A-Share Market Recap | 2026-08-21 Friday

0. One-Line Recap

Sentiment temperature: divergent warming (index barely up, limit-up count contracting from 79 to 55, but the failed-limit-up rate falling from 36.8% to 24.7% — "fever breaking, not tide receding"). Today's strongest main line was nonferrous metals across the board — silver sector +8.68%, lithium +5.91%, gold +5.09%, tungsten +4.90%, copper +3.21%; the nonferrous metals industry took +RMB 8.829 billion of net main-force inflow, first in the whole market.

Pre-market list scorecard in brief: the branch call was right, the single-stock criterion was wrong. The pre-market list ranked "precious metals · silver" as the branch with the strongest evidence, and today all 3 constituents of the silver sector (湖南白银 Hunan Silver 002716, 盛达资源 Shengda Resources 000603, 兴业银锡 Xingye Silver & Tin 000426) rose, with the sector +8.68%, the top industry gain in the entire market — the branch direction played out exactly. But the core criterion used pre-market to rank within that branch, "self-mined ore vs. smelting," was directly falsified by the market that day: 湖南白银 (Hunan Silver, 002716), which had been Passed and judged "not what its name suggests," today sealed the limit at 10:41, +10.03%, main-force net inflow RMB 1.056 billion (4th in the whole market), while 盛达资源 (Shengda Resources, 000603), listed as a priority deep-dive, only sealed at 13:57, broke the limit 3 times, and drew just +RMB 383 million of main-force inflow. Both hit the limit, but seal quality and capital intensity were exactly inverted.

The most complete pre-market misjudgment was on the branch with the strongest catalyst: 沪八条 ("Shanghai Eight Measures") took effect today, yet the real-estate development sector fell −0.89%, 18 up / 62 down, 0 limit-ups, main-force net outflow RMB 245 million, and 城投控股 (Shanghai Chengtou Holding, 600649) −6.73%. The call-auction gate set pre-market ("flat or gap-down open → front-running money distributing, downgrade the whole branch on the spot") correctly triggered at 09:25 — the risk-control criterion held, the directional call was voided. At 12:22 today there was also an incremental negative that could not have been foreseen pre-market: the Guangzhou Intermediate People's Court accepted the bankruptcy liquidation case of Evergrande Real Estate Group Co., Ltd.

Tomorrow's tone (next Monday 8/24): nonferrous continues, pharma diverges, real estate avoided. But a coverage gap must be admitted first: today's single largest line by capital volume (optical modules / PCB / copper-clad laminate / copper foil) got not one word in the pre-market list.

---

1. Market Overview

Data basis: Eastmoney quote interface, read after the 2026-08-21 close (15:00), local read moment 15:30–15:45.

Combined turnover of the two exchanges: RMB 1.8923 trillion (sum across all 5,901 A-share listings, stock by stock).

Breadth and sentiment indicators:

Basis validation (executed item by item per Iron Rule 3): limit-up pool declared tc=55, actually retrieved 55 rows (pass); limit-down pool tc=13, retrieved 13 rows (pass); failed-limit-up pool tc=18, retrieved 18 rows (pass). Limit-up/limit-down status is judged strictly by each board's own bound — main board ±10%, ChiNext/STAR ±20%, BSE ±30%; every stock below is tagged with its board.

Sentiment call: warming, but "shrinking volume, improving quality" warming, not euphoria. Three parallel pieces of evidence: ① limit-up count contracted from 79 to 55 while the failed-limit-up rate simultaneously fell from 36.8% to 24.7% — fewer participants but a higher success rate, money shifting from "casting a wide net for limit-ups" to concentrating on a single line; ② the peak height fell from 4 boards to 3 boards, and the 3-board name 汉森制药 (Hansen Pharmaceutical, 002412, Shenzhen main board ±10%) turned over only RMB 230 million, so there is no high-board anchor; ③ 2,505 advancers vs. 2,862 decliners — indices up while more individual stocks fall, meaning the money effect is concentrated inside the nonferrous and compute-hardware lines while most stocks outside them are still bleeding.

