A-Share Market Recap | 2026-09-09 (Wednesday)

Scope and data-basis boundaries of this report (please judge the strength of the conclusions accordingly) ① A-share quotes are on a 2026-09-09 close basis, read at 15:30–15:33 (Beijing), sourced from Tencent qt.gtimg.cn. ② The limit-up / limit-up-break / limit-down pools come from the official East Money pool push2ex, with len(pool)==tc verified (limit-up 48==48, limit-up breaks 27==27, limit-down 7==7). ③ Advance/decline counts are self-computed after pulling the entire market name by name (5,911 targeted, 5,911 actually retrieved, coverage 100%); the top-N rows of a sorted, paginated query are NOT used. ④ Limit-up determination is tiered by board price limit: Shanghai/Shenzhen main board ±10%, ChiNext/STAR Market ±20%, Beijing Stock Exchange ±30%. Every stock in this report is tagged with its board. ⑤ ⚠️ This report does NOT include today's Dragon-Tiger List. The Dragon-Tiger List is published only after 18:00, while this report was generated at 15:30–16:10, so it is physically unobtainable. All capital-flow conclusions in §5 come from main-force net inflow (an intraday order-flow basis). This is an information disadvantage of this report relative to the pre-market note, not a data failure. ⑥ ⚠️ No northbound flow data (daily net buy disclosure was discontinued as of 2024-08-19); this report gives no northbound figures. ⑦ ⚠️ The 长飞光纤 H-share reading (06869.HK) is an intraday 15:32 price; Hong Kong does not close until 16:00, so that figure is not a closing price — flagged in §3. ⑧ Domestic futures are on a 09-09 day-session 15:00 close basis; the white sugar change is given on two baselines (prior settlement / prior close) — reason in §3.

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

I. Channels that worked

• Quotes via Tencent qt.gtimg.cn (indices, individual stocks, and the Hong Kong line r_hk06869 all succeeded on the first try); daily bars via web.ifzq.gtimg.cn/appstock/app/fqkline.
• Limit-up / limit-up-break / limit-down pools via push2ex.eastmoney.com; all three pools passed the tc==len(pool) check; additionally pulled the 9/8 limit-up pool (73==73) to compute how yesterday's limit-up names performed today.
• Sector / concept / single-stock capital flows via push2delay.eastmoney.com — succeeded on the first try for the second consecutive trading day; the backup-domain memory keeps closing the loop.
• Futures via ak.futures_zh_realtime; urea / white sugar / cotton / corn / soybean meal / rapeseed meal all retrieved.
• Announcements via np-anotice-stock, using eiTime rather than notice_date to fix the publication moment — that memory paid off directly this time; see item 3 below.
• Crude via CNBC @CL.1/@LCO.1, with volume=33,189 and last_time=09-09T03:23 ET — active with real trades, usable for adjudicating a verification point.

II. Channels that failed

• ⚠️ ak.stock_zh_index_spot_em returned RemoteDisconnected; the East Money akshare cluster has now been broken for seven consecutive trading days, so the entire index line was rerouted to Tencent.
• ⚠️ ak.stock_lhb_detail_em(20260909) returned TypeError: NoneType not subscriptable — this is not a broken interface; there simply is no data before 18:00. Anticipated per memory [[lhb-publishes-after-1800]] and written into boundary ⑤; no time wasted debugging.
• ⚠️ The column names of ak.futures_zh_realtime no longer match the old code (no settle/date; they are actually settlement/tradedate), so the first call threw KeyError; succeeded after renaming the columns.

III. Four methodological cross-checks that were made deliberately and that changed the conclusions

1. ⚠️⚠️ [Most important item in this report] The advance/decline count was nearly written the wrong way round because of pagination bias. The first pull looped with pz=1000, but each page actually returned only 100 rows; after 12 pages it had 1,200 rows that were all advancers (because the sort was by descending change), and using that directly would have produced the absurd conclusion "1,200 up, 0 down"; even taking only the ratio would have badly overstated breadth. After switching to a loop until len(allr)>=total (60 pages, 5,911 rows, coverage 100%), the true result is 1,794 up / 3,642 down. This is exactly the one-sided bias described in memory [[em-clist-pagination-one-sided-bias]], and this time it was caught before it reached the body text.

