A-Share Market Recap | 2026-08-06 Thursday

Data-basis disclosure first (please read this section before anything else)

1. Quote basis: all A-share single-stock closing prices, percentage changes, turnover rates, traded value, 20-day gains and main-force net inflows in this report are as of the 2026-08-06 close, taken from the Tencent Finance single-stock snapshot (stock_zh_a_spot_tx, 5538 records). Main-force net inflow is on the East Money basis (large + extra-large orders), which may point in the opposite direction from the Tonghuashun "active buy / active sell" basis; the two must not be mixed or added together.
2. The limit-up pool / failed-limit-up pool / limit-down pool / consecutive-limit-up ladder are taken from East Money (stock_zt_pool_em etc.); measured values were limit-ups 79, failed limit-ups 20, limit-downs 1.
3. Industry percentage changes and industry net inflows are taken from the Tonghuashun industry basis (90 industries). Industry net inflow and single-stock main-force net inflow belong to two different bases and must not be added together.
4. Price-limit bands differ by board, so every stock in this report is tagged with its board: SSE/SZSE main board ±10%, ChiNext (30/301) ±20%, STAR Market (688) ±20%, BSE ±30%. On ChiNext/STAR, +12% is not a limit-up — please do not read a ±20% stock with ±10% intuition.
5. Items not obtained / left blank in this report:
   • 2026-08-06 Dragon-Tiger List: not included in this report. The East Money Dragon-Tiger List is only published around 18:00 after the close, while this report closed its copy at 16:00; all four interfaces returned no data. This report therefore lists no Dragon-Tiger seats and draws no "hot-money preference" inference from them.
   • Single-day northbound net inflow: cannot be provided. Since August 2024 the exchanges have discontinued real-time/post-close disclosure of single-day northbound net buying (measured this time: both Shanghai Connect and Shenzhen Connect returned "net buy turnover" of 0 with trading status 3). This is a gap in the disclosure basis, and this report does not pass off any substitute data as northbound flow.
   • New same-day catalyst for the coal sector: not found. This report treats it as "no new hard news on the day"; see Section 3.

⚠️ Data-pull and implementation notes for this run (internal only):