---

2. Pre-Market List Scorecard

Executable-price basis: the truly executable price for an A-share pre-market list is that day's opening price, so "open→close %" is given alongside "close change %". Looking only at the closing gain systematically overstates the hit rate — today the 5 recommended names averaged +2.94% on the close, but only +2.17% if executed as written at the open.

① Recommended Slots (§9① The 5 Most Worth Watching Today)

Recommended-slot hit rate: 3/5 = 60% (delivered = closed up, beat the SSE Composite, and met the verification point set for it pre-market). On direction alone (closed up), it is 4/5 = 80%.

② "Watch Only" List (16 names)

Watch-only hits: 14/16 = 87.5%, but both misses were 20cm names, and those two alone pulled the group average up to +2.60%.

③ Pass List (19 names)

Pass-list hits: 15/19 = 78.9% (hit = did not rise, or rose less than the SSE Composite). There were also two names §9③ explicitly flagged as "do not chase": 安科生物 (Anke Biotechnology, 300009, ChiNext ±20%) −11.63% and 康泰生物 (Kangtai Biological, 300601, ChiNext ±20%) −9.95% — both called correctly and by a wide margin.

④ Portfolio-Level Results

Recommended-group median +4.96% vs. Pass-group median −0.78%, a spread of 5.74 percentage points; recommended group 4/5 advancers vs. Pass group 7/19. The stratification works.

⑤ The 17 Verification Points Set Pre-Market, Adjudicated One by One

⑥ One-Line Post-Mortem: What Was Right, What Was Wrong, What Changes Tomorrow Morning

The right category: exclusionary criteria. Every negative judgment based on "already pre-announced, no expectation gap," "loss already locked in," "concept mismatch," or "poor seal quality" held today — 志邦家居 (603801) open→close −9.15%, 金健米业 (600127) −12.67%, 安科生物 (300009) −11.63%, 康泰生物 (300601) −9.95%, 拓荆科技 (688072) +0.33%, 中华企业 (600675) −3.91%, 京投发展 (600683) −6.01%. On the question "this stock should not be bought," the pre-market methodology was quite accurate.

The wrong category: ranking criteria. Of the two "who is strongest within the branch" criteria set pre-market, one was falsified today and one was bypassed:

1. "Self-mined ore vs. smelting" is not a ranker on a single-day time scale. It measures "whose incremental gross margin gains more when silver rises further" — a quarterly-scale fundamental elasticity criterion; what money bought today was "first-disclosed headline growth + low price + small cap." 湖南白银 (002716) only first disclosed H1 +170.34% at 18:03 on 8/20 and sealed the limit at 10:41 today; 盛达资源 (000603) had already disclosed its interim report on 8/1 and only sealed at 13:57. The very fact written down pre-market — "the only one of the 7 that had not pre-announced, i.e. a first disclosure" — was used by me as a bearish reason, when it was in fact the strongest bullish reason of the day. This is the second occurrence of the same pattern as the remembered lesson "financial quality is not a short-term ranker" — last time it was 盛美上海, ex-non-recurring −15.31%, finishing +12.16% and first on the board that day.
2. The "strongest catalyst ≠ strongest branch" distinction was right, but the position-sizing implication was never spelled out. The pre-market correctly wrote "catalyst strength points to 沪八条, evidence strength points to precious metals," yet still put 城投控股 (600649) in recommended slot 2. Once it was characterized as a "pure policy-sentiment trade," it should never have occupied a recommended slot — it fell −6.73% today, the only negative contributor among the 5 recommended names, dragging the portfolio average down by roughly 1.35 percentage points on its own.

Three changes for tomorrow morning's pre-market list (directly executable):

• Change one: within-branch ranking must carry a separate "same-day trading dimension" column, parallel to fundamental elasticity, and the two must not be merged. Three concrete fields: ① is the name's key earnings release/filing a first disclosure that day or the prior evening (first disclosure > already pre-announced); ② absolute share price and free-float market cap (under RMB 10 / RMB 10 billion carries more elasticity); ③ is it the cheapest name in the sector that day. 湖南白银 hit all three today.
• Change two: any name characterized as a "sentiment trade" is barred from the recommended Top5 and goes only into the watch pool. Recommended slots are reserved for names with all three of "new overnight driver + results already delivered + verifiable business purity." Recomputing today under that rule: the recommended slots should have been 盛达资源, 兴业银锡 and 山东黄金, giving a group average of +7.00% instead of +2.94%.
• Change three: the §8 coverage self-check (item 17) must be moved forward to before the pre-market list is finalized, not left for the recap. Today it accurately caught four missed industries, but by then the market had already closed. Pre-market should instead: pull the top 10 industries by main-force net inflow before finalizing, and for every industry not on the branch list, write out explicitly "why we are not looking at it today."

