U.S. Market Recap | 2026-08-17 (ET) Monday

Coverage window: 2026-08-17 09:30–16:00 ET regular session, including the 16:00–17:00 ET after-hours session. Quote basis: closing prices, opening prices, intraday highs/lows and volumes are all taken from the stockanalysis.com quote API (u timestamp = Aug 17, 2026, 4:00 PM EDT close basis); indices are taken from CBOE delayed quotes (last_trade_time 16:14–16:15 ET) and the official Nasdaq API; Treasury yields are taken from the U.S. Treasury daily yield curve for the day (15:30 ET settlement basis); after-hours prices are as of around 17:00 ET, with timestamps separately annotated at each point.

Reconciliation baseline: the pre-market list is reports/us/2026-08-17.md (08:02–08:40 ET readings). This recap gives both a "vs. Friday close" and a "vs. pre-market quote" column — using only the former systematically overstates the hit rate.

---

0. Recap in One Line

1. Today was risk-off, but not a flight to safety — it was "reflation + term premium": S&P −0.52% to 7,745.06, Dow −0.51%, Nasdaq Composite −0.32%, Russell 2000 −0.35%; the S&P opened at the day's high and closed at the day's low (open 7,790.68 / high 7,790.68 / low 7,744.88), a one-way decline all day. But bonds fell at the same time (TLT −0.84%, 30Y 5.31%), while gold +1.0%, crude +2.65%, the dollar flat — this is not the traditional risk-off in which haven money flows into Treasuries; it is supply-shock pricing.
2. The pre-market list ranked "war" No. 2 and called it "underpriced by the market" — and it was delivered the same day: Trump, on the expiry of the Iran ceasefire memorandum, said he "cannot see the war ending soon" and threatened force if Oman obstructs the Strait of Hormuz. Brent +2.65% to $90.87, WTI +2.5% to $84.50; the 30Y Treasury at 5.31%, the highest since June 2007. Yahoo Finance's own market headline for the day attributed the whole-day decline to this.
3. The pre-market list got the strongest theme right, but got the loudest one wrong: biotech was, as expected, the strongest of the day (IBB +1.85%, XBI +1.35%, ARGX +16.04%), and war/energy, retail weakness, social-media legal liability and the RDDT index event were all four directionally correct; but the storage chain was labeled "avoid / watch only" across the board and in fact rallied across the board (SNDK +8.88%, AMAT +5.55%, WDC +5.35%, MU +4.13%, STX +2.19%). Of the 6 names called wrong on the day, 4 are concentrated in this one line.
4. The real split was not "storage vs. non-storage" — it was "those who collect AI capex vs. those who spend AI capex": today's gainers were the supply side (MU/SNDK/WDC/AMAT/SMH +1.06%), and the decliners were the spending side (MSFT −3.04%, META −3.54%, GOOGL −0.55%, AMZN −0.51%). The pre-market list avoided storage on the grounds that "the narrative anchor is soft," but that narrative was simply not what the market traded that day.
5. VIX 15.19 (+6.60%) — up, but still at a low level. The contradiction the pre-market list flagged against itself (war + a 30Y breakout + confidence at 51, yet VIX only 14.9) was corrected by just 0.94 points today; the contradiction still stands.
6. Tone into the next session: the HD earnings report (pre-market 8/18) + META opening statements + RDDT's S&P 500 inclusion at the open all land on the same day; on the macro side, watch whether the long end can hold 5.31%. There are no major earnings after the close; all after-hours movers before 17:00 ET are micro-caps, and the S&P/Nasdaq ETFs were essentially flat after hours (SPY −0.02%, QQQ −0.06%).

---

1. Market Overview

Indices

Pattern note: the S&P opened at the day's high and closed at the day's low, with no meaningful bounce in between. This kind of "one-way down candle" usually means the selling pressure came from a persistent macro input (today: oil prices and long-end rates) rather than a news shock at a single point in time.

Market Breadth

Breadth was negative, and worse than the index. The equal-weight index fell nearly twice as much as the cap-weighted one, meaning the median stock was weaker than the index; gains were concentrated in a small number of large-cap semiconductor and energy names. Note: mega-caps such as MSFT/META were also falling today, yet cap-weighting still outperformed — which means the gains were concentrated in the "large but not largest" tier (the MU, AMAT, XOM type).