• The East Money whole-market snapshot stock_zh_a_spot_em threw ConnectionError three times in a row (RemoteDisconnected; retries at 5-second intervals did not help), and stock_zh_index_spot_em failed the same way. Following existing experience I switched straight to the Tencent snapshot stock_zh_a_spot_tx (5538 rows) + the Sina index feed stock_zh_index_spot_sina; both backup channels succeeded on the first try. The East Money interfaces for the limit-up / failed-limit-up / limit-down pools were themselves fine, which shows that what went down was the quote-snapshot group of interfaces, not East Money as a whole.
• I stepped on a trap while writing the script myself, worth recording: when converting to numeric I excluded columns with for c in df.columns: if c not in ('代码','名称'), but the Tencent snapshot's code column is named in English, code (values look like sh600489), so it was not excluded and the entire column was turned into NaN by pd.to_numeric — the first version of the reconciliation table simply returned an empty table. It has been changed to str.extract(r'(\d{6})') to pull the digits first and then exclude. This is the second trap on the same interface as the one recorded on 8/5, that "the Tencent snapshot's numeric columns are strings, and sorting them as strings produces a fake money-flow ranking" — neither its column names nor its column types can be taken for granted.
• Units were verified by measurement, not guessed: in the Tencent snapshot turnover is in units of RMB 10,000, zljlr is in units of RMB 10,000, and ltsz/zsz are in units of RMB 100 million. The verification method was to take 有研新材 (Grinm Advanced Materials, 600206) with turnover=441333 (units of RMB 10,000 → RMB 4.413 billion) and check it against the East Money limit-up pool's 成交额=4413325568 (RMB → RMB 4.413 billion) — an exact match. This kind of cross-source unit alignment must be done once, otherwise RMB 4.4 billion gets written up as RMB 441.3 billion.
• All four Dragon-Tiger List interfaces failed: stock_lhb_detail_em and stock_lhb_jgmmtj_em both reported TypeError: 'NoneType' object is not subscriptable (indicating an empty response body rather than a network fault), and stock_lhb_detail_daily_sina reported KeyError. The root cause is timing, not the interfaces — the East Money Dragon-Tiger List is published around 18:00, and this report closed its copy at 16:00. stock_lhb_stock_statistic_em(symbol="近一月") returned 858 rows normally, which serves as counter-evidence. Recommendation: if the scheduler pins the post-close recap to 15:30–16:30, the Dragon-Tiger List will never be obtainable; either push that slot back past 18:30, or hard-code into the template that "the day's Dragon-Tiger List has not been published" and switch to the past-month statistics instead. This is a process-design problem, not a same-day failure.
• Sina's AI-generated copy stitched old news into a "reason for today's limit-up" — the same pattern for the third consecutive trading day, this time on 长电科技 (JCET, 600584). Two AI pieces (doc-inimivft0808249, doc-inimivfm4741737) attributed the limit-up to "AI compute demand + high-end packaging-and-test positioning + earnings pre-announcement growth". Checking item by item: the interim earnings pre-announcement was published on 2026-07-15 (net profit attributable to parent RMB 770–950 million, +63.48%+101.70% year on year; ex-non-recurring RMB 740–910 million, +68.95%+107.76% year on year); the RMB 7.8 billion Lingang advanced packaging-and-test project was announced on 2026-06-25. Both items are old news. This report therefore does not list 长电科技 (JCET) as a news-driven name, citing it only as evidence of the day's largest money flow. This is the third consecutive day of refusing to accept this kind of after-the-fact attribution (8/5 工业富联 (Foxconn Industrial Internet), 8/6 pre-market 工业富联, today 长电科技). → Recommendation: harden this into a process assertion — whenever citing "the reason a stock hit the limit today", first check the announcement date of the cited reason; if the publication date does not fall within the day's trading session, downgrade it to "money-flow behaviour" without exception. This trap now shows up every day; it is not sporadic.
• There is a cross-source conflict on 长电科技's traded value: the AI copy says "total traded volume RMB 10.017 billion", while the locally measured Tencent snapshot gives turnover=1726046 (units of RMB 10,000) = RMB 17.26 billion. This report uses the measured value of RMB 17.26 billion and does not cite the AI copy's turnover figure. Also note that a 6/29 piece returned by search claims it was "up more than 150% year to date, from RMB 32 to RMB 105", whereas the measured close today is RMB 75.87 with a 20-day change of −26.71% — that piece is badly out of date, and citing it directly would turn "a rebound after a deep decline" into "acceleration from a high", exactly inverting the characterization.
• The catalyst for the coal storyline could not be found, so this report states truthfully that "no new hard catalyst was found for the day" and does not force-fit one. The Securities Times piece of 8/6 (stcn.com/article/detail/4061264.html) was confirmed by WebFetch to be published that day and to describe that day's trading, but the body only describes the phenomenon and gives no attribution at all. Searching for "动力煤价格 2026 年 8 月" (thermal coal prices, August 2026) returned nothing but March/June-vintage industry reports and undated price pages. With no same-day catalyst obtainable, writing an annual-horizon storyline such as "summer peak demand + supply tightening" as "the reason for today's rise" would be fabrication, so Section 3 records only the money-flow characteristics (low turnover rate, large caps, no consecutive limit-ups) and no reason.
• The reconciliation basis was defined before it was computed, not cherry-picked: the constituents of the four tiers (priority deep-dive / watch closely / watch only / Pass) were taken directly from the pre-market Section 3 master list and Section 6 Pass list, fixed before today's trading was seen; means/medians are equal-weighted across each full tier, with no outliers removed. The Pass tier having the highest mean (+3.35%) — a result unfavourable to us — was produced this way, not picked afterwards. See the self-critique in Section 2.
• yfinance was not used at any point in this run (A-share data always goes through akshare/Tencent); skipped directly per existing experience, with no wasted trial and error.

---

0. One-Line Recap

Sentiment temperature: warming, but a "structural warming" — the money-making effect improved while the indices did not follow. The failed-limit-up rate fell from yesterday's 29.5% to 20.2% (limit-ups 79 / failed limit-ups 20 / limit-downs 1), the healthiest reading in recent sessions; but the number of limit-ups fell from 103 to 79, combined turnover across the two exchanges shrank from RMB 2.68 trillion to RMB 2.53 trillion, and the Shenzhen Component −0.24% and the ChiNext Index −0.55% both closed in the red, with only the SSE Composite +0.57% and the STAR 50 +0.45% in the green. Across the whole market 2789 advanced / 2590 declined, with a median gain of just +0.05% — this was a day of "single-stock divergence taken to the extreme while the indices barely moved".