---

3. Verification of Today's Main Lines

Did the Pre-Market Call the Strongest Main Line? — Half Right

Right: ranking "precious metals · silver" as the strongest evidence. Today the silver sector's +8.68% was the top industry gain market-wide, precious metals +5.99% second, gold +5.09% fourth. The driver chain given pre-market (U.S. Treasury doubles the size of a single long-bond buyback → credit concerns → silver stronger than gold) was continuously validated on the commodity side during the session: at 10:50 spot silver moved above USD 69/oz and the SHFE front-month silver contract rose over 4%; as of the 15:35 Beijing-time read, COMEX silver @SI.1 was 68.95 (+1.24%), COMEX gold @GC.1 4,618.70 (+1.03%), COMEX copper @HG.1 6.559 (+1.39%).

Wrong: ranking "沪八条" as the strongest catalyst and giving it a recommended slot. The pre-market spent an entire section arguing that "the policy tools and the listed companies' inventory value are not in the same market" and that "all three main names have already pre-announced H1 losses" — the argument was entirely correct, yet the conclusion was still "watch closely." Today's outcome shows: once a chain's fundamentals have been negated by your own analysis to that degree, the correct conclusion is Pass, not "trade it on the sentiment line."

There was also an incremental negative that could not have been foreseen pre-market: at 12:22 on 8/21 the Guangzhou Intermediate People's Court accepted, in accordance with law, the bankruptcy liquidation case of Evergrande Real Estate Group Co., Ltd. (the petitioner being the Huaxia sub-branch of Guangzhou Rural Commercial Bank, on the grounds of inability to repay debts due and assets insufficient to cover all debts). This landed at midday and further suppressed the property sector's scope for an afternoon repair.

Any Main Lines the Pre-Market Missed? — Two, and Both Large

First: the breadth of nonferrous metals. The pre-market wrote up only two sub-segments, "silver" and "gold," while today was a systematic re-rating of the entire nonferrous complex — lithium +5.91%, tungsten +4.90%, copper +3.21%, lead-zinc +3.74%, molybdenum +4.39%, nickel +3.15%, all up, and the nonferrous metals industry's main-force net inflow of +RMB 8.829 billion was first market-wide. The pre-market wrote the driver out completely as "U.S. Treasury credit concerns → precious metals," but never extrapolated that driver's spillover into the entire commodity pricing complex. 白银有色 (Baiyin Nonferrous, 601212, Shanghai main board ±10%) was completely absent from the pre-market list of 7 silver names because it is classified under "copper," and today it sealed the limit at 09:42, +9.93%, main force +RMB 707 million.

Second: AI compute hardware was today's true battlefield for capital, and the pre-market wrote not one word about it. Communications equipment +RMB 8.532 billion, components +RMB 4.490 billion, printed circuit boards +RMB 4.118 billion — RMB 17.1 billion of main-force net inflow across the three; the top two names market-wide by main-force net inflow (中际旭创 +RMB 3.595 billion, 新易盛 +RMB 2.529 billion) are both on this line, followed by 生益科技 (Shengyi Technology, 600183, Shanghai main board) +RMB 1.011 billion, 歌尔股份 (GoerTek, 002241, Shenzhen main board) +RMB 923 million, 沪电股份 (WUS Printed Circuit, 002463, Shenzhen main board) +RMB 785 million, 东山精密 (Dongshan Precision, 002384, Shenzhen main board) +RMB 703 million, 国瓷材料 (Sinocera, 300285, ChiNext) +RMB 623 million. Note: this line has the shape of "inflows far exceeding limit-up count" — institutions are buying, hot money is not chasing limits.