Rates, Volatility and Cross-Asset

Sentiment read: this is "term premium + supply shock" pricing, not a growth scare. Three pieces of evidence: ① stocks and bonds fell together — on a genuine haven day Treasuries should rise, yet today TLT was −0.84% and pressing its 52-week low; ② bear steepening of the curve (2s10s went from 51bp to 53bp, with the long end rising more than the short end) — a growth scare produces bull steepening or bull flattening; only inflation/term premium produces bear steepening; ③ the dollar was flat while gold rose — in a global flight to safety the dollar usually strengthens. Oil +2.65% is the upstream variable common to all three.

One counter-piece of evidence that must be acknowledged: under this combination the VIX rose only from 14.25 to 15.19, an absolute level still in the low range. The contradiction the pre-market list flagged against itself — "the market does not accept this risk pricing" — was not resolved today, only mildly corrected.

Today's Macro Data (both beat expectations, stocks still fell)

This is the key link for understanding today: both data points came in better than expected, yet stocks fell and long-end rates rose. The market is currently in a reaction function of "good data = higher rates = lower valuations." NAHB rebounded modestly, but it has now been below 40 for 16 consecutive months, with at least 30% of builders cutting prices for 16 straight months — a bearish backdrop for tomorrow's HD and Wednesday's LOW earnings.

---

2. Pre-Market List Reconciliation

Basis note: the "pre-market" column is the 08:02–08:40 ET reading recorded in the pre-market list; the "vs. pre-mkt" column = today's close ÷ pre-market quote − 1, representing the result of "actually placing the order pre-market per the pre-market list." The two columns often give opposite conclusions, which is precisely why both must be shown.

Long-Side Calls

Short / Avoid Calls

Hit Rate

Post-Mortem: the errors were not random — they all sit on one line

Of the six names called wrong, four are storage/semis (SNDK, AMAT, WDC, STX) and two are crypto miners (HIVE, BTDR). Not one of the eight correct calls sits on either of these two lines. That means the day's errors were not a matter of scattered luck but a systematic misread of one theme. Three specific self-criticisms:

1. [Most serious] Equating "the narrative anchor is soft" directly with "the price will fall." The pre-market skepticism of that Anthropic revenue forecast (an unlisted company, anonymous sources, not an SEC filing, not an order) was entirely correct on the facts, and nothing today refutes it. But the market was not trading that narrative today — it was trading the transfer of AI capex from buyers to sellers (see Section 3). "This piece of news is not solid enough" and "this sector will fall today" are two independent propositions, and the pre-market list merged them.

2. The "good news that fails to lift the price = sector top" thermometer had its reading overturned. The pre-market list took AMAT's Friday beat-but-fell −5.12% as the core evidence that "semis have entered a good-news-fails-to-lift phase," and downgraded the entire storage chain on that basis. Today AMAT was +5.55%, not only recovering the entire decline but closing above the pre-decline close of $534.54. A single day of "good news failing to lift" is noise; upgrading it into a sector characteristic requires more samples — using one trading day's price reaction to infer a sector's phase is insufficient evidence.

3. The "HDD is not AI storage" tiering did not play out on the day, but it was not falsified either. The pre-market list explicitly pre-specified the test: "if MU/SNDK rise while WDC/STX weaken, that is the correct differentiation." In fact all four rose together, and WDC (+5.35%) even outgained MU (+4.13%); the market traded the four as a single sector that day. That said, STX (+2.19%) was indeed the weakest of the four, so the tiering partly holds on relative strength. The difference in business type is an accounting fact and does not change because of one day of beta — but on that day it was not a tradable ranking criterion, and the pre-market list overrated it.

But one thing was done right, and it matters more than the hit rate: the quality of the verification points exceeded the quality of the conclusions

The pre-market list set one falsifiable verification point for each of 5 names. Four of them produced a clear reading on the day, and 2 directly overturned the headline conclusion attached to them:

Takeaway: writing a falsifiable verification point first and reasoning back to a conclusion is more reliable than putting the conclusion first. In today's OABI and AMAT cases, the pre-market list's own verification points negated its own conclusions — which shows the verification points are effective independent information, not decoration on the conclusion. MU is the most valuable example: the price condition passed while volume did not, and looking only at "held $1,000" would have produced a stronger signal than reality warrants.

---

3. Theme Verification

① Biotech: right, but it must be qualified — this was a single-stock event, not a sector rotation

XLV (the full healthcare sector) was −0.19% today — it fell. Only the biotech sub-sector rose (IBB +1.85% / XBI +1.35%), and a substantial part of IBB's gain came from ARGX alone (+16.04%). Large-cap pharma barely moved (LLY +0.25%, PFE +0.30%, GILD +0.40%, BIIB −0.32%).