2. ⚠️⚠️ White sugar has two change baselines, differing by 0.45pp and with opposite signs. akshare's changepercent gives +0.07%, while back-computing from preclose gives −0.38%. Field-by-field reverse computation confirms: that field is based on the prior settlement price of 5,481, not the prior close of 5,506; both numbers are correct, on different bases. Per memory [[two-pct-figures-may-differ-by-base]] this report shows both columns rather than reporting a "data conflict". More important still: today's high of 5,505 < the 9/8 in-window high of 5,510 cited in the pre-market note — on either baseline, the verification condition "white sugar keeps making new highs" was NOT met.

3. ⚠️⚠️ The trigger for today's collapse in publishing was missed by the pre-market note, and it was inside the window. 龙版传媒 (605577) filed a trading-halt-for-verification announcement with eiTime = 2026-09-08 21:24:06, landing squarely in the middle of the news window the pre-market note set for itself (9/8 15:00 → 9/9 07:10); the stock's turnover today was 0 (halted), while in the same sector 出版传媒 (601999) hit limit-down, 读者传媒 (603999) −7.82%, 中国出版 (601949) −5.25%, and 中信出版 (300788) −6.46%. Passing on publishing was correct, but the stated reason was "the catalyst has no primary source" rather than "a hard negative had just been published." Right conclusion, one link missing from the evidence chain.

4. ⚠️ The catalyst for the agriculture theme was traced back to its first-disclosure date, and it turned out to be outside the window. Lianbanwang attributed today's broad agriculture move to "Thai sugar output may fall below 10 million tonnes"; tracing it to the original Jiemian News article, the publication timestamp is 2026-09-08 10:21, i.e. during the 9/8 session, outside the pre-market window. This explains why 中粮糖业 (600737) had already hit limit-up +9.99% on 9/8 — today is day 2 of pricing that news, not day 1. Handled as "pricing progress day 2" per memories [[trace-to-first-disclosure-date]] and [[scan-window-misses-multiday-themes]].

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

• Today's Dragon-Tiger List, institutional seats and hot-money seats were all unobtained (published after 18:00). §5 stands on the single leg of main-force net inflow, one independent evidence chain fewer than the pre-market note; this is a structural disadvantage of this slot.
• The catalyst for the general-retail branch (百大集团 (600865) 5 consecutive limit-ups, 中百集团 (000759) 3 consecutive limit-ups, 友好集团 (600778)) remains unidentified. Pulling all three companies' announcements one by one: 百大集团 has only abnormal-movement / risk-warning notices (9/8 17:05, 9/7 16:30), neither of which discloses any reason; 中百集团's most recent announcement dates back to 8/26, i.e. zero announcements inside the window. This is the same item the pre-market note has failed to identify for four consecutive days; this report likewise failed to identify it, and says so plainly.
• N 洛轴股份 (301699) listed today at +101.45% with main-force net inflow of RMB 696 million, which contaminates the bearing / general-equipment sector capital-flow readings; §5 of this report does not use that sector's data, so no exclusion was performed, but readers who look up that sector themselves must exclude it.
• The performance of yesterday's limit-up names was checked stock by stock against the official 9/8 limit-up pool of 73 names, without using stock_zt_pool_previous_em (that interface has the known trap of returning the same day rather than the next day, per memory [[zt-pool-previous-returns-same-day]]).

---

0. One-sentence recap

Sentiment turned from "euphoric" to "ebbing", yet all three recommendation slots in the pre-market list delivered — and this was the cleanest reconciliation this slot has had in recent memory.

Today's A-share market was a textbook case of the index being defended while individual stocks fell broadly: the Shanghai Composite +0.28%, yet market-wide 1,794 up / 3,642 down, with a median stock change of −0.68% — a complete reversal from yesterday's 3,257 up / 1,859 down. Limit-ups fell from 73 to 48, limit-downs rose from 0 to 7, the limit-up failure rate was 36.0%, and the highest consecutive-limit-up count dropped to 5. All three growth indices gapped up and faded: the ChiNext Index opened +1.15% and closed −0.14%; the STAR 50 opened +1.08% and closed −0.69%.