⚠️ One uncertainty that must be labeled honestly: this report was unable to confirm any new catalyst for this line on 8/21. The only related leads found were two pre-market media reports — 07:27 "the lithium iron phosphate cathode industry sees a full recovery, leading companies' earnings up more than 8x" and 07:28 "the copper foil industry saw a strong reversal in H1, some producers are already preparing another round of price increases" — plus an 08:43 filing from 国际复材 (Chongqing Polycomp International, 301526, ChiNext) that its low-dielectric LDK second-generation product had entered a leading CCL maker's supply chain. These three explain copper foil (诺德股份 600110 limit-up) and CCL; they do not explain the RMB 6.1 billion of main-force net inflow into 中际旭创 and 新易盛. Per Iron Rule 2, this is written as "driver not yet confirmable," with no fabricated attribution. "No catalyst found" is not the same as "no catalyst" — this line must have its news flow filled in before any participation is discussed on Monday.

---

4. Limit-Up Tiers / High Boards

Tier structure: 1 name on 3 boards, 11 names on 2 boards, 43 first boards. Overall a "short and wide" structure with no anchor.

Advancement rate: of the 79 limit-ups on 8/20, 28 rose today (35.4%) and 12 hit the limit again (15.2%). The 12 that advanced to a second board are: 中关村 (Zhongguancun, 000931), 双鹭药业 (Shuanglu Pharmaceutical, 002038), 汉森制药 (002412, advancing to 3 boards), 康希诺 (688185), 键凯科技 (688356), 近岸蛋白 (688137), 贝瑞基因 (Berry Genomics, 000710), 深中华 A (Shenzhen China Bicycle A, 000017), 哈森股份 (Hasen Shoes, 603958), 通鼎互联 (002491), 科森科技 (Kesen Technology, 603626), 宇环数控 (Yuhuan CNC, 002903). Seven of them belong to the pharma chain — while yesterday's main line was receding as a whole, its leaders instead completed their advancement, which is the standard shape of "divergence" rather than "receding tide."

Broken streaks: yesterday's 4-board leader 金健米业 (Jinjian Cereals, 600127, Shanghai main board ±10%) never touched the limit price all day (high 9.45, limit price 9.46), closing −4.65%, open→close −12.67%, turnover rate 42.41%. The space board snapping is the direct reason the day's peak height fell from 4 boards to 3.

Three names in the failed-limit-up pool worth recording: 华正新材 (Huazheng New Material, 603186, Shanghai main board ±10%, CCL) turned over RMB 4.632 billion yet closed only +5.10%; 志邦家居 (603801, Shanghai main board ±10%) spiked +9.96% then closed −3.87%; 康希诺 (688185) went back and forth between limit-up and break 15 times and still closed at +20.01%.

---

5. Capital Flows

① Cross-Border Flows

• Northbound: daily net buying has no longer been disclosed since 2024-08-19; today the interface (push2/api/qt/kamt) returned dayNetAmtIn = 0.0 with status = 3, which is the signature of "disclosure discontinued," not "net inflow of zero." This report gives no northbound figure.
• Southbound: the 8/21 data had not been published at this report's generation time (15:30); the latest available from the interface is 8/20's net selling of HKD 10.412 billion (a second consecutive day of net selling; 8/19 was −HKD 10.621 billion, 8/18 was +HKD 14.002 billion).

② Industry Main-Force Net Inflows / Outflows

In one line: today was a clean rotation out of "pharma" and into "nonferrous + compute hardware." Pharmaceuticals & biotech lost RMB 9.111 billion, nonferrous metals gained RMB 8.829 billion — almost one for one in magnitude.

③ Concept Sectors

Top inflows: lithium mining concept +RMB 3.854 billion (+3.30%), gold concept +RMB 6.784 billion (+2.63%), minor metals concept +RMB 6.531 billion (+1.97%), scarce resources +RMB 4.153 billion (+2.11%), F5G concept +RMB 4.884 billion (+1.85%), high-speed copper cable +RMB 1.601 billion (+1.99%). Top outflows: innovative drugs −RMB 5.835 billion (−3.96%), pharma & healthcare style −RMB 5.638 billion (−3.78%), pathogen control −RMB 3.432 billion (−3.76%), CRO −RMB 3.189 billion (−4.09%), weight-loss drugs −RMB 2.577 billion (−4.57%), monoclonal antibody concept −RMB 2.313 billion (−4.80%).