The correct read: there was no such thing today as "money rotating into healthcare" — there was only "a Phase 3 dataset repricing one stock and its directly related names." Ranking it the No. 1 theme pre-market was correct on evidence grade, but inferring from it that "defensive pharma is taking the baton" would be a faulty extrapolation.

ARGX cross-listing check (used pre-market, redone at the close — and it reverses one of the pre-market reservations):

The pre-market list noted that "about 3 percentage points of the U.S.-line record high came from euro appreciation." Today that is reversed: the euro depreciated 0.92% on the day, so the U.S. line's +16.04% understates this move — the true gain in euro terms is +17.12%. And the Brussels close of €855.20 has now clearly cleared the prior high of €830.80 from 7/3 — both lines made new highs simultaneously, and the pre-market pitfall of "the two lines giving different conclusions" was filled in by the price action itself today.

Volume expanded to 2.3–2.6x on both listings simultaneously, which is the single most solid confirmation of this advance.

② Storage/semis: the market was not trading the narrative the pre-market list questioned

The pre-market list's core argument was: the anchor of the AI storage theme (the Anthropic 2028 revenue forecast) is soft, so it should not be chased. That factual judgment was not refuted today — but it was not the day's pricing mechanism. What actually happened today was a clean split between the buyers and sellers of capex:

MSFT was one of the single largest drags on the S&P 500 today (close $480.35, −3.04%). Per Motley Fool's report the same day, the driver was investors' reassessment of its AI capex trajectory and margins, plus profit-taking after the post-earnings rally, rather than any new negative.

This explains an apparent contradiction: XLK (technology) closed +0.16% today while its largest weight, MSFT, fell 3% — the semiconductor gains offset the declines in software and mega-caps. Looking only at XLK's +0.16%, you would see nothing of this internal split.

Correction to the pre-market framework: the pre-market list used "MU forward PE 6.76 / PB percentile 100%" to argue a peak-earnings pattern, and that remains a valid cycle-position judgment; but cycle position is not an intraday pricing mechanism. What was priced today is "whose cash-flow statement the money moves from and to," and that direction is clearly positive for storage.

One sourcing note that must be disclosed (a mirror image of the same problem in the pre-market list): several aggregator sites published, under 2026_08_17 URLs, "New Street upgrades MU to buy, target $1,250" and "UBS reiterates buy, target $1,625," which look like same-day catalysts. But both are falsified by their own price anchors: the New Street item references "compared with Thursday's close of $949.83" (i.e., the 8/13 close), and the UBS item references "the current share price of about $879" (where MU traded around 8/11) — while MU closed at $1,011.75 today. This is the mirror version of the very pitfall the pre-market list already stepped into (treating an 8/3 rating as an 8/17 catalyst). This piece could not confirm any new sell-side rating action dated 8/17, and therefore does not list ratings as a cause of today's storage rally; the confirmable background is TrendForce (8/12) and Micron management (8/11) commenting on tight memory supply/demand and price increases, both of which predate today.

③ Two things the pre-market list missed entirely

Both misses share one root cause: the pre-market framework asked "which narrative is more solid," while what the market rewarded that day was "which asset benefits from the same macro shock." War → oil ↑ → inflation expectations ↑ → real rates pressured → gold and bitcoin benefit in the same direction; the pre-market list traced this chain only halfway and stopped.

---

4. After-Hours Earnings Moves

No major earnings after the close today (16:00–17:00 ET). Monday is the earnings gap day of this week; retail season starts with HD tomorrow.

Basis note: after-hours liquidity is extremely thin, so the after-hours moves above serve only as a reference for the next open and do not carry the same pricing credibility as the regular session. As of 17:00 ET, no after-hours news constitutes a major catalyst for the next session — the next session's catalysts all come from tomorrow morning's calendar (see Section 6).

---

5. Flows and Sentiment

Sector ETF Panorama (sorted by change %)

Rotation Characterization

Today's flows can be summed up in one sentence: out of "rate-sensitive + consumer + mega-cap platforms" and into "real assets + their supply chains."

