A-Share Market Recap | 2026-07-29 Wednesday

⚠️ Risk warning (up front): This recap is a post-close information review and research record. It does not constitute investment advice and contains no ratings, target prices, position sizes or buy/sell recommendations. The "hit / wrong call" labels are only an after-the-fact reconciliation against this morning's pre-market list and do not represent any judgement about future returns.

Data basis disclosed up front (iron rule: if it cannot be found, write "no reliable data")

1. Quotes, the limit-up/limit-down pools, sector moves and capital flows are all as at the 2026-07-29 close, sourced from akshare (stock_zt_pool_em / stock_zt_pool_dtgc_em / stock_zt_pool_zbgc_em / stock_zt_pool_previous_em / stock_zh_a_spot_em / index_zh_a_hist) and the Eastmoney push2 flow interface, all measured live between 15:30 and 16:00.
2. No reliable data for today's top-trader list. The exchanges usually disclose it after 17:30–18:00; at this report's press time (about 16:00) Eastmoney's RPT_DAILYBILLBOARD_DETAILSNEW returned success:false for 2026-07-29, and the "most recent listing date" in akshare's stock_lhb_stock_statistic_em was still 2026-07-28. This section therefore contains no top-trader content for today and does not pass yesterday's data off as today's.
3. No reliable data for the single-day northbound net buy. stock_hsgt_fund_flow_summary_em returns "net buy turnover = 0" for northbound flows via the Shanghai and Shenzhen Connect — since August 2024 the exchanges have stopped disclosing real-time and single-day northbound net buying, so that 0 means "no longer disclosed", not "zero inflow", and must not be read as flat flows. Southbound data remain valid; see Section 5.
4. Limit-up / limit-down counts use the Eastmoney limit-up/limit-down pool basis (the same source as the 7/28 pre-market list, so directly comparable). One basis difference must also be noted: screening the full A-share quote table for "closing change ≤ −9.8%" gives 30 names today, while the limit-down pool has only 9 — the difference comes mainly from ChiNext / STAR Market names (±20% limits) and BSE names (±30%) where a 10% fall is not a limit-down, so the two bases must not be mixed.
5. Two-market turnover is the sum of turnover across all A-shares (RMB 2.3117 trillion), cross-checked as consistent with the turnover fields of the SSE Composite and SZSE Component.
6. Dimensions not covered by this report (honest disclosure): today's top-trader seat preferences and speculative-desk activity are entirely uncovered (the data are not disclosed); the FOMC decision will be published at 02:00 Beijing time on 7/30 and had not landed at press time, so every FOMC-related statement in Section 6 is a pre-event judgement.

---

0. Recap in One Sentence

Today was a textbook "deep-V repair after a sentiment freeze", and A-shares traded independently of Korean equities and the SOX. In the morning the tech chain kept collapsing, dragged down by Korea's KOSPI hitting a second intraday circuit breaker (closing −5.98% at 5663.24, about 40% off its June high, and the first time in history it has triggered breakers on two consecutive days); the STAR 50 touched 1608.14 (−5.05% at the intraday low) and the ChiNext Index a low of 3249.65 (−2.33%). The afternoon brought a full counter-attack, closing with the SSE Composite +0.40%, the SZSE Component +1.10%, the ChiNext Index +1.55% and the STAR 50 −0.87%, with the ChiNext Index up 3.97% from its low and the STAR 50 up 4.39% from its low. 4253 advancers versus 1214 decliners, a median gain of +1.58%, limit-ups rising from 61 yesterday to 81 and limit-downs falling from 49 to 9 — sentiment moved from freezing back to clearly warming. Today's strongest mainline was "the reallocation of risk-averse money": dairy (+7.06%), beverages and dairy (+6.90%), soft drinks (+6.50%), training and education (+5.33%), snack foods (+4.65%) and gaming (+4.49%) led, joined by an afternoon rally in the large financials (securities +2.31%, large-order net inflows of RMB 1.520 billion).