④ Top 10 Stocks by Main-Force Net Inflow

Top main-force outflows: 恒瑞医药 (Hengrui Pharmaceuticals, 600276, Shanghai main board) −RMB 699 million, 康龙化成 (Pharmaron, 300759, ChiNext) −RMB 697 million, 贵州茅台 (Kweichow Moutai, 600519, Shanghai main board) −RMB 594 million, 长电科技 (JCET, 600584, Shanghai main board) −RMB 488 million, 石药创新 (CSPC Innovation, 300765, ChiNext) −RMB 385 million, 药明康德 (WuXi AppTec, 603259, Shanghai main board) −RMB 351 million.

⑤ Block-Trade Leaderboard (Dragon-Tiger List)

The 8/21 dragon-tiger list was not yet available at this report's generation time (15:30) — the exchanges' and Eastmoney's daily boards are typically published after 18:00, and RPT_DAILYBILLBOARD_DETAILSNEW returns success: false for TRADE_DATE=2026-08-21; the same interface returns 78 rows normally for 8/20. This is a scheduling problem, not an interface failure.

This is a structural defect of the recap slot: a 15:30 generation moment can never retrieve the same day's dragon-tiger list. It is the same item as the remembered lesson "the dragon-tiger list is only published after 18:00." To keep seat-level analysis in the recap there are only two options: ① move the A-share recap slot to after 18:30; ② keep the 15:30 slot but split the dragon-tiger list out as a fixed back-fill item in the next morning's pre-market list. Option ② is recommended, because the timeliness of 15:30 is worth more for the "tomorrow outlook."

This report does not include 8/21 dragon-tiger seat data: the exchanges publish the daily board relatively late after the close, and this recap's generation moment precedes publication. The seat composition of today's likely board candidates (湖南白银 002716, 盛达资源 000603, 白银有色 601212, 通鼎互联 002491, 旭光电子 600353, and others) will be back-filled and cross-checked in the next pre-market list.

---

6. Tomorrow's Outlook

⚠️ Time basis: today is Friday (8/21). Below, "tomorrow" always means the next trading day = 2026-08-24 (Monday). The sole exception is the interim report of 上实发展 (600748), whose scheduled disclosure date is 2026-08-22 (this Saturday), falling on the weekend.

① Main-Line Continuation Calls

② Focus for Next Monday (8/24)

③ Avoid on Monday (8/24)

1. The entire Shanghai property chain — 城投控股 (600649), 中华企业 (600675), 光明地产 (600708), 陆家嘴 (600663), 浦东金桥 (600639), 外高桥 (600648), 上海临港 (Shanghai Lingang, 600848). On the very day the policy took effect the sector fell −0.89% with 0 limit-ups and net main-force outflow — falsification in its strongest form. Compounded by the Guangzhou Intermediate Court's 8/21 acceptance of Evergrande Real Estate Group's bankruptcy liquidation.
2. The tail of yesterday's main line, not its head — the distinction matters: the pharma chain's median has already receded today, and 24 of yesterday's 37 pharma limit-ups fell, among them 安科生物 (300009, ChiNext) −11.63%, 莱美药业 (Lummy Pharmaceutical, 300006, ChiNext) −11.28%, 博腾股份 (Porton Pharma, 300363, ChiNext) −10.58%, 优宁维 (Univ Biotechnology, 301166, ChiNext) −10.04%, 康泰生物 (300601, ChiNext) −9.95%. This layer is explicitly to be avoided.
3. High-position names with poor seal quality — 康希诺 (688185, STAR Market ±20%, broke the limit 14 times + seal order to zero), 通鼎互联 (002491, Shenzhen main board ±10%, seal ratio 0.021 + main force −RMB 258 million + turnover rate 30.46%), 近岸蛋白 (688137, STAR Market ±20%, seal ratio 0.004). All three are on 2 boards, and none of the three has a seal order that can support a relay the next day.
4. Space boards after a broken streak — 金健米业 (600127, Shanghai main board ±10%) went open→close −12.67% today, the 4-board tier is broken, do not catch the rebound.
5. Names with compliance doubts — 智飞生物 (300122, ChiNext ±20%): its interim report was scheduled for 8/18 and is now overdue by 4 trading days, still undisclosed with no delay notice issued (the latest Eastmoney filing on record is 8/10). Disclosure compliance at an RMB 34.6 billion market-cap name outranks any thematic narrative.
6. Isolated one-off limit-ups — 旭光电子 (Xuguang Electronics, 600353, Shanghai main board ±10%, +10.00%, turnover RMB 3.783 billion but broke the limit 4 times, seal ratio 0.002), 诺德股份 (Nuode Investment, 600110, Shanghai main board ±10%, first sealed only at 13:27, seal ratio 0.003): very large turnover with an extremely thin seal order is the textbook "high-churn relay" shape, not locked-up capital.