• Buy side: energy (XLE +1.08%), precious metals (GDX +2.13%), semis and storage (SMH +1.06%), biotech (IBB +1.85%). The first two are direct beneficiaries of oil prices and inflation expectations; semis are the recipients of AI capex; biotech is purely event-driven and unrelated to macro.
• Sell side: communication services (−1.89%), consumer staples (−1.64%), consumer discretionary (−1.23%), financials (−1.00%), real estate (−0.97%). Of these, real estate and financials are direct victims of long-end rates, and the two consumer groups map to retail earnings week and weakening consumer data.
• The single most notable item: consumer staples (XLP) fell 1.64%, more than three times the S&P 500's decline. Staples is the archetypal defensive sector and should hold up on a genuine haven day. It led the decline today, which shows money was not "reducing risk" but "repricing rates and input costs" — fully consistent with the stocks-and-bonds-down, bear-steepening signals.

risk-on / risk-off Characterization

Conclusion: neither a typical risk-on nor a typical risk-off day — it is a "reflation + rising term premium" combination.

Only one of the six items is fully consistent with a traditional risk-off. The correct description is: the market is repricing the chain "energy supply disruption → sticky inflation → long-end rates higher for longer," and the decline in stocks is the result of a rising discount rate, not a collapse in risk appetite. This also explains why the VIX rose only 0.94 points — a discount-rate shock does not require panic, only recalculation.

---

6. Next-Session Outlook (2026-08-18, Tuesday)

① Theme Continuity

② Tomorrow's Earnings and Macro Calendar (all times Eastern)

Earnings

Macro Data

Other Events

• Before the open: RDDT formally joins the S&P 500, replacing AvalonBay.
• META opening statements in the federal court for the Northern District of California (Oakland), with four states going first out of 29 states in total; the trial runs about seven weeks, and Zuckerberg and Instagram head Mosseri are expected to testify; the plaintiffs' maximum claim is $1.4 trillion (note: this is the upper bound of the claim, not a judgment; the first comparable jury verdict this March actually awarded $6 million).
• Effective 8/19 at 00:01: the 50% Section 338 tariff on Canada (including furniture), the same day as TGT/LOW earnings.
• 8/20: FOMC meeting minutes (the 7/28–29 meeting, the first time in a decade that three members argued for a hike in the same direction) + WMT earnings.

③ Tomorrow's Focus List (Ticker + verification point)

④ What to Avoid

1. Chasing a fourth gap day in storage. The pre-market reasons for "should not chase" (soft narrative anchor, high cycle position) were shown today not to be the day's pricing mechanism, but they still hold as a cycle judgment. SNDK has gapped three days running and rose another 5.06% after today's open, and MU's volume was only 0.99x its average — price is moving, volume is not following. The theme direction has been corrected to neutral, but the odds at this price level are still wrong.
2. Extrapolating today's "biotech is strongest" into a sector rotation. XLV was −0.19% today and large-cap pharma did not move at all. This is one stock's event, not the start of a sector move.
3. Rate-sensitive assets (XLRE / XLU / KRE). The 30Y is at its highest since 2007 and still in a breakout; until the long end stabilizes, every rebound in these three fights the discount rate.
4. Buying META for a bounce on the day of opening statements. A seven-week trial means headline risk is supplied continuously; today's −3.54% prices "the trial date," not the content of the testimony.
5. The RDDT index event. Passive buying was executed in today's closing auction; this catalyst is fully played out.
6. Using today's crypto rebound to trade a trend. HIVE +14.13% happened against a backdrop of BTC still −49% from its peak and miners still losing money; this is a beta rebound aligned with macro, not a fundamental reversal — the pre-market fundamental judgment was not wrong; the error was using it as a single-day directional forecast.

⑤ Input Notes for Tomorrow's Pre-Market List

1. Put "war / energy / long-end rates" straight at the top. It is the only main line fully verified today whose catalysts are not yet exhausted (the Bessent measures have not landed this week). And the transmission path missed pre-market must be added: gold and precious-metals equities (GDX/GLD), plus bitcoin under pressured real rates.
2. Change the storage/semis conclusion from "avoid" to "neutral watch," but swap in a different argument. Do not use "is the Anthropic forecast credible" to characterize this main line tomorrow — use "does the AI capex buyer/seller split continue" instead, with the relative strength of MSFT vs. MU as the main observable.
3. On HD earnings, forecast only the direction of guidance, not EPS. The reason is in Section 6②: comps and margin guidance are where the information content lies; with EPS expected up only +0.6% y/y, a beat or miss does not change the narrative.
4. Timeliness items that must be verified: ① whether there is any genuinely same-day sell-side rating action tomorrow — today's lesson is that aggregator sites repackage two-week-old ratings under same-day URLs, so the publication date must be reverse-checked using the price anchors in the report; ② the first-hand numbers in HD's earnings should come from the company press release / 8-K, not media paraphrase.
5. Basis discipline: tomorrow's pre-market list must continue to write a falsifiable verification point for every name — of today's 5 verification points, 2 directly overturned their own headline conclusions (OABI, AMAT), and the information value of the verification point exceeds that of the conclusion itself.
6. Do not flip wholesale to bullish just because 6 names were called wrong today. Today's overall hit rate was 57% (50% on the executable basis), and the errors are entirely concentrated in storage and crypto; the other eight directional calls were all correct. What needs correcting is the direction of one theme, not the whole framework.