For the pre-market list, today's result was "half a slap and half a triumph", and both halves are extremely clean:

• The triumphant half — the avoidance calls went 10/10. The pre-market list classified the whole AI hardware chain (memory / optical modules / PCB / CCL / semiconductor equipment) as "avoid", and today not one of that group of 10 names beat the median: 通富微电 (Tongfu Microelectronics, 002156) and 紫光股份 (Unisplendour, 000938) both hit limit-down, 兆易创新 (GigaDevice, 603986) −6.81%, 长飞光纤 (YOFC, 601869) −6.81%, 东山精密 (Dongshan Precision, 002384) −5.73%, 生益科技 (Shengyi Technology, 600183) −4.59%. Getting the avoidance group entirely right on a broadly rising day when 74% of stocks advanced and the median was +1.58% is the most valuable output of today's pre-market list.
• The slapped half — only 2 of 11 recommendations beat the median. The top name 圣晖集成 (Shenghui Systems Integration, 603163) closed at a one-word limit-up (+10.01%), sealed at 09:25:03 with zero broken seals all day, the only high-quality delivery; 京东方 A (BOE Technology, 000725) at +2.33% just beat it. The other 9 all underperformed: 春秋航空 (Spring Airlines, 601021) −0.07%, 吉祥航空 (Juneyao Airlines, 603885) −0.36%, 农业银行 (Agricultural Bank of China, 601288) −0.43%, 工商银行 (ICBC, 601398) −0.25%, 中国建筑 (China State Construction Engineering, 601668) 0.00%, 赛轮轮胎 (Sailun Tire, 601058) 0.00%, 贵州茅台 (Kweichow Moutai, 600519) +0.08%.

There is only one core lesson to record today: the pre-market directional judgement that "risk aversion is the mainline" was right, but the judgement about "where the risk-averse money would go" was entirely wrong. The pre-market list defined defence as banks / baijiu / telecom operators, on the grounds that large state-owned banks had large-order net inflows of RMB 1.201 billion on 7/28 with volume ratios of 1.44~1.51; today the banking sector had large-order net outflows of RMB 1.874 billion (a net share of −5.26%), with ABC and ICBC both turning red, while the money that actually absorbed the risk-averse flow went to food and beverage (large-order net inflows of RMB 1.779 billion), gaming (+RMB 1.663 billion, a net share of +10.01%) and brokers (+RMB 1.520 billion) — low-positioned domestic-demand and sentiment sectors that the pre-market list did not mention with a single word. The extrapolative assumption that "yesterday's inflow direction = today's mainline" was falsified today.

Two further structural misjudgements must be owned: ① the pre-market characterisation of the day's tone was "structural, low-elasticity, defensive, and the negative increment is stronger than the positive", whereas it was in fact a broad deep-V advance with 4253 gainers — extrapolating the overseas negative (Korean equities / the SOX) linearly into A-shares underestimated A-shares' independence; ② 长鑫科技 (ChangXin Technology, 688825), listed pre-market as "avoid (not main board)", was +12.66% today on turnover of RMB 44.814 billion with large-order net inflows of RMB 3.454 billion, the highest in the whole market, turning straight from "the largest single drain on liquidity" into the biggest magnet for money — the single worst call of the day.

Tomorrow's tone: sentiment has been repaired but no mainline is established — the strength in consumption and media is "the mirror image of tech's blood loss" rather than an independent industry driver, and the first verification point tomorrow is whether it can hold up once tech stops falling. At the same time no report of the July Politburo meeting convening has been found as of press time (by convention it lands this week) and the FOMC decision is published at 02:00 tonight, so two event windows sit together ahead of tomorrow's open.

---

1. Market Overview

1.1 Indices and turnover

• Two-market turnover was RMB 2.3117 trillion (Shanghai RMB 1.0873 trillion / Shenzhen RMB 1.2091 trillion), 14.1% above 7/28's RMB 2.0258 trillion — this was a rebound on expanding volume, not a low-volume oversold bounce, which is the key support for the quality of today's reversal.
• Yesterday (7/28) for comparison: SSE Composite −1.16%, SZSE Component −4.52%, ChiNext Index −7.35%, STAR 50 −6.33%. Today both the ChiNext Index and the STAR 50 recovered part of yesterday's collapse, but the STAR 50 still failed to turn green and is the only broad index closing red — the tech chain's repair is the weakest.

1.2 Sentiment temperature

Sentiment reading: warming (clearly repaired from freezing, but not yet exuberant).