Today's genuinely strongest theme was optical fibre/cable — which was precisely branch #2 of the pre-market note. Communication cable and accessories +4.98%, main-force net inflow RMB 4.495 billion (net ratio 8.78%); the optical-fibre concept +1.69%, net inflow RMB 6.490 billion. The single most distinctive judgement of the pre-market note — "read optical fibre and optical modules separately" — was fully confirmed today: all four fibre names rose (长飞 +6.04%, 亨通 +5.56%, 中天 +4.95%, 烽火 +1.99%), while all four optical-module names (中际旭创 (300308) +0.72%, 光迅 +1.60%, 天孚 −0.18%, 新易盛 (300502) −0.38%) gapped up 3%–4.5% and then faded back to flat.

The second line, oil shipping, also delivered: 招商轮船 (601872) hit limit-up, the shipping sector +3.48% with net inflow of RMB 1.071 billion, 3 limit-ups in shipping & ports plus 1 in marine equipment.

Reconciliation results (all with board tags, on both bases):

The three tiers are strictly monotonic on both bases, and the recommendation group clearly outperformed. ⚠️ But it must be said at the same time: the "open basis" is the executable basis — the list is published before the open, and the earliest price a reader can transact at is today's open, not yesterday's close. On the open basis, the Pass group averaged −0.05% and the recommendation group +6.25%, so the real spread is 6.3pp, not the 5.5pp shown on the close basis (this time the gap was in fact understated by the close basis).

Tone for tomorrow: relay within an ebbing phase; the main line shifts from "breadth of numbers" to "capital concentrating in the optical-fibre/copper-cable physical layer"; agriculture demands high vigilance — the stocks are rising while the commodity futures are falling across the board.

---

1. Market overview

1.1 Indices (2026-09-09 close, read at 15:30)

⚠️ The column to read in this table is "open→close", not "change". All four growth indices gapped up and faded; the STAR 50 slid all the way from +1.08% to −0.69%, an intraday drawdown of 1.75pp. The gap-up imported from the overnight rally in US semiconductor equipment stocks was fully digested within the first hour of A-share trading — the pre-market note's §5.4 judged that "the rally in US equipment stocks speaks to future supply, not current demand, and is not a reason to be long A-share semis", and today's intraday action is consistent with that judgement.

1.2 Sentiment indicators

Sentiment reading: ebbing phase. Five indicators deteriorated at once — limit-ups halved, limit-downs went from zero to seven, first-board count halved, advance/decline reversed, and turnover contracted by RMB 100 billion. The only counter-indicator is the highest consecutive-limit-up count rising from 4 to 5 (百大集团 (600865)), and it is a lone name: today there were only 4 names on their 3rd board and none on a 4th, so the ladder has a void at the 3rd board.

⚠️ One structure that must be spelled out: today's "index up" and "stocks down" are not a contradiction, they are two faces of the same thing. The top of the main-force net-inflow list is almost entirely large-cap weights (紫金矿业 (601899) RMB 896 million, 宁德时代 (300750) RMB 595 million, 中国巨石 (600176) RMB 825 million), while capital exited AI/computing power/chips on a large scale (artificial intelligence −RMB 10.650bn, AI applications −RMB 7.558bn, domestic chips −RMB 6.919bn, memory chips −RMB 5.712bn). The index is being held up by a handful of weights while breadth deteriorates.

---

2. Reconciliation of the pre-market list

⚠️ This table gives two bases. "Change%" is based on yesterday's close (market convention); "open→close%" is based on today's open — the list is published before the open and the earliest transactable price for a reader is today's open, so this is the executable column. Per memory [[recap-vs-premarket-price-column]], reconciling on the close-based change alone systematically overstates the hit rate.

2.1 Recommendations (3 names; 2 slots publicly left empty in the pre-market note)

Recommendation group: 3/3 delivered, close-basis mean +7.33%, open-basis mean +6.25% (whole-market median −0.68%).