④ Input Notes for Monday's Pre-Market List

1. Today's (8/21) dragon-tiger seat data must be back-filled. This report lacks it because the generation moment (15:30) precedes the daily board's publication (after 18:00). Focus on four names: 湖南白银 (002716), 盛达资源 (000603), 白银有色 (601212), 通鼎互联 (002491). And per the remembered lesson "divide the seal order by turnover, break institutional net buying down by seat," institutional net buying must be broken down to the seat level and not substituted with the whole-board net figure.
2. Fill in the news flow for the AI compute hardware line. 中际旭创 (300308) +RMB 3.595 billion and 新易盛 (300502) +RMB 2.529 billion were today's top two main-force net inflows, yet this report could not confirm the day's catalyst. Monday's pre-market must first answer "why did this line rise on 8/21," and only then decide whether to give it a branch slot.
3. Precious-metals coverage must expand from 7 names to 13 (all constituents of the precious-metals sector BK0732), and "copper (BK1615, 17 names), lithium (BK1621, 8 names), tungsten (BK1625, 4 names), lead-zinc (BK1614,11 names)" must be brought in as parallel sub-lines. Today's lesson: the driver is "U.S. Treasury credit concerns," and it acts on the entire commodity pricing complex, not just on gold and silver.
4. Add the three "same-day trading dimension" fields to within-branch ranking (first disclosure / absolute price and free-float cap / relative cheapness within the sector), scored in parallel with fundamental elasticity, never merged. Concrete quantitative suggestion: fundamental elasticity and the trading dimension each carry 50%, and when the two conflict (as with 盛达资源 vs. 湖南白银 today), the conflict must be stated explicitly in the body with a parallel conclusion for both names, rather than using one criterion to Pass the other.
5. Any name characterized as a "sentiment trade" is barred from the recommended Top5. Recomputing today under that rule lifts the recommended group's average from +2.94% to +7.00%.
6. Move the §8 coverage self-check forward to before the pre-market list is finalized: pull the top 10 industries by main-force net inflow, and for every industry not on the branch list, write out item by item "why we are not looking at it today." Today this criterion accurately caught 4 missed industries after the close, but by then they could no longer be traded.
7. Next week's calendar (each item dated; do not write "tomorrow"): 8/22 (Sat) interim reports of 上实发展 (600748) and 紫金矿业 (601899); 8/27 (Wed) the review meeting for 中金公司 (601995) absorbing 东兴证券 (601198) and 信达证券 (601059) by share swap, plus 中华企业 (600675) interim report the same day; 8/28 (Fri) interim reports of 光明地产 (600708) and 志邦家居 (603801); 8/29 (Sat) interim reports of 兴业银锡 (000426), 山东黄金 (600547) and 沃森生物 (300142); 8/31 (Mon) interim reports of 城投控股 (600649) and 赤峰黄金 (600988).