The basis and the counter-evidence are presented side by side, to avoid a one-sided narrative:

• Supporting "warming": limit-downs plunged from 49 to 9 (−82%); limit-ups +33%; a broken-seal rate of only 14.7% with high sealing quality; 74.2% of stocks advanced; volume +14.1%.
• Not supporting "exuberant" (which must be acknowledged simultaneously): yesterday's limit-ups rose only 0.65% on average today with an 18.0% promotion rate, showing relay money is not strong and the money-making effect is concentrated in first boards and new directions rather than the consecutive ladder; the highest board, 爱丽家居 (603221) on 7 boards, was a volumeless one-word board with a turnover ratio of just 0.46% all day and a RMB 343 million bid queue, an isolated high name rather than the top of a ladder; the STAR 50 still closed red. This was a day of "the retreat stopping + new directions producing first boards in numbers", not a day of a main upleg.

---

2. Reconciliation Against the Pre-Market List

2.1 Note on the reconciliation basis

With 74.2% of stocks advancing today and a median gain of +1.58%, "the stock closed green" does not mean the call was right on a day like this. This section therefore uses a uniform relative-to-median basis:

• "watch closely" (the bullish class) → a hit = beating the median of +1.58%;
• "watch only / Pass" (the bearish class) → a hit = underperforming or matching the median (i.e. the "should not be bought" assertion holds).

And to restate the self-imposed constraint set in the 7/28 recap: the Pass list asserts only "should not be bought" (a fundamental judgement), not "will fall" (a price judgement). A "❌" for a Pass name in the table below means "it rose more than the market that day, so the 'should not be bought' assertion gained no trading advantage that day", not that the fundamental conclusion has been overturned.

2.2 Name-by-name reconciliation of the 27 on the main list

2.3 Hit rates

The "avoid" group in detail (explicitly classified as avoid pre-market; not one beat the median today):

2.4 Observation zone (non-main-board) reconciliation

2.5 The review in one sentence

What was right was "which ones must not be touched"; what was wrong was "which ones would rise".

• The right calls share a common feature: verifiable negative or limiting evidence issued by the companies themselves. Aopu Optoelectronics (two denials in an exchange document) −7.59%, Yongxin Optics (a filing stating <1% itself) −8.46%, Zhangjiang Hi-Tech (a one-word board on shrinking volume with a register that had not turned over) underperforming — every judgement supported by the double basis of "hard company-level evidence + a defective trading structure" was right today without exception. The 10/10 avoidance group belongs in this category too.
• The wrong calls also share a common feature: judgements built on the extrapolative assumption that "yesterday's flow direction will persist". The defensive group (banks / baijiu / telecom operators) underperformed across the board, and the source was treating 7/28's RMB 1.201 billion of large-order inflows into large state-owned banks as a signpost for today; today banks had large-order net outflows of RMB 1.874 billion. By the same token, the oil branch had zero reaction for a second consecutive day — the pre-market list had already cut it from S to B+ and from "priority deep-dive" to "watch closely", yet it still kept 4 airline names on the list, so the downgrade was not thorough enough.

Three specific changes for the next pre-market report (executable, not slogans):

1. Delete the "yesterday's inflows → today's recommendation" pathway, or at minimum give it its own weight ceiling far below filing-grade positives. Today's evidence: banks went from RMB 1.201 billion of net inflows yesterday to RMB 1.874 billion of net outflows today — the sign flipped in a single day. Flows are a description of state, not a predictive variable.
2. For a branch where "the news has happened and the sector has had zero reaction for N days", it should be dropped from the list outright once N≥2, rather than downgraded and retained. The oil branch had zero reaction on both 7/28 and 7/29 yet occupied 4 slots on the list (Spring, Air China, Juneyao, China Southern) — those 4 slots crowded out observation slots that could have gone to new directions.
3. A positive scan of "where the money went" must be added, not just a negative exclusion of "where the money left". Today's pre-market report argued completely (and correctly) why AI hardware should be avoided, but never answered "where will the money squeezed out go?" — the answer was food and beverage, dairy, education, gaming and brokers, five directions that appear 0 times in a pre-market report of more than 8,000 characters. That is the root cause of 100% accuracy on avoidance versus 18.2% on recommendations.

---

3. Verifying Today's Mainlines

3.1 Branch reconciliation table

3.2 Lithography: the "β contamination" question left open pre-market has an answer today

This is the most valuable methodological conclusion in today's recap.