2.2 Watch only (12 names)

Watch-only group: +2.37% on the close basis (8 up, 4 down); +1.29% on the open basis (7 up, 5 down).

2.3 Pass list (15 items)

Pass group: +1.85% on the close basis (8 up, 7 down); −0.05% on the open basis (6 up, 8 down, 1 flat). ⚠️ That one flat name is 金正大 (002470), which was locked limit-up from the open — on an open-price basis its return is identically 0, because no transactable price was offered all day.

2.4 ⚠️ Hit rate and self-critique

Hit rate (measured by "was the direction of the conclusion correct"):

⚠️ Where the three right calls and the three wrong calls respectively came from:

Methods that worked and are worth keeping:

1. "Write the verification points before the recommendation conclusions" once again outperformed the conclusions themselves today. Of the 7 verification points in pre-market §8, 6 were right today (details in §3.3), above the 66.7% of the recommendation conclusions. Point 1 in particular — "watch the volume, not the size of the gap-up" — 招商轮船 (601872) opened flat at 0.00% (which the old criterion "a gap-up of more than 3% is a one-day-trip pattern" would have misjudged), but turnover expanded 123%, so under the new criterion it qualified — and the result was a limit-up. That criterion rescued a pattern that would otherwise have been misjudged.
2. "Read optical fibre and optical modules separately" was the most valuable independent judgement of this cycle (§3.1); it came neither from price change nor from capital flow, but from checking the scope of applicability of the FCC draft.
3. Fundamental checks vetoed three originally intended recommendations in a row; today two of those vetoes proved right and one wrong: 奥浦迈 (688293) did not hit limit-up (right), 中天/亨通 rose (wrong). Net effect is positive, but see the next section.

The three types of wrong calls, and the specific fixes for tomorrow's pre-market note:

1. ⚠️⚠️ [The one most in need of change] "Share of revenue" was used on its own as a long/short switch, when it is only a weight. The pre-market note pulled 中天科技 (600522) (14.0%) and 亨通光电 (600487) (8.4%) out of the fibre branch entirely; today the two were +4.95% / +5.56%, ranking 5th and 3rd market-wide in main-force net inflow (RMB 0.988bn / RMB 1.310bn). A low revenue share means low elasticity, not that a stock will not rise — in a sector-beta move, names in the same sector get bought together. ⚠️ But nor can one simply say the removal was wrong: pre-market §8.1 item 4b set a reverse self-check — "if the two rally hard and outperform 长飞, the decision to remove them was wrong" — and today's readings are 长飞 +6.04% > 亨通 +5.56% > 中天 +4.95%, with 长飞 still the strongest, so that condition was not triggered. The accurate statement is: the branch assignment was right (长飞 is the strongest), but deleting them from the "may rise" list was an over-reaction. Fix: use share of revenue only for within-branch ranking and position weighting, not as a binary "remove/keep" switch.
2. ⚠️ The mapping "capital outflow yesterday → bearish tomorrow" failed twice today. 沪电股份 (002463) was Passed on the grounds that "the PCB sector saw RMB 1.304 billion of net outflow on 9/8, a distribution pattern"; today the PCB concept had main-force net inflow of RMB 9.845 billion and the printed-circuit-board sector rose +2.73%, with 沪电 +2.12%. This matches memory [[single-day-fund-flow-mean-reverts]] ("single-day extremes in main-force flow are a contrarian indicator") exactly — the pre-market note used single-day capital flow for a directional call, when this slot's own memory explicitly states it cannot be used that way. Fix: single-day sector capital flow may only be used to characterise same-day strength/weakness, never extrapolated to the next day's direction; extrapolation requires a 3–5 day cumulative basis.
3. ⚠️ The "liquidity trap" criterion in the Pass list only caught half the cases. 亚盛集团 (600108) (volume ratio 0.09) broke its board today and was −1.21% on the open basis — the criterion worked; but 百大集团 (600865) (turnover RMB 59 million) ran to a 5th board at +9.97% — the criterion failed. The difference: 亚盛 had 3 risk announcements suppressing it, while 百大 had only abnormal-movement notices that disclosed no reason. Fix: separate out "has a substantive risk announcement self-disclosing valuation or earnings already been published" as its own Pass rationale; it discriminates better than a turnover threshold.