---

Data Retrieval and Pipeline Failure Log (internal)

A. yfinance was intermittently rate-limited this time; the root cause was neither leftover processes nor a pure IP ban

• The first single-ticker call (SPY, period='5d') succeeded; two subsequent concurrent batches (12 indices + 21 ETFs) all returned 429 YFRateLimitError.
• Leftover processes were checked first per the established procedure: ps aux | grep -E "python3|yfretry" showed no leftover yfretry.py / /tmp/yfretry_fund.py, ruling out root cause A.
• But it was not a pure IP ban (root cause B) either: about 20 minutes later history() for ARGX.BR and ARGX recovered normally, while MU in the same batch still returned 429 — the rate limiting is at symbol/endpoint granularity, not machine-wide.
• Actionable conclusion (suggest updating memory): on this machine, yfinance rate limiting has a third form — short-lived throttling triggered by concurrency, characterized by "single call succeeds, concurrency fails immediately, partial recovery after about 20 minutes." Retrieval scripts should not issue concurrent yfinance batch requests; they should go serial with 0.4–8s intervals. After switching to stockanalysis this time, there were zero failures throughout (about 70 calls).

B. Channels verified working this time (consistent with the 08-12 log; can remain the default combination)

1. stockanalysis.com/api/quotes/s/<SYM> — the workhorse, 70+ calls with zero failures. In the after-hours session ep/ecp/eu return normally, and the u field confirms the 16:00 PM EDT close basis. Works for both ETFs and single stocks.
2. cdn.cboe.com/.../quotes/_SPX|_NDX|_RUT|_VIX.json — normal. last_trade_time was checked symbol by symbol: all four were 2026-08-17T16:09–16:15, fresh, and did not hit the _DJI/_COMP trap (neither of those was called this time).
3. home.treasury.gov daily yield curve CSV — normal, with 08/17 as the first row, all tenors present.
4. api.nasdaq.com/api/quote/COMP/info?assetclass=index — normal, returning 26,644.91 / −0.32%, independently corroborating the search results.
5. query2.finance.yahoo.com/v8/finance/chart/MU — returned normally while query1/yfinance was still 429 on MU, again confirming that query1 and query2 are banned independently. MU's daily series and prior 5-day average volume were obtained this way.

C. Conflicts among the three sources and how they were handled

1. [Most important] Oil prices: all three text sources were stale, and were overturned by an ETF cross-check.
   • TradingEconomics returned "Brent $88.31, −0.24%" — exactly the same number cited by this morning's pre-market list ($88.31), i.e., the page had not updated all day and was a stale reading.
   • Yahoo's live blog for the day: "Brent around $89, +0.4%; WTI $82, little changed" — this was a morning-session blog entry, not the close.
   • Actual: Brent +2.65% to $90.87, WTI +2.5% to $84.50 (NBC News).
   • It was three oil ETFs that exposed the contradiction first: USO +2.91%, BNO +2.63%, DBO +3.06%, all three pointing consistently to +2.6~3.1%. BNO (Brent) +2.63% vs. Brent's actual +2.65%, a deviation of only 0.02pp.
   • Takeaway: ETFs that settle at 16:00 ET are the best cross-check tool for cross-asset readings, because they naturally carry the correct timestamp, whereas spot quote pages and live blogs can both lag silently. This is the same class of failure as the existing "silently one day stale" note.
2. Market breadth: search summaries diverged from measured data, and the search value was discarded. One search summary claimed "NYSE advancers 1,141 vs. decliners 1,040 (positive breadth)," which could not be traced to a primary source; whereas measured RSP (equal-weight S&P) −0.89% vs. SPY −0.47% shows equal-weight clearly lagging — the opposite direction. The body of this piece uses the RSP/SPY basis, which can be recomputed independently, and does not cite that A/D figure. (The NYSE composite contains many bonds/preferreds/ETF-type securities and is not directly comparable to the S&P 500 basis to begin with, but since the source cannot be verified, it is simply not used.)
3. Aggregator errors in single-stock percentage moves: tradingkey said RDDT "−7.14%" and MSFT "−3.17%"; sundayguardianlive said MSFT closed at $479.15 / −3.28%. The three contradict each other. This piece uses the stockanalysis quote API throughout (RDDT $164.50 / −7.63%, recomputable from the $178.09 prior close; MSFT $480.35 / −3.04%, recomputable from the $495.40 prior close). Percentages from low-quality aggregators are unusable, but their "event narratives" are usable when corroborated by other sources.