The 7/28 pre-market list observed that "the genuine chain names all fell 7~11% while all 6 limit-ups were low-turnover concept shells", but risk-auditor honestly pointed out at the time that the inference had one unresolved weakness: board-level β contamination — the genuine chain names were all on the STAR Market / ChiNext (the STAR 50 was −6.33% and the ChiNext Index −7.35% that day) while the concept shells were all on the main board (the SSE Composite −1.16%), so "genuine chain falls / concept shells rise" might just be a by-product of "STAR and ChiNext fall / main board rises". The pre-market report wrote explicitly that "there is no reliable source for same-board comparison data, and this report could not rule out that contamination".

Today's data rule that contamination out. Today the ChiNext Index was +1.55%, the STAR 50 only −0.87%, 74.2% of the market advanced and the median was +1.58% — the β environment reversed completely. And the result was:

Conclusion: in an environment where ChiNext was +1.55% and the whole market rose broadly, the genuine lithography chain averaged −6.10% and the concept shells −4.82%, with both groups falling heavily together. This shows that 7/28's "genuine chain falls" was not board β but the lithography theme itself retreating — the β hypothesis has been rejected by today's data. Yesterday's 6 concept-shell limit-ups left not one behind today (0 limit-ups); the theme jumped straight from "fermenting" to "retreating" with no divergence phase in between.

One piece of pre-market wording must also be acknowledged as conservative in hindsight, though directionally right: the pre-market rating was "theme strength A, stock-level delivery C", leaving the theme itself at A. Today proves the theme strength does not hold either — a theme driven by a secondary foreign-media item (The Information → Reuters → domestic media), with no domestic official or company filing following up, has a life cycle of exactly 1 trading day. This should be written into tomorrow's pre-market scoring basis.

3.3 An unexpected mainline: passive components / MLCC — one the pre-market list missed entirely

Today's largest industry large-order net inflow was not consumption but passive components (+RMB 2.594 billion, a net share of +8.32%), and the pre-market list had zero coverage of it.

• Performance: 风华高科 (Fenghua Advanced Technology, 000636) hit limit-up with a bid queue of RMB 415 million, the thickest in the market, a 14.92% turnover ratio and large-order net inflows of RMB 1.838 billion (2nd among all stocks); 三环集团 (Sanhuan Group, 300408) +7.99% with large-order net inflows of RMB 801 million; 江海股份 (Jianghai Capacitor, 002484) and 国瓷材料 (Sinocera, 300285) +4.79% followed.
• The driver (its timeliness must be labelled clearly): this is not news from today but the second fermentation of an industry trend already running for months — since April 2026 domestic MLCC channel prices have risen sharply, with some high-specification models up 3–5 times from the start of the year; Yageo raised prices across its entire capacitor range from 1 July 2026; and Japanese and Korean majors have raised MLCC prices for AI servers by 15%~35% (36Kr¹, China Fund News²). This report could not verify any new price-increase letter or company filing today, so the question of "why today" has no reliable data.
• An internal contradiction that must be pointed out (this is tomorrow's key verification point): the core driver of MLCC price increases is the explosion in AI server demand, while the market today is heavily selling off the entire AI hardware chain (electronics had large-order net outflows of RMB 6.440 billion). That is: money is selling the mid- and downstream of AI hardware while buying the upstream passive components. These two things cannot both be right for long. What must be verified tomorrow is whether this is a rotation within the AI chain "from expensive downstream to price-raising upstream" (sustainable) or purely a risk-averse catch-up within the sector (unsustainable). This report does not pre-judge the direction.
• Risk (not to be ignored): Fenghua Advanced Technology hit a one-word limit-down at the end of June 2026 over a "Rashomon" of MLCC rumours, while Sanhuan Group publicly denied them over the same period (JRJ³) — this branch has a documented history of falsified rumours, so the verification bar for news should be higher than for an ordinary theme.

3.4 Shenghui delivered, but the branch did not diffuse — an important distinction

圣晖集成 (603163) hit a one-word limit-up at +10.01%, a clean delivery for the pre-market rank 1. But all three other names in the same branch fell: 亚翔集成 (603929) −0.78%, 柏诚股份 (601133) −3.47%, 太极实业 (600667) −5.28% (a 20.58% turnover ratio and RMB 7.378 billion of turnover, clearly falling on expanding volume).