---

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

Data-Retrieval and Run Log (not sent to clients)

1. Retrieval Failures

• push2.eastmoney.com quote/snapshot interfaces recovered today (first availability after three consecutive days of 502s), but /api/qt/clist/get still 502s continuously, so industry/concept sectors and the full A-share list went through push2delay.eastmoney.com throughout. Availability differs across paths under the same host, so the next failure diagnosis must go down to path granularity and not be judged host-wide.
• getTopicDTPool (limit-down pool) failed silently: with sort=fbt:asc it returned tc=13 but an empty pool array. Switching to sort=fund:asc retrieved all 13 rows in one go. This is the second occurrence of the "tc non-zero but pool empty is a silent failure" item inside the remembered lesson "limit-up/limit-down status must use the official pool" — without the len(pool)==tc validation, this report would have said "0 limit-downs today." The limit-up and failed-limit-up pools work fine with fbt:asc; the limit-down pool's sort field is not interchangeable with the limit-up pool's.
• The one-sided pagination bias in the full A-share breadth data reappeared: fs=m:0+t:6,... declares total=5901, but paging descending (po=1) to the end yielded only 3,900 rows, with everything missing coming from the loser tail — computed on that data the result would have been "2505 up / 861 down / 0 names down more than 5%," while the market plainly had 13 limit-downs. After an added ascending pass (po=0) and merging back to 5,901 rows, the truth is "2505 up / 2862 down / 217 names down more than 5%." This bias flips the sentiment call outright (the up-down ratio goes from 2.9:1 to 0.88:1). It is the exact same shape as the remembered lesson "the one-sided bias of Eastmoney clist pagination," and this time it was caught only because the total == len(got) validation was done up front; that validation must stay a fixed step.
• The 8/21 dragon-tiger list is unavailable (the daily board publishes after 18:00; at the 15:30 generation time it returned success: false). The same interface returns 78 rows normally for 8/20. This is not a failure, it is a scheduling conflict — see the handling recommendation in §5⑤.
• api/qt/slist/get (a stock's sector membership) returned data: null, so the concept classification of 高新发展 (Hi-Tech Development, 000628) could not be obtained, and the body therefore gives no thematic attribution for that limit-up.

2. Freshness Cross-Check (symbol by symbol)

• CNBC commodities: last_time for @SI.1 / @GC.1 / @HG.1 was all 2026-08-21T03:35 -0400 = 15:35 Beijing time; XAG= / XAU= was 03:45 -0400 = 15:45 Beijing time. All are post-A-share-close reads, and the read moment is labeled at every occurrence in the body (following "cross-market data must carry a read timestamp"). No stale data this time.
• A-share quotes: limit-up pool date=20260821; quote snapshots read between 15:30 and 15:45, all post-close final values.
• The latest available date on the southbound interface is 2026-08-20 (NET_DEAL_AMT = −10412.07, in millions of HKD = −HKD 10.412 billion, consistent with yesterday's pre-market list record; cross-basis check passed). 8/21 not yet published, as declared in the body.
• Northbound: kamt returned dayNetAmtIn = 0.0, status = 3. On the established fact (disclosure discontinued since 2024-08-19) this is judged "not disclosed" rather than "zero inflow," and the body does not use that number. Without that judgment, this report would have produced a fabricated conclusion like "northbound net inflow of zero, a neutral stance."

3. Full-Coverage Basis Validation

4. Methodological Takeaways from This Report

1. "Financial quality is not a short-term ranker" recurred for the second time within the same month. Last time it was 盛美上海, ex-non-recurring −15.31%, finishing +12.16% and first on the board that day; this time it is 湖南白银 (002716) — 97% smelting, operating cash flow −RMB 1.377 billion — sealing the limit 3 hours 16 minutes earlier than 100%-self-mined 盛达资源 (000603). The common thread: I used a correct criterion on the wrong time scale (quarterly elasticity) to solve a same-day problem. The fix is not to scrap the fundamental criterion but to force both criteria to be output in parallel and forbid a one-sided ruling when they conflict.
2. A fact I wrote down myself pre-market can be pointed in the wrong direction. Pre-market §1 item 8 explicitly recorded that "湖南白银 is the only one of the 7 precious-metals names that had not pre-announced, i.e. a first disclosure," and I wrote that fact into the bearish paragraph "brightest headline, most fragile structure." The same fact is a risk inside a fundamental framework and the strongest bullish signal inside a trading framework. The checklist needs one more question: "Is the fact I am using to be bearish actually bullish on the trading dimension?"
3. A correct argument whose conclusion fails to follow is a more insidious failure than a wrong argument. The 沪八条 section argued "inventory-value mismatch," "all three names have already pre-announced losses," and "the subsidy pool is only RMB 200 million" completely and thoroughly, yet still gave 城投控股 (600649) recommended slot 2 on the grounds of "highest trading recognizability." Once a chain's fundamentals have been negated by your own analysis to that degree, the correct conclusion is Pass. New rule: any name characterized in the body as a "pure sentiment trade" is barred from the recommended Top5.
4. The coverage self-check must run pre-market, not be left to the recap. The pre-market §8-17 criterion was well designed and triggered accurately today (4 of the industries with the most limit-ups were absent from the branch list), but its execution moment was after the close, which means it had only diagnostic value and no trading value.
5. The spillover radius of a driver must be explicitly extrapolated one step. The pre-market wrote the chain "U.S. Treasury doubles long-bond buybacks → U.S. Treasury credit concerns → precious metals" very accurately, but never asked "where else will this driver land." The answer is the entire commodity pricing complex — today lithium +5.91%, tungsten +4.90%, copper +3.21%, lead-zinc +3.74%, molybdenum +4.39%, all up, with the nonferrous metals industry's main force at +RMB 8.829 billion, first market-wide.