---

3. Verification of today's main themes

3.1 Branch reconciliation table

3.2 Three items that must be spelled out separately

① ⚠️ Agriculture is the most dangerous line today: the stocks are rising while all of its commodities are falling.

The verification point set in pre-market §8.1 item 6 reads verbatim: "If white sugar keeps making new highs while urea lags, the layering of 'trade soft commodities only, not fertiliser' holds; if both fall, judge the whole line as ebbing." Today's day-session readings:

⚠️ Giving two baselines for white sugar is necessary, not vague phrasing: akshare's changepercent is based on the prior settlement of 5,481 and yields +0.07%, while the prior close of 5,506 yields −0.38%. Both numbers are correct. But on either one, today's high of 5,505 is below the 9/8 high of 5,510, so the condition "keeps making new highs" was not met.

Verdict: close to "both fall", so on the verification point the whole agriculture line should be judged as ebbing. Meanwhile, on the equity side: other cultivation +5.96%, nitrogen fertiliser +3.28%, and all three agricultural sub-sectors had main-force net outflows. The most glaring combination is 中粮糖业 (600737): share price +9.03% while main-force net outflow was RMB 711 million with a net ratio of −12.98% (4th on the market-wide net-outflow list). Per memory [[commodity-futures-falsify-equity-theme]], a divergence between the commodity and the equity means "capital rotation rather than fundamentals" — and this is the clearest such divergence today.

⚠️ One additional check traced to the first-disclosure date: Lianbanwang attributed today's broad agriculture move to "Thai sugar output in the 2026/27 crushing season may fall below 10 million tonnes (12 million tonnes last season, a decline of about 17%)", citing the Thai Sugar Millers Corp. Tracing to the original Jiemian News article, the publication timestamp is 2026-09-08 10:21, i.e. during the 9/8 session — this explains why 中粮糖业 (600737) had already hit limit-up +9.99% on 9/8. Today is day 2 of pricing that news, not day 1.

② ⚠️ Today's collapse in publishing had a hard catalyst, it was inside the pre-market window, and the pre-market note missed it.

龙版传媒 (Longban Media, 605577, SH main board) published its trading-halt-for-verification announcement at 2026-09-08 21:24:06 (on the eiTime basis), landing squarely in the middle of the news window the pre-market note set for itself, 9/8 15:00 → 9/9 07:10. Its turnover today was 0 (halted), while in the same sector 出版传媒 (601999) hit limit-down at −9.98%, 读者传媒 (Reader Publishing, 603999) −7.82%, 中国出版 (601949) −5.25%, 中信出版 (300788) −6.46%, and the publishing sector −4.29%.

Passing on publishing pre-market was right, but the stated reason was "the only catalyst was a Sina AI-generated page, with no primary source." The reality is: not only was there no primary source for a positive, there was a primary-source negative that had just been published — and it was not picked up. Right conclusion, one link missing from the evidence chain — per memory [[no-news-found-is-not-no-news]], "no catalyst found" and "there is a hard negative" are two entirely different Pass rationales, and the latter also supports the stronger action of shorting/avoiding the whole sector.

③ ⚠️ The divergence between optical fibre and optical modules is the cleanest confirmation of a pre-market judgement today.

Pre-market §8.1 item 4: "If optical modules are markedly stronger than optical fibre, the market is buying 'optical communications' as a single basket, and the core 'read them separately' judgement of this note is falsified." Actual readings:

All four optical-module names gapped up 3.3%–4.5% and all closed 1.7–4.6 percentage points below their opening price; all four fibre names closed above their opening price. The divergence not only holds, its direction matches the pre-market call exactly. ⚠️ One counterweight from the capital side: the optical-communications-module concept had main-force net inflow of RMB 8.749 billion today — capital has not left — which is consistent with the pre-market wording that "this is a do-not-chase, not a write-off." The precise conclusion is: the money is still there, but today it could not push prices up; this is high-level churn, not distribution.