This shows the market priced the RMB 858 million order as "a Shenghui company event", not as "a recovery in cleanroom industry conditions". The pre-market report had in fact laid the correct groundwork, writing explicitly that "a single plant order is not enough to overturn an industry judgement; it is only a counter-example to track". Today verified that sentence: the order is real, the industry extrapolation is not. Other names in the same branch should not be recommended on this basis tomorrow.

One tracking item also remains outstanding: no filing about the formal contract signature for Shenghui Systems Integration was found before press time, so the company's own statement that "no formal contract has yet been signed and the terms remain uncertain" still stands even after the limit-up.

3.5 Software / IT localisation: under the rule the pre-market report set for itself, this branch should not be deleted

The pre-market report wrote an explicit self-test rule for this branch: "it only holds if a first consecutive board appears today, otherwise it should be deleted from the branch list".

Today's result: 天融信 (002212) posted 2 consecutive boards (+10.00%, a RMB 116 million bid queue, a 6.44% turnover ratio) — the first consecutive board appeared.

Under the rule the pre-market report set for itself, this branch should be retained rather than deleted. This must be executed honestly even where the conclusion is unfavourable to the pre-market report — falsification conditions a report sets for itself cannot be quietly changed just because the result is inconvenient.

The other half of the facts must be presented alongside, to avoid reading it as "the branch has turned around": internal divergence is severe — 格尔软件 −1.78%, 岩山科技 −2.05% and 电科数字 +1.11% all underperformed the median; 南天信息 rose +2.77% but its seal broke once intraday on a turnover ratio as high as 11.10%; 浪潮软件 rose +3.18% but on a 20.15% turnover ratio. In other words: the branch is alive, but the only thing alive is Topsec, while the rest bleed out through turnover. Nothing today overturns any of the pre-market fundamental conclusions about these companies (4 posted single-quarter 2026Q1 losses, and not one has disclosed a verifiable IT-localisation revenue breakdown).

3.6 Did the pre-market report identify the strongest mainline?

No. And it was not a mis-ordering — the whole direction was absent from the list.

• Of the 8 branches given pre-market, not one of today's genuine leaders — consumption, education, gaming, brokers and passive components — appeared.
• The branch ranked 1 pre-market (cleanroom) delivered but did not diffuse; the 3rd (oil) had zero reaction for a second consecutive day; the 4th (bank / baijiu defence) had the right direction and entirely the wrong vehicles; the 5th (lithography) retreated outright; the 8th (AI hardware) was entirely correct as an avoidance call.
• The closest thing to correct was the abstract judgement of "risk aversion"; the most wrong was making risk aversion concrete as banks.

---

4. The Limit-Up Ladder / Highest Boards

4.1 Ladder structure (81 names in the limit-up pool)

Ladder assessment: badly fractured, missing the middle layer. It jumps straight from 7 boards to 3 boards, with no 4/5/6 boards; consecutive-limit-up names number only 11, 13.6% of all limit-ups, so the other 86.4% are first boards. This is the classic structure of "new directions just starting while an old high name hangs alone", not a healthy ladder of continuous promotion.

• The quality of the highest board, 爱丽家居 (603221), is doubtful: 7 boards but a turnover ratio of just 0.46% and total turnover of only RMB 21 million (its RMB 343 million bid queue is 16 times its turnover) — a volumeless one-word board. The pre-market Pass list had already noted that "it traded only RMB 13 million on 7/28 with a 0.33% turnover ratio, so there is no absorption once the seal breaks", and the structure is unchanged today, so the relay risk is not cleared.
• Turnover diverges enormously within the 3-board tier: Chuanzhi Education at a 3.74% turnover ratio (a well-locked register) versus Xingxin New Materials at 23.86% (violent divergence) — the same 3 boards, different quality.