The strongest storyline was not the precious metals the pre-market brief bet on, but the AI-hardware chain of [semiconductor packaging-and-test / PCB / components] — and that is precisely the direction pre-market Section 9.3 explicitly said "we do not recommend chasing". 长电科技 (JCET, 600584, SSE/SZSE main board) hit the limit, traded RMB 17.26 billion, with main-force net inflow of RMB 2.297 billion, the largest in the whole market; 景旺电子 (Kinwong Electronic, 603228, SSE/SZSE main board) hit the limit with main-force net inflow of RMB 1.06 billion; 方正科技 (Founder Technology, 600601, SSE/SZSE main board) +9.96% with net inflow of RMB 975 million; 胜宏科技 (Victory Giant Technology, 300476, ChiNext) +5.84% with net inflow of RMB 890 million. Pre-market judged this group as "more likely to gap up and dump than to continue"; in reality they rose across the board and delivered the day's largest money inflow. This is the one error today that most deserves to be owned.

Pre-market list hit summary: half the direction right, the ranking essentially void. The pre-market's first branch [precious metals] rose only 2.66% today (versus +7.72% yesterday), with zero limit-ups in the entire branch, and the anchor 紫金矿业 (Zijin Mining, 601899) up just +1.17% with main-force net outflow of RMB 431 million — the branch did not break, but it clearly decelerated and entered divergence. The second branch [memory] stalled completely: 兆易创新 (GigaDevice, 603986) −0.11%, 佰维存储 (Biwin Storage, 688525) −0.06%, 长鑫科技 (ChangXin Memory, 688825) −4.31% with main-force net outflow of RMB 3.003 billion, the largest in the whole market (second day running). Uglier still is the tier breakdown: the 9 Pass-tier names averaged +3.35% today, above the "priority deep-dive" tier's +2.75% — on a mean basis the pre-market's four-tier ranking was inverted. (On a median basis the ranking still holds; the divergence between the two bases is itself today's single most important conclusion — see Section 2.)

Tomorrow's keynote: switch from "betting on branches" back to "following the money". Precious metals are downgraded from "strongest branch" to divergence watch, memory is clearly ebbing, and the two lines the pre-market did not write up — packaging-and-test/PCB/components, and coal — need to be tested tomorrow on "whether there is a second day". What most needs guarding against tomorrow is not missing out, but treating the first day's huge net inflow into a name like 长电科技 (JCET, 600584) — extremely strong money with no same-day new catalyst — as a guarantee for the second day.

---

1. Market Overview

1.1 Indices and Turnover

• Combined turnover across the two exchanges was about RMB 2.53 trillion, a contraction of about 5.6% from yesterday's RMB 2.68 trillion.
• Index divergence was clear: the SSE Composite and STAR 50 closed green while the Shenzhen Component and ChiNext Index closed red. This was not a broad decline but a difference in weighting structure — the SSE Composite was pulled up by large-cap names in coal and non-ferrous metals, while ChiNext was dragged down by previously strong names in memory, optical modules and consumer electronics.
• Across the whole market 2789 advanced / 2590 declined / 159 unchanged (all-A basis, 5538 names in the Tencent snapshot), with a median gain of +0.05% and a mean of +0.56%. Advancers slightly outnumbered decliners, but the near-zero median shows that the gains were concentrated in a few sectors while most stocks went nowhere.

1.2 Limit-Up Ladder and Sentiment Indicators

• Limit-ups by board: SSE/SZSE main board (±10%) 70 names, ChiNext (±20%) 7 names, STAR Market (±20%) 2 names. Limit-ups were highly concentrated on the main board, consistent with today's leaders in coal, packaging-and-test and components being mostly mid- to large-cap main-board names.
• Distribution of first limit-seal times: 09:00 hour 41 names, 10:00 hour 16, 11:00 hour 6, 13:00 hour 12, 14:00 hour 4. More than half of the limit-ups were completed within the first hour of the morning session, with limited additions in the afternoon — a structure of "surge in the morning, converge in the afternoon", meaning sentiment intensity is not as good as the limit-seal count makes it look.

1.3 Sentiment Temperature Assessment

Verdict: warming, but of the "quality-improvement" type rather than the "intensity-increase" type.