3.3 Item-by-item verdict on the pre-market verification points (6 of 7 right)

---

4. Limit-up ladder / height boards

Highest consecutive board: 百大集团 (600865, SH main board ±10%) on its 5th board, +9.97%, turnover RMB 1.039 billion, turnover rate 18.79%, board broken twice.

⚠️ Its catalyst has now been unidentified for four consecutive trading days, and this report did not identify it either. Pulling announcements one by one: an abnormal-movement notice at 9/8 17:05, a risk-warning notice at 9/7 16:30, an abnormal-movement notice at 9/4 18:07 — none of the three discloses any reason. In the same branch, 中百集团 (000759, SZ main board) is on its 3rd board at +10.00% with a turnover rate of 28.37% and zero announcements inside the window (its latest announcement dates back to 8/26). A 5th-board leader plus a 3rd-board follower, and the entire branch has no verifiable fundamental anchor.

4.1 Consecutive-board ladder (official pool basis)

⚠️ The void at the 4th board is the most important ladder feature of the day. Yesterday there were 3 names on a 4th board; today there are none — of yesterday's three 4th-boarders, 亚盛集团 (600108) broke its board (+8.14%), 爱仕达 (002403) fell −9.08%, and 百大集团 (600865) advanced to a 5th board. One of three advanced, a promotion rate of 33%, and the only one that advanced has an unidentified catalyst.

⚠️ Another item easily misread: Lianbanwang records 亚盛集团 (600108) as being on a "4th board", which does not match the official limit-up pool. In the official push2ex pool there is no 4th board today, and 亚盛集团 sits in the limit-up-break pool (board broken 3 times, closed +8.14%, closing price 5.71 < limit-up price 5.81). This report goes with the official pool, per memory [[limit-updown-must-use-official-pool]].

4.2 How yesterday's limit-up names (73) performed today

Grouped by yesterday's consecutive-board count:

⚠️ The combination "mean +1.11% but median −0.60%" is itself the conclusion: a handful of big gainers lifted the mean, while more than half of yesterday's limit-up names fell today. The relay profit effect no longer holds. ⚠️ The 4th-board group's mean of +3.01% looks strongest, but n=3 (百大 +9.97%, 爱仕达 (002403) −9.08%, 亚盛 +8.14%) — the sample is far too small to constitute a criterion that "extended names are stronger" — per memory [[subsample-monotonic-is-coincidence]], no qualitative conclusion should be drawn from a sub-sample of n≤3.

Representative streak-breakers today: 出版传媒 (601999) −9.98% limit-down, 我爱我家 (5i5j, 000560) −9.30%, 爱仕达 (ASD, 002403) −9.08%, 百合花 (Lily Group, 603823) −8.55%, 读者传媒 (603999) −7.82%, 中曼石油 (603619) −0.60%.

---

5. Capital flows

⚠️ This section stands on the single leg of main-force net inflow. Today's Dragon-Tiger List and institutional seat data are published only after 18:00, while this report was generated at 15:30–16:10, so they are physically unobtainable — this is a structural information disadvantage of the post-close slot relative to the pre-market slot, not a data failure. Tomorrow morning's pre-market note should make "what did the institutional seats buy today" a mandatory question. ⚠️ Daily northbound net buy disclosure has been discontinued since 2024-08-19, so this report gives no northbound figures.

5.1 Sector main-force net inflow / outflow

On a concept basis, the direction of capital migration is very clear:

⚠️ Today's capital flow in one sentence: out of "the software and chip layers of AI" and into "the physical connectivity layer of AI". Artificial intelligence, AI applications, AIGC, Xinchuang and memory chips saw combined net outflows of over RMB 37 billion, while 5G, communications technology, PCB, optical modules, CPO and optical fibre saw combined net inflows of over RMB 52 billion. This is not a decline in risk appetite; it is a shift of layers within the same AI theme.

5.2 Top single-stock main-force net inflow / outflow

⚠️ Four of the top 10 net inflows are optical fibre/cable names (亨通, 永鼎, 中天, 长飞), totalling RMB 4.225 billion. This is the most concentrated capital line of the day. ⚠️ N 洛轴股份 is a new listing today, and its RMB 696 million net inflow contaminates the capital-flow readings of the bearing / general-equipment sector, so it must be excluded when reading that sector's data (memory [[new-listing-pollutes-sector-fund-flow]]).