4.2 Promotions and broken boards

• 61 limit-ups yesterday → 11 still at limit-up today, a promotion rate of 18.0%, with an average gain of +0.65%.
• The names that broke their boards and fell heavily are concentrated in lithography and thematic shells: 创新医疗 (Innovative Medical Management, 002173) −10.02% (limit-down), 波长光电 (301421) −8.30%, 永新光学 (603297) −8.46%, 奥普光电 (002338) −7.59%, 海南海药 (Hainan Haiyao, 000566) −7.20%, 南京熊猫 (Nanjing Panda Electronics, 600775) −6.96%.
• 14 broken seals, a broken-seal rate of 14.7% (on the low side, showing reasonable willingness to seal). Among them, note that 北京文化 (Beijing Culture, 000802) broke its seal 4 times and 银龙股份 (Yinlong, 603969) 3 times, while 南天信息 (000948) and 卓翼科技 (Zhuoyi Technology, 002369) broke theirs once each.

4.3 Identifying the core leaders (large turnover + rising + recognisable)

A structural fact worth pointing out: today's top two by turnover (ChangXin Technology at RMB 44.8 billion and Innolight at RMB 43.5 billion) were both classified in the pre-market list as "avoid" or "not main board, no primary recommendation", and both rose sharply. Among the main-board names available, the largest-turnover gainers were Fenghua Advanced Technology and 京东方 A (000725, RMB 9.723 billion). The "main board first" iron rule carried an obvious cost in today's market structure — this report does not argue for changing it (it is a risk-control rule, not a stock-selection rule), but the pre-market report should explicitly label "how today's opportunity distribution differs between main board and non-main board", so readers know for themselves what has been filtered out.

---

5. Capital Flows

5.1 Northbound / southbound

• Northbound flows: no reliable data. stock_hsgt_fund_flow_summary_em returns "net buy turnover = 0.00" for northbound flows via the Shanghai and Shenzhen Connect — since August 2024 the exchanges have stopped disclosing single-day northbound net buying, so a 0 in that field means "no longer disclosed" and must not be read as "zero inflow". This report cannot verify the source or basis of the "northbound net inflow of RMB XX" readings circulating on several financial channels today and does not cite any of them.
• Southbound flows (valid data): the Shanghai Stock Connect southbound leg had net selling of RMB 3.941 billion and the Shenzhen leg RMB 1.330 billion, about RMB 5.271 billion in total, while the Hang Seng Index was +1.8% the same day. That is: Hong Kong stocks rose while southbound money flowed out, with mainland money choosing to take profits into Hong Kong strength — consistent with the direction of capital returning in A-shares today.

5.2 Industry large-order net inflows / outflows (Eastmoney basis, full day)

Three key readings:

1. Banks are the textbook case of "the index is green, the money is leaving": the sector was +0.70% yet had large-order net outflows of RMB 1.874 billion at a net share of −5.26%. The pre-market defensive logic not only failed to deliver today; money was actively retreating in the opposite direction.
2. Computing and IT services are even more extreme cases of "the index is green, the money is leaving heavily": computing +0.81% with net outflows of RMB 4.913 billion, and IT services +0.12% with net outflows of RMB 3.836 billion (a net share of −7.79%, the most negative in the market). This confirms the Section 3.5 judgement — the software sector's broad advance today was driven by retail and speculative money while the large money used the bounce to distribute. Topsec's 2 consecutive boards must be read against that backdrop.
3. Digital chip design had net inflows of RMB 1.515 billion against the trend, the opposite direction to semiconductors' overall net outflow of RMB 5.337 billion — divergence has already appeared inside semiconductors, and it is no longer an indiscriminate sell-off. This is a signal to track tomorrow (see 6.1).

5.3 Stock-level large-order flows

贵州茅台 (600519) deserves a separate note: large-order net inflows of RMB 552 million ranked 7th in the market, yet the share price rose only 0.08%. Big money was buying while selling pressure of equal size was coming out — the classic "institutional rebalancing absorption zone", not a launch pattern. Listing it as a defensive name pre-market was not directionally wrong, but there was no price delivery on the day.

5.4 Top-trader list

No reliable data for today's top-trader list. At press time (about 16:00) the exchanges had not disclosed the 2026-07-29 list, the Eastmoney interface returned success:false for that date, and the "most recent listing date" in akshare's statistics table was still 2026-07-28. This section does not substitute yesterday's data for today's, and cites no unverified seat rumours.