• The improved side (real): the failed-limit-up rate fell from nearly three-tenths to two-tenths and there was only 1 limit-down, showing that limit-seal quality improved markedly and the money-losing effect converged.
• The un-improved side (also real): 24 fewer limit-ups, turnover down 5.6%, a median gain of only +0.05%, and both major growth indices closing red. The money-making effect became "steadier", but the range in which money was made became "narrower".
• One contradictory signal that must be flagged separately: the highest consecutive-limit-up count rose from 9 boards to 10 boards (爱丽家居, Ailly Home, 603221, SSE/SZSE main board, 11 limit-ups in 21 days), yet this very stock itself announced on the evening of 8/5 an expected interim net loss of RMB 40.50 million–34.50 million. The top of the ladder being carried by a stock that has explicitly warned of a loss is not evidence of healthy sentiment; it is evidence that money has nowhere else to go.

---

2. Reconciliation Against the Pre-Market List

Reconciliation rules stated up front: the constituents of the four tiers are taken directly from the pre-market Section 3 master list and Section 6 Pass list, fixed before today's trading was seen; "delivered or not" is judged against the conclusion basis given pre-market, not reverse-engineered from "did it go up today". The whole-market median gain of +0.05% is the baseline.

2.1 Pre-Market Top 5 Recommendations (pre-market Section 9.1)

Top 5 hit rate: 2/5 = 40% on a "same-day gain > 3%" basis; 3/5 = 60% on a "closed up" basis. Both bases are listed because in a market with a median of only +0.05%, "up a little" is not enough to count as a hit.

2.2 Four-Tier Breakdown (the one table most worth looking at today)

The two bases give opposite conclusions, and that contradiction is itself today's core finding:

• On a median basis the pre-market ranking holds: priority deep-dive +3.53% > Pass +2.30% > watch closely +1.13% > watch only +0.85%. The four tiers decline monotonically (Pass excepted), showing the ranking judgement about the "typical stock" was valid.
• On a mean basis the pre-market ranking is inverted: the Pass tier at +3.35% overtakes priority deep-dive at +2.75%. The cause lies entirely in three pure money-game names inside the Pass tier — 云南锗业 (Yunnan Germanium, 002428, SSE/SZSE main board) 3rd consecutive limit-up, +10.00%, 光智科技 (Guangzhi Technology, 300489, ChiNext) +10.72%, and 爱丽家居 (Ailly Home, 603221, SSE/SZSE main board) 10th consecutive limit-up, +9.98%. All three hit or came within a hair of the limit.

2.3 Self-Critique: What Was Right, What Was Wrong, How to Change Tomorrow Morning

Right (three items, none of them luck):

1. Every single risk warning about the memory branch hit. Ranking memory as the No. 2 branch pre-market was a ranking error, but the graded judgement about it was correct: "already-realized earnings confirmed, forward guidance below expectations", "the leader is absent", "the expectation gap is negative". Today 兆易创新 (GigaDevice, 603986) and 佰维存储 (Biwin Storage, 688525) both closed green-side down, and 长鑫科技 (ChangXin Memory, 688825) was −4.31% with main-force net outflow of RMB 3.003 billion, the largest in the whole market (second day running), fully confirming the pre-market line that "a branch with no anchor deserves a discount on persistence".
2. The ebb warning on 盛达资源 (Shengda Resources, 000603) hit with great precision. Pre-market verbatim: "if it gaps up, pushes higher and then fades today, that is the first ebb signal to appear within the branch". What actually happened: it touched the limit price of RMB 32.07 intraday, failed the limit 18 times, and finally closed at RMB 31.30, +7.38%, with the turnover rate doubling from yesterday's 7.67% to 18.15%. Up 7.38% and yet the most textbook divergence pattern in the branch — looking only at the percentage change would have misread this one as "the strongest".
3. The characterization of precious metals as "a rebound, not a new high" was not overturned by today's trading. The branch rose only 2.66%, had zero limit-ups, and its anchor saw net outflow — consistent with the pre-market characterization that "the catalyst is real, but it sits on a medium-term downward price curve".