⚠️⚠️ 中粮糖业 (600737) is the one row that most deserves to be read out today: one of the biggest gainers in the whole market (+9.03%), and simultaneously 4th on the main-force net-outflow list (−RMB 711 million, net ratio −12.98%). Turnover rate 13.44%, volume up 195% from yesterday. Rising on expanding volume plus heavy main-force net outflow = a textbook distribution pattern, corroborating the commodity-side divergence in §3.2. The pre-market "watch only" (do not chase) conclusion missed a 9% gain, but from the standpoint of holder structure the direction was right.

---

6. Outlook for tomorrow

6.1 Judgement on theme continuation

6.2 Focus for tomorrow (branches + stocks + verification points)

⚠️ Every verification point must test "the variable that would make this trade lose money", not an indicator that will probably pass (memory [[verification-point-must-test-the-loss-driver]]).

6.3 What to avoid tomorrow

1. ⚠️ The height boards in general retail (百大集团 600865, SH main board; 中百集团 000759, SZ main board). A 5th board plus a 3rd board, while the catalyst has been unidentified for four consecutive trading days and none of the three companies' announcements disclose a reason. ⚠️ Note this report's stance: this is not "we judge it will fall" — today it was the strongest thing in the market. It is "with no verifiable anchor whatsoever, this slot has no ability to offer a reason to participate."
2. ⚠️ Indiscriminate relay-buying of yesterday's limit-up names. Across the 73 names that hit limit-up yesterday, today's median was −0.60% and more than half fell; of the 39 that gapped up, 11 closed down. The relay profit effect is already negative.
3. ⚠️ Semiconductors, AI computing, memory chips, Xinchuang. The magnitude of the outflows (over RMB 37 billion combined) is consistent with the STAR 50's pattern of opening +1.08% and closing −0.69% today; there is insufficient evidence for bottom-fishing against the trend.
4. ⚠️ Media/publishing across the board. 龙版传媒 (605577) is halted for verification, and the sector faces ongoing regulatory pressure until it resumes; 出版传媒 (601999) already hit limit-down today.
5. ⚠️ Chasing agricultural fertiliser (金正大 002470, SZ main board). On its 3rd board, and today it was a locked limit-up from the open (volume ratio 0.34, turnover RMB 296 million), so it could not be bought at the open; the fundamentals (RMB 497 million half-year loss, borrowings overdue 5.7 years, debt ratio 86.30%) have not improved at all, and urea futures were −1.15% today.

6.4 ⚠️ Input notes for tomorrow morning's pre-market list (in descending order of importance)

