean-revert, they do not trend. As a description of that day's relative strength it is accurate; as a forecast of the next day's direction it is inverse. Three of the four comparisons reversed, and the only one that did not — banks — happens to be the branch the pre-market report scored lowest (B+) and valued least.

Concrete fix: when the pre-market report cites "yesterday's top sector net inflow", it cannot be used as a bullish reason, only as a factual statement of "who was strongest yesterday"; to use it as a next-day criterion, look at whether the money direction is consistent over 2–3 consecutive days, not at a single-day ranking.

---

6. Outlook for tomorrow

6.1 Main-line continuity assessment

6.2 Key focus for tomorrow (branch + stock + verification point)

⚠️ The following are objects of observation and falsifiable verification points, not buy recommendations. The purpose of a verification point is that it can be definitively judged met or not after tomorrow's close.

6.3 What to avoid tomorrow

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

1. [Highest priority] The industry distribution of the limit-up pool must be listed in full and sorted by consecutive-limit-up height. The root cause of today's miss is: the data was in hand, but the industry distribution was only used to count how many names the target branch had. "Which industry does the market's highest consecutive-limit-up stock belong to" should be made a permanent must-answer item. Today's answer is household products (海鸥住工, 7 boards); tomorrow morning this question must be answered before any branch discussion.
2. [Scoring model] Split it into a "direction table" and a "position table". Fundamentals and price position no longer enter the ranking of the stock set, they only determine position weight. All three Pass categories in this report beat the recommended group — the second occurrence of the same error.
3. [Pass list] List by reason in separate columns, and back-test each separately. "Attribution refuted by primary-source evidence" may be removed; "poor financial quality" and "position too extended" must uniformly become position cuts, not removals, and their chg% must be retained for next-day reconciliation.
4. [How to use fund-flow data] Yesterday's sector net-inflow ranking can only serve as the fact of "who was strongest yesterday", not as a bullish reason for the next day. To use it as a criterion, the direction must be consistent over 2–3 consecutive days. Today's RMB −25.8 bn reversal in semiconductors is the Nth proof.
5. [Verification points first] Write the verification points first, then derive the recommendation conclusions backwards from them. All 5 verification points in this report called the direction right, with markedly higher discriminating power than the score ranking.
6. [Data that must be added] The dragon-tiger list after 8/31 (disclosed after 18:00 today) should be filled into tomorrow morning's pre-market report, with a focus on the seat profile of the agriculture and consumption sectors — determining whether institutions or hot money are driving this main line directly determines its persistence.
7. [External variables to track] The Bloomberg Agriculture Spot Index and wheat, sugar and cocoa prices. The driver of this main line is in external commodity markets, and A-share agriculture names are followers. If commodities pull back overnight, the domestic main line will most likely cool in step the next day.
8. [Yesterday's leftovers] The overnight US headwinds noted in the pre-market report (Nasdaq, the Golden Dragon China index −2.19%, the 10Y Treasury yield at 4.75%) exerted no substantive pressure on A-shares today, with the SSE Composite only −0.16% and 62.4% of stocks advancing. The transmission of overseas headwinds to A-shares essentially failed today, and the overall tone should not be marked down on that basis tomorrow morning.

---

⚠️ Risk disclaimer: this recap is post-close information compilation and observation 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.

Sources:

• Jiemian News 9/1 10:28 — Grain ETF underlying index up over 6.4%, global agricultural product prices rising²
• Jiemian News 9/1 midday — Midday limit-up/limit-down analysis: 68 limit-up stocks, 1 limit-down stock¹
• UkrAgroConsult — Global agricultural commodity prices post strongest monthly rise since 2012³
• Chinanews 9/1 — A-share open: over 2,600 stocks in the red, SSE Composite and ChiNext Index open lower, SZSE Component opens higher⁴
• Quote and limit-up pool data: Tencent qt.gtimg.cn (close-settlement basis, read at 15:30–15:43), East Money official limit-up pool (push2ex, 9/1 basis, tc=len=83), THS 90-industry-sector closing snapshot