Wrong (three items, and they must be written out clearly):

1. [The one most deserving to be owned] Listing the entire AI compute-hardware chain as "not recommended to chase" was a misjudgement. The pre-market reasoning was "Musk's remarks generate no revenue + overnight AMD −7.04%, SpaceX −13.61%, Nasdaq −0.83%, so gapping up and dumping is more likely than continuation". Actual result: 胜宏科技 (Victory Giant Technology, 300476, ChiNext) +5.84%, 生益科技 (Shengyi Technology, 600183, SSE/SZSE main board) +3.63%, 工业富联 (Foxconn Industrial Internet, 601138, SSE/SZSE main board) +3.56%, 沪电股份 (WUS Printed Circuit, 002463, SSE/SZSE main board) +1.38%, 深南电路 (Shennan Circuits, 002916, SSE/SZSE main board) +1.38% — all five rose and none dumped; moreover the top of the inflow ranking was occupied by names from this chain. The error was not in reading the news but in treating "overnight offshore sentiment" as the determining variable for A-share performance that day. The explanatory power of a 0.83% overnight Nasdaq decline over an independent A-share storyline was overrated.
2. [The most systemic one] "Not recommended for participation" was written in a way that implied "will weaken". 云南锗业 (Yunnan Germanium, 002428), 光智科技 (Guangzhi Technology, 300489) and 爱丽家居 (Ailly Home, 603221) on the Pass list all hit or came within a hair of the limit today. The Pass reasons given pre-market (P/E 2,835x, P/E −927, company-guided loss) were all factually accurate, and not one of them was overturned today — but those reasons answer "is it worth buying", whereas what drives this kind of stock on the day was never valuation but chips and sentiment. Taking a valuation conclusion directly as a price forecast is this report's single most systemic methodological error.
3. The day's two strongest lines were missed: coal and semiconductor packaging-and-test. Coal mining and processing was +4.88%, ranked first among all industries, with 7 limit-ups, and was not mentioned anywhere in the pre-market brief; 长电科技 (JCET, 600584) hit the limit and took the largest main-force net inflow in the whole market, RMB 2.297 billion, yet was not among the 25 names on the pre-market master list. The reason for missing them is that the pre-market stock-selection entry point is "derive branches from overnight news", and neither of these lines had news driving it that day — both were money re-allocating on its own. A purely news-driven entry point is structurally blind to them.

Three concrete changes for tomorrow morning's pre-market list:

• ① Add a "money-driven candidate" tier alongside the "news-driven candidate" tier. Of today's 79 limit-ups, 70 were on the main board, 7 in coal, and several in packaging-and-test/components; what they have in common is no new news on the day. Suggested extra pre-market step: take the previous session's stocks that are in the top 20 by main-force net inflow and have a negative 20-day change as a separate observation pool — 长电科技 (JCET, 600584, 20-day −26.71%) and 方正科技 (Founder Technology, 600601, 20-day −15.71%) would both be caught by this sieve.
• ② Enforce a wording distinction in the Pass-list conclusions: split "Pass" into "not recommended for participation (valuation/fundamentals do not support it)" and "expected to weaken (with explicit evidence)". Today's three Pass names hitting the limit belongs to the former and is not a judgement error; but with pre-market lumping them under the same "do not chase" heading, readers will read it as the latter.
• ③ The weighting on overnight offshore markets must be cut. For two consecutive days (8/5, 8/6) we saw "overnight US tech selloff → the corresponding A-share sector rises anyway". Recommendation: downgrade overnight offshore markets from a "directional criterion" to an "opening-volatility note", and stop issuing sector-level avoid conclusions on that basis.

---

3. Verification of Today's Storylines

3.1 Did the Pre-Market Brief Call the Strongest Storyline: No

Pre-market listed [precious metals] as branch No. 1 and [memory] as No. 2. Today's actually strongest money storyline was [semiconductor packaging-and-test/PCB/components], followed by [coal] and [electronic chemicals]. Two of those three were not written up pre-market at all, and the third was written into "do not chase".

But one detail deserves to be recorded fairly: pre-market judged that "the catalyst for the precious-metals branch is real and not yet priced by A-shares" — today precious metals still ranked 6th among all industries at +2.66% with 12 up and 2 down, so the direction was right; what was wrong was the intensity and the ranking. The branch has not broken down, it is simply nowhere near enough to support a "branch No. 1" designation.