5. Not Done / Not Obtained / Doubtful in This Report

• 8/21 dragon-tiger seat data is entirely missing (scheduling reason, not a failure). §5 of this report has no seat analysis, as declared to clients in the body.
• The same-day catalyst for the AI compute hardware line (中际旭创 300308 +RMB 3.595 billion, 新易盛 300502 +RMB 2.529 billion) is unconfirmed. The three related leads found (LFP recovery, copper foil price increases, 国际复材's low-dielectric product entering a leading CCL maker's supply chain) explain copper foil and CCL only, not optical modules. Per Iron Rule 2 the body says "driver not yet confirmable" and fabricates no attribution. But beware the remembered lesson "no news found ≠ no news" — most likely this is a search miss, not an absence of news.
• Several of today's limit-ups have no thematic attribution: 旭光电子 (600353, RMB 3.783 billion turnover), 士兰微 (600460, main force +RMB 1.148 billion), 中瓷电子 (Zhongci Electronic, 003031), 诺德股份 (600110), 飞龙股份 (Feilong Auto Components, 002536), 江钨装备 (Jiangwu Equipment, 600397). These are first boards with relatively large turnover, and the missing attribution means that on Monday it will be impossible to tell whether they are part of a branch or single-stock moves.
• The cause of 康希诺 (688185) "breaking the limit 14 times and ending with a zero seal order yet still closing at +20.01%" was not established. The official failed-limit-up pool records 15 times for it while the limit-up pool records 14 — two official bases in conflict (the same class of phenomenon as 城投控股 on 8/20 closing at the limit price while absent from the limit-up pool). The body characterizes it as "extremely unstable shape" and draws no conclusion about the discrepancy in counts.
• 上实发展 (600748)'s interim report is scheduled for 2026-08-22 (Saturday). The Eastmoney schedule table shows IS_PUBLISH=0, RESIDUAL_DAYS=1, confirming it is not yet disclosed. Saturday disclosure is uncommon on the Shanghai exchange, so a rollover to 8/24 (Monday) is possible; the next issue must confirm the actual disclosure date before discussing falsification.
• 智飞生物 (300122)'s interim report is still undisclosed with no delay notice, its scheduled date of 8/18 now overdue by 4 trading days. The latest entry in the Eastmoney filings library (np-anotice-stock) is a clinical-trial acceptance notice from 8/10.
• This report did not run an independent risk-auditor quality check. The run log of the previous pre-market list already flagged the same item, making this the second consecutive issue without it. This report's high-risk assertions (especially the three improvement proposals in §2⑥ and the open question "AI compute hardware catalyst unconfirmed") have not been falsified by any independent party.
• Today's combined turnover of RMB 1.8923 trillion is on the stock-by-stock summation basis for all A-shares, differing by RMB 13.1 billion from the exchange basis of "SSE RMB 883.4 billion + SZSE RMB 995.8 billion = RMB 1,879.2 billion"; the gap should be the BSE plus some statistical-basis differences, and was not reconciled item by item.