One thing to add to tomorrow's pre-market work: refill today's top-trader list after 18:00, focusing on three things — ① the nature of the buying seats behind 天融信 (002212)'s 2 consecutive boards (institutional vs speculative, which determines whether it is a coordinated push or a single speculative desk's solo act); ② the seats behind 风华高科 (000636)'s RMB 415 million bid queue (to judge the capital tier of the passive-components branch); ③ whether the food-and-beverage limit-up cluster (Yiming Food, Ziyan Foods and others) includes institutional seats — if the consumption limit-ups are all speculative money with no institutions, that mainline's persistence must be discounted heavily.

---

6. Outlook for Tomorrow (2026-07-30 Thursday)

Important preface: two event windows have not yet landed before tomorrow's open, and every judgement in this section assumes both outcomes are unknown.

• The FOMC decision: published at 02:00 Beijing time on 7/30 (not landed at press time).
• The July Politburo meeting: as of press time (about 16:00) no authoritative report of it having convened has been found. By convention it is held at the end of July to arrange second-half economic work, could land at any time, and is directionally two-way. This item has now sat in "pending" status for a second consecutive day (Caixin preview⁴). If the readout lands before tomorrow's open and mentions expanding domestic demand or promoting consumption, today's consumption mainline upgrades from "risk-averse rotation" to "policy + capital resonance"; if the readout emphasises tech self-reliance, today's tech sell-off could reverse. This is tomorrow's single largest variable.

6.1 Judgement on mainline continuation

6.2 Key focus tomorrow (branch + main-board stock + verification point)

The following are research priorities, not investment ratings, and carry no buy or sell meaning.

6.3 What to avoid tomorrow

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

Six things that must be carried into tomorrow morning's pre-market report:

1. Delete the "yesterday's inflow direction → today's recommendation" pathway. Today's hard proof: banks went from RMB 1.201 billion of net inflows on 7/28 to RMB 1.874 billion of net outflows on 7/29, a sign reversal in one day. Put flows in the "description of the present", not in "reasons to recommend".
2. Add a mandatory section, "where did the money go". The gap between today's 100% avoidance accuracy and 18.2% recommendation accuracy stems from the pre-market report doing only negative exclusion without a positive scan. Tomorrow morning it must answer: the money leaving AI hardware went today into passive components (+RMB 2.594 billion), food and beverage (+RMB 1.779 billion), gaming (+RMB 1.663 billion) and brokers (+RMB 1.520 billion) — is it still there tomorrow?
3. Enforce the drop-from-list rule newly set in Section 2.5: news has happened and the sector has had zero reaction for ≥2 consecutive days → drop it from the list outright. The oil / airline branch should be dropped tomorrow morning, with the freed slots given to the new directions verified today.
4. The AI hardware "avoid" must change from indiscriminate to tiered. Divergence signals have already appeared today (digital chip design +RMB 1.515 billion, Innolight +4.74%, ChangXin Technology +12.66%), and continuing to avoid indiscriminately would repeat today's error of missing ChangXin Technology's +12.66%. It is suggested that it be split into "the price-increase beneficiary end" and "the de-rating end" and rated separately.
5. How to handle the two event windows: the FOMC lands at 02:00 tonight and the Politburo meeting could land at any time. Tomorrow morning's pre-market report must verify the actual outcomes of both before setting the branch ordering; carrying over today's ordering is prohibited. And per pre-market iron rule ①-b, when citing overnight overseas data, the trading day the data correspond to must be displayed (Korea's KOSPI closed today at 5663.24 / −5.98%, which is the same session as A-shares, not overnight).
6. Refill today's top-trader list after 18:00 (see the three priorities in 5.4), especially the nature of the seats behind Topsec's 2 consecutive boards and whether institutions took part in the consumption limit-up cluster — the latter directly determines what strength the consumption mainline should be given tomorrow.

Finally, one revision suggestion for the scoring model: the "theme persistence" dimension in Section 4 of that report currently reads "an existing consecutive-limit-up ladder 7–10 pts". Today the lithography theme went from 6 limit-ups to 0 in a single trading day, and its primary source was a secondary foreign-media relay (The Information → Reuters → domestic media). It is suggested that a clause be added: for a theme whose primary source is a secondary foreign-media relay with no domestic official or company filing following up, the "theme persistence" dimension is capped at 4 pts.

---

⚠️ Risk warning: this recap is only a post-close information review and observation and does not constitute investment advice. Data may differ in timeliness or basis; the exchange's and companies' disclosures prevail, and this must not be used directly as a basis for trading.