3.2 The Unexpected Storyline the Pre-Market Missed: Coal

Coal mining and processing was +4.88%, first among the 90 industries, with 7 limit-ups and industry net inflow of +RMB 1.433 billion (Tonghuashun industry basis). But its money-flow characteristics are completely different from those of a thematic stock, and this must be spelled out or it will be easy to chase it as a theme tomorrow:

• Turnover rates are generally only 1.5%–2.5%, yet the limits held — this is large-cap, low-turnover, institutional-allocation buying, not a hot-money relay.
• 20-day gains are all positive (+10%~+21%), showing that coal did not start today; it is a trend name that has already been rising for a month, and today was merely acceleration.
• All 7 limit-ups were first boards with zero consecutive limit-ups, and 郑州煤电 (Zhengzhou Coal Industry & Electric Power, 600121) and 晋控煤业 (Jinkong Coal, 601001) additionally failed their limits.
• No new hard catalyst for the day was found: the Securities Times piece of 8/6 was verified to describe the phenomenon only, with no attribution; the "summer peak demand, supply tightening, lower imports" items returned by search are all annual-horizon industry logic from March/June and cannot be treated as the reason for today's rise.

This report therefore characterizes coal as: an acceleration day for a trend name, not the launch day of a theme. The two call for exactly opposite playbooks tomorrow — a pullback in a trend name can be looked at, whereas a first-board theme name has to be guarded against a failed limit on day two.

3.3 A Structural Change: Optical Modules Are Starting to Split Internally

Pre-market passed on optical modules as a whole (reason: the FCC draft is not finalized and the uncertainty is not lifted). That conclusion is maintained today, but an internal split appeared that pre-market did not anticipate:

• The system makers keep getting sold: 中际旭创 (Innolight, 300308, ChiNext) main-force net outflow of RMB 634 million, 新易盛 (Eoptolink, 300502, ChiNext) net outflow of RMB 1.15 billion.
• Upstream optical components are being bought: 光迅科技 (Accelink, 002281, SSE/SZSE main board) +2.30%, net inflow RMB 357 million; 天孚通信 (T&S Communications, 300394, ChiNext) +3.89%, net inflow RMB 1.197 billion; 光库科技 (Advanced Fiber Resources, 300620, ChiNext) +5.63%, net inflow RMB 571 million; 仕佳光子 (Shijia Photons, 688313, STAR Market) +3.61%, net inflow RMB 405 million.

This is money pricing "exposure to exports to the US" more finely — system makers with direct export exposure are sold, while upstream components with more indirect exposure are bought. Pre-market handled optical modules as a single block, which is not granular enough. Tomorrow morning "optical-module systems" and "optical components/upstream" should be split into two branches.

---

4. Limit-Up Ladder / Top of the Ladder

4.1 Full View of the Consecutive-Limit-Up Ladder

4.2 Promotion Rates and Promotion Quality

• The ladder is complete but the break is obvious: between 10 boards → 5 boards → 4 boards there are no 6/7/8/9-board names — the middle is empty. This shows the top of the ladder is "old carry-over names holding on by force" rather than the product of a new round of thematic promotions.
• 3 boards and 2 boards have 9 names each, a solid base, and this is the healthiest segment of the ladder.
• First boards at 57 names are 72% of the total, a high proportion, indicating the day was mainly new themes being tested rather than relay continuation.

4.3 Anchors (High-Turnover Advancers with Clear Identity)

The AI-hardware chain has anchors (长电科技 (600584), 胜宏科技 (300476) and 天孚通信 (300394) all with expanded volume and net inflow), precious metals have no anchor (紫金矿业 (601899) with expanded volume and net outflow), and memory's anchor is being net sold (长鑫科技 (688825) −RMB 3.003 billion). Those three sentences are essentially the entire branch-strength ranking for today.

4.4 Broken Boards and Failed Limits

• 20 failed limit-ups, a failed-limit-up rate of 20.2% (yesterday 29.5%), an overall improvement.
• The three most violent failed limits (worth noting individually):
  • 盛达资源 (Shengda Resources, 000603) (SSE/SZSE main board, precious metals): failed the limit 18 times, turnover rate 18.15%, closed +7.38%. The ebb-signal name called out pre-market — the pattern matched exactly.
  • 中钨高新 (China Tungsten & Hightech, 000657) (SSE/SZSE main board, minor metals): failed the limit 4 times, traded RMB 10.522 billion, closed +8.83%.
  • 利通电子 (Litong Electronic, 603629) (SSE/SZSE main board, consumer electronics): failed the limit 4 times, traded RMB 9.274 billion, closed +8.76%.
• One more extreme pattern: 百合花 (Lily Group, 603823, SSE/SZSE main board) still closed at the limit after failing it 46 times, on RMB 1.134 billion of trading. Opening off the limit 46 times and finally sealing is textbook high-divergence with strong absorption; volatility tomorrow will be large.
• Only 1 limit-down, so the money-losing effect is very low.