D. One "old rating masquerading as a same-day catalyst" caught — the mirror image of the same pitfall in the pre-market list

The first draft was at one point going to write "New Street upgrades MU to buy / target $1,250" and "UBS reiterates buy / target $1,625" as the direct cause of today's storage rally, sourced from stockstotrade / timothysykes articles with 2026_08_17 in the URL. The falsification came from the price anchors the articles carry: the New Street item says "compared with Thursday's close of $949.83" (i.e., 8/13), and the UBS item says "the current share price is about $879" (where MU traded around 8/11) — while MU closed at $1,011.75 today. Both are old research repackaged under a same-day URL. Takeaway: an aggregator's URL date is not the publication date. Whenever citing a rating/target price, use the price anchor inside the article to reverse-check the corresponding trading day — this shares a root with the memory note "recurring column headlines hide the publication date," but the new identification technique, "reverse-check via the price anchor," is more reliable than checking the headline, because the price anchor is endogenous and cannot be faked. The body therefore explicitly states that "no new sell-side rating action dated 8/17 could be confirmed" rather than filling the gap with old news.

E. WebFetch failure list: the CNBC oil page 403; TheStreet's daily market page 403 (consistent with the previous log; these two domains can be treated as permanently 403 and need not be retried). The stockanalysis.com after-hours movers page and Yahoo's live blog page both returned normally.

F. Items not obtained / left blank in this piece

• The Dow's OHLC: only the close of 53,459.78 and the point change were obtained (from search sources); open/high/low were not obtained; the DIA ETF (−0.49%) served as a consistent directional cross-check. Existing memory already records that "the Dow has no reliable free API," and this is confirmed again.
• Official NYSE / Nasdaq advance-decline counts: no verifiable primary source was obtained, so the RSP/SPY substitute basis was used instead; see C-2.
• Pre-market quotes for ZLAB, RH and BTDR: the pre-market list itself did not obtain them, so the "vs. pre-mkt" column cannot be computed for these three; the table marks them "not obtained" rather than leaving them blank or inventing them.
• HIVE's pre-market price: the pre-market list recorded only the change of +12.64% without a price, so this piece back-computes approximately $3.03 from $2.69×1.1264, marked "approx." in the table, with the vs. pre-mkt column marked "approx. +1.3%."
• First-hand attribution for MSFT's decline: only a same-day Motley Fool article and two low-quality aggregators were available; no sell-side report or company-level primary event was obtained. The wording in the body is limited to "per Motley Fool's report the same day," not written as established fact.
• Consensus for HD's 8/18 earnings: $47.5B / EPS $4.71 comes from media paraphrase, not from Visible Alpha / company IR primary basis; it must be re-verified against the company press release in tomorrow's pre-market.

G. A self-check on the reconciliation basis

This piece gives two hit-rate numbers (57% on the close basis / 50% on the executable basis). Note that the denominator on the executable basis is 20, not 21 — ZLAB, RH and BTDR have no pre-market price, and HIVE's pre-market price is back-computed. To keep the denominator from being silently shrunk, missing items are annotated line by line in the tables. Reporting only the 57% close basis would overstate by 7 percentage points, which is exactly the bias the existing memory note "reconciliation must add a vs. pre-market price column" guards against, and it holds again in practice this time: of the 4 long-side names that "rose," 2 (OCUL −3.62%, OABI −4.74%) actually lost money if executed per the pre-market list.

---

⚠️ Risk disclaimer: this recap is an after-hours information review and observation only and does not constitute investment advice. Data may differ in timeliness or basis; please refer to company disclosures / SEC filings, and do not use this directly as a basis for trading.