1. ⚠️⚠️ The question tomorrow morning must answer with the Dragon-Tiger List: of the main-force net inflow into today's four fibre names (长飞 601869, 亨通 600487, 中天 600522, 烽火 600498), how much came from institutional seats? This report cannot answer because there is no data before 18:00, and this is the only hard evidence that separates "sector beta" from "institutional accumulation". If institutional seats are absent, today's RMB 4.2 billion of net inflow was hot-money behaviour and its persistence must be discounted.
2. ⚠️⚠️ The way "share of revenue" is used must change. Pre-market, 中天 (14.0%) and 亨通 (8.4%) were pulled out of the fibre branch entirely; today both rose and their capital inflows ranked 3rd and 5th market-wide. Fix: use the share only for within-branch ranking and position weighting, not as a binary "remove/keep" switch. At the same time, keep the part that was proved right today — 长飞 has the highest purity and was indeed the strongest today, so the within-branch ranking was correct.
3. ⚠️⚠️ Single-day sector capital flow must not be used to extrapolate the next day's direction. 沪电股份 (002463) was Passed because of "RMB 1.304 billion of net outflow on 9/8"; today PCB had net inflow of RMB 9.845 billion. This slot's memory already states plainly that single-day capital extremes are a contrarian indicator — the pre-market note violated its own memory. Extrapolation must switch to a 3–5 day cumulative basis.
4. ⚠️ The news-window scan must cover negative announcements of the "halt / verification / inquiry" type. 龙版传媒 (605577)'s 9/8 21:24 trading-halt-for-verification announcement was inside the window yet was missed, and it was the direct trigger for today's collapse in publishing. Suggestion: add a separate mandatory check in the window scan for the keywords "halt, verification, inquiry letter, risk warning".
5. ⚠️ The time threshold on a verification point was set too tight. "Oil shipping holds if 3 or more names hit limit-up before 10:00" — in reality the 4 limit-ups were spread across 09:58–10:55, failing the strict reading while substantively having diffused. Suggest changing it to 3 names before 11:30, so that an otherwise valid branch is not rejected by an over-tight threshold.
6. ✅ Two practices proved effective today and worth keeping: ① write the verification points before the recommendation conclusions (6 of 7 right today, more accurate than the recommendation conclusions' own 66.7%); ② when a branch holds but no single stock passes the checks, publicly leave the recommendation slot empty rather than forcing a name in — the pharma/CRO branch was −1.29% with RMB 3.946bn of net outflow today, and forcing in a representative stock at the time would have produced a loss today.
7. ⚠️ Make the catalyst behind the general-retail 5th board a standalone mandatory check tomorrow morning. It has been unidentified for four consecutive days, and it is the market's highest consecutive board. Per memory [[ztpool-industry-mix-only-counts-target-branch]], "which industry does the market's highest consecutive-board name belong to, and what is its catalyst" must be a mandatory question, not skipped merely because it falls outside the pre-set branches.

---

⚠️ Risk disclaimer

1. This recap is post-close information organisation and observation only; it does not constitute investment advice, a securities recommendation, or any form of offer to buy or sell. The labels "watch closely / watch only / Pass / avoid" used herein are purely internal information-strength gradings for this report; they are not investment ratings and carry no implication as to position size, entry/exit timing, or target price.
2. Data timestamps: A-share quotes and sector/single-stock capital flows are on a 2026-09-09 close basis, read at 15:30–15:33 (Beijing); domestic futures are at the 15:00 day-session close; WTI/Brent are live quotes at 15:23 (Beijing); the 长飞光纤 (601869) H-share figure is an intraday 15:32 price and Hong Kong does not close until 16:00, so that figure is not a closing price.
3. ⚠️ This report does not include today's Dragon-Tiger List or institutional seat data (published only after 18:00); the capital-flow conclusions in §5 rest solely on main-force net inflow, so the evidence is weaker than in the pre-market note that includes the Dragon-Tiger List. Daily northbound net buy disclosure has been discontinued since 2024-08-19.
4. Incomplete-coverage statement: This report only reconciled the stocks listed in yesterday's pre-market list against today's limit-up pool and sector capital flows; the vast majority of the more than 5,900 A-share listings were not scanned name by name. Not being mentioned does not mean there is no opportunity, and being mentioned does not mean it has been sufficiently researched. The catalyst for the general-retail branch could not be identified in this report, and that is stated plainly.
5. Stock-specific risks: The stocks discussed are generally exposed to earnings misses, shareholder sell-downs, lock-up expiries, regulatory inquiries, verification halts, abnormal price movements, insufficient liquidity and similar risks; the STAR Market (±20%) and the Beijing Stock Exchange (±30%) additionally impose investor-suitability thresholds such as RMB 500,000 in assets, so some readers are not eligible to trade them. Price limits are tiered by board — please do not read ±20% / ±30% names with a ±10% intuition.
6. Reconciliation-basis statement: The hit-rate statistics here are based on the publicly stated conclusions of yesterday's pre-market list; the sample is only 18 directional conclusions and a single day's result is not statistically significant, so it should not be used to infer the long-run accuracy of this report series.
7. This report is generated by an automated process and may contain biases arising from information lag, supply-chain mapping errors or data-interface failures; it must not be used directly as a basis for trading. Please rely on official disclosures from the exchanges and the companies.
8. Position statement of the producer: The producer of this report holds none of the securities discussed herein and has no interest relationship with the companies mentioned.