---

5. Money Flows

5.1 Northbound Capital: Disclosure Basis Missing, Not Provided in This Report

Since August 2024 the exchanges have discontinued real-time/post-close disclosure of single-day northbound net buying. Measurement of the interfaces this time confirms: both Shanghai Connect and Shenzhen Connect returned "net buy turnover" of 0.00 with a trading-status field of 3. This report does not pass off any substitute data as northbound net inflow.

Southbound is available (for reference only, not A-share money): Stock Connect (Shanghai) net buying +RMB 3.697 billion, Stock Connect (Shenzhen) −RMB 4.327 billion, for combined net selling of about RMB 630 million; over the same period the Hang Seng Index was −1.58%.

5.2 Dragon-Tiger List: Not Yet Published at Copy Deadline, Not Included in This Report

The East Money Dragon-Tiger List is published around 18:00 after the close, while this report closed its copy at 16:00; none of the four interfaces returned data for the day. This report therefore lists no Dragon-Tiger seats and infers no hot-money preference from them.

5.3 Main-Force Net Inflow TOP 15 (East Money basis: large + extra-large orders, Tencent snapshot)

Reading: 9 of the 15 belong to the AI-hardware chain of "semiconductor packaging-and-test / PCB / components / optical components" (长电科技, 天孚通信, 景旺电子, 方正科技, 胜宏科技, 光库科技, 仕佳光子, 江丰电子, 工业富联). The direction of the money was extremely concentrated today, and that direction is the one the pre-market brief wrote up as "do not chase".

5.4 Main-Force Net Outflow TOP 10

The single most important item: 长鑫科技 (ChangXin Memory, 688825) topped the main-force net outflow ranking for the second consecutive trading day (yesterday −RMB 2.542 billion, today −RMB 3.003 billion, −RMB 5.545 billion over the two days), with the share price −4.31% today. The domestic memory maker in the memory chain is being systematically reduced, and this is the hardest evidence that the memory branch must not be touched tomorrow.

5.5 Industry Net Inflows (Tonghuashun industry basis, must not be added to single-stock main-force net inflows)

• Top net inflows: electronic chemicals +RMB 1.811 billion, coal mining and processing +RMB 1.433 billion, precious metals +RMB 348 million, planting and forestry +RMB 105 million.
• Top net outflows: power −RMB 5.85 billion, brokers −RMB 5.449 billion, culture and media −RMB 3.747 billion, components −RMB 3.085 billion, minor metals −RMB 2.612 billion, whole-vehicle autos −RMB 2.436 billion, semiconductors −RMB 2.221 billion.

A contradiction that must be pointed out: components (+2.97%) and semiconductors (+2.14%) were among the top gainers, yet their industry net inflows were −RMB 3.085 billion and −RMB 2.221 billion. This shows severe internal divergence within these two industries — a handful of leaders (长电科技, 景旺电子, 方正科技) took huge net inflows while large numbers of small and mid-caps within the industry were being sold. "The industry went up" does not equal "the industry was bought", and tomorrow, picking names in these two industries must come down to single-stock money flow rather than substituting the industry reading.

---

6. Outlook for Tomorrow

6.1 Persistence Assessment for Tomorrow's Storylines

6.2 Key Names to Watch Tomorrow

The following are observation names and verification points; they do not constitute buy recommendations. All are tagged with board and price limit.

6.3 Avoid Tomorrow

1. The entire memory chain — 兆易创新 (GigaDevice, 603986), 佰维存储 (Biwin Storage, 688525), 江波龙 (Longsys, 301308), 德明利 (Demingli, 001309), 普冉股份 (PUYA Semiconductor, 688766), 北京君正 (Ingenic Semiconductor, 300223), 长鑫科技 (ChangXin Memory, 688825). Reason: 长鑫科技 had two-day net outflow of RMB 5.545 billion and fell 4.31%, and 兆易创新 flipped from +RMB 1.774 billion to −RMB 989 million in a single day, so the branch's "judged finished" conditions are fully satisfied. Add in SanDisk's FY27Q1 guidance coming in below consensus and TrendForce having already cut its Q3 contract-price increase forecast, and fundamentals and money flow point down together.
2. High-level consecutive-limit-up names — 爱丽家居 (Ailly Home, 603221, 10 boards, the company has already guided to an expected interim net loss of RMB 40.50 million–34.50 million), 豪尔赛 (Hao'ersai Technology, 002963, 5 boards), 风范股份 (Fengfan Power Equipment, 601700, 4 boards, failed the limit 5 times today), 云南锗业 (Yunnan Germanium, 002428, 3 boards, P/E(TTM) 3,118x). Reason: the ladder is broken between 5 boards and 10 boards, and with no intermediate relay at the top of the ladder, a broken board would produce a concentrated stampede.
3. High-level / high-divergence precious-metals names — 盛达资源 (Shengda Resources, 000603, failed the limit 18 times today, turnover rate 18.15%, 20-day +42.86%), 赤峰黄金 (Chifeng Gold, 600988, +1.13% today but with main-force net outflow of RMB 192 million, 20-day +31.74%). Reason: the branch has entered divergence, and these two are the highest-positioned with the worst patterns today.
4. Optical-module systems — 中际旭创 (Innolight, 300308), 新易盛 (Eoptolink, 300502). Reason: the FCC draft is still not finalized, so the uncertainty is not lifted; and there has been sustained large net outflow over two days (combined −RMB 1.784 billion today). Do not treat "it has stopped falling" as good news.
5. Innovative drugs — 百济神州 (BeiGene, 688235), 恒瑞医药 (Hengrui Pharmaceuticals, 600276). Reason: the good news of the earnings-delivery window was fully priced on 8/5, today was the first pullback day after delivery, and the biological products industry ranked eighth from the bottom.

6.4 Input Notes for Tomorrow Morning's Pre-Market List

1. [Highest priority] The pre-market brief must add a "money-driven candidate pool". Today's three strongest lines (packaging-and-test/PCB, coal, electronic chemicals) all had no news on the day, and a purely news-driven stock-selection entry point is structurally blind to them. Concrete method: take the previous session's stocks in the top 20 by main-force net inflow with a negative 20-day change and list them as a separate tier. Back-testing this sieve on today, 长电科技 (JCET, 600584, 20-day −26.71%) and 方正科技 (Founder Technology, 600601, 20-day −15.71%) would both be caught.
2. Split the Pass-list wording into two categories: "not recommended for participation (valuation/fundamentals do not support it)" and "expected to weaken (with explicit evidence)". Today's three Pass names hit the limit and not one of their Pass reasons was overturned, but lumping them under the same heading gets them read as a price forecast.
3. Downweight overnight offshore markets. Two days running we have seen "overnight US tech selloff → the corresponding A-share sector rises anyway". Recommendation: downgrade overnight offshore markets from a "sector-direction criterion" to an "opening-volatility note".
4. Split "optical modules" into two branches: "systems" and "upstream optical components". Today the two had opposite money directions (systems net outflow RMB 1.784 billion, while upstream 天孚通信/光库科技/仕佳光子/光迅科技 all saw net inflow), and handling them together misjudges both sides at once.
5. Three things to check first thing tomorrow morning: ① whether 长电科技 (JCET, 600584) issued any new announcement after the 8/6 close (today's limit-up had no same-day catalyst, so a post-close announcement would change its character); ② the seat composition for 长电科技, 云南锗业 and 盛达资源 in the Dragon-Tiger List (published after 18:00 today) — institutions vs hot money, which directly determines tomorrow's playbook; ③ whether coal shows any same-day price or policy catalyst (none found today).
6. One methodological reminder (recorded for the third time this week): whenever citing "the reason a stock hit the limit today", first check the announcement date of the cited reason. Today's two attributions for 长电科技 (earnings pre-announcement 7/15, RMB 7.8 billion project 6/25) are both old news, exactly the same pattern as 工业富联 on 8/5 and 8/6. This is no longer sporadic; it should be hardened into a mandatory verification step in the pre-market process.

---

⚠️ Risk disclaimer: this recap is a post-close information review and observation only and does not constitute investment advice. Data may differ in timeliness or basis; please refer to exchange/company disclosures as authoritative. It must not be used directly as a basis for trading.