US Market Recap | 2026-08-31 (ET) Monday

Coverage window: 2026-08-31 09:30–16:00 ET regular session; the reconciliation target is the same-day pre-market list reports/us/2026-08-31.md (pre-market read moment 08:00–08:17 ET).

Data conventions (source and read moment stated item by item)

• Close / open-high-low / volume: CNBC restQuote API, read moment 17:00–17:05 ET; last_time values all fall within 2026-08-31 16:00–16:20 ET, i.e. the regular-session 16:00 ET close convention. Key names (PCG/EIX/NOW/TSLA/VLO/SLB) were independently cross-checked against stockanalysis.com, with close, previous close, and open-high-low all matching item by item.
• Percentage changes are adopted only after this report reverse-computes them as (close − previous close) ÷ previous close, and are cited only when they agree with the API's own change_pct.
• The "vs pre-market price" column is computed by this report: using the 08:00–08:17 ET pre-market price recorded in the pre-market list as the base, it computes (close − pre-market price) ÷ pre-market price. This column answers "if you executed the pre-market list and filled at the pre-market price, are you up or down at the close" — a different question from "the day's percentage change".
• Treasuries: CNBC restQuote; the change is computed by this report as last − previous_close. The CNBC Treasury change_pct field once again contradicts its own last/prev (the 2Y field shows +0.0156% while last 4.344% < prev 4.350%), so it is not accepted.
• Crack spread: 3-2-1 computed by this report from CNBC restQuote second-month contracts (@CL.2/@RB.2/@HO.2); the 16:50 ET read moment is an electronic-session price, not the 14:30 ET settlement price, while the comparison base is Friday's settlement, so a slight convention mismatch exists and is flagged in §3.
• This report does not use yfinance; the reason is in the internal record at the end.

---

0. One-Line Recap

Today was a textbook "almost every direction right, but no money in it" day — the pre-market list hit 26/33 (78.8%) on the day's direction, yet 8 of the 15 long-side names closed below their own pre-market price: the direction was right, and that pre-market gap had already taken all the money off the table.

• Risk appetite: mild risk-off, but masked by one violent single-sector event. S&P −0.33%, Dow −0.70%, Nasdaq −0.12%, Russell −0.62%; VIX 14.92 (+3.40%), 10Y 4.752% (+3.0bp), DXY 99.414 (−0.29%). 9 of the 11 sector ETFs closed lower, with only XLE (+2.04%) and XLK (+0.44%) up.
• Strongest theme: the California wildfire liability repricing — the S-grade assigned pre-market not only delivered, it was far from enough. EIX closed at $53.98, −23.07% (described in the media as the worst single day since the 2001 California electricity crisis); PCG closed at $13.27, −20.06%. The pre-market readings were EIX −13.28% and PCG −16.39%, meaning these two fell a further 11.3% and 4.4% respectively during the regular session.
• The finest call on the pre-market list was not either of those two shorts — it was SLB. Pre-market it was re-classified from "oilfield services · oil price pass-through" to "AI data-center infrastructure", on the basis of the $4.1 billion Kelvion acquisition 8-K the company itself filed that day. SLB closed +4.83%, outperforming XLE by 2.79pp — that excess is precisely the price evidence for "not pure oil-price β", and it was also the largest gain versus pre-market price (+3.89%) on the entire list.
• The biggest miss was the software give-back: the whole leg reversed after the open. Pre-market listed NOW as the representative of "largest give-back" (pre-market −2.56%); it did indeed gap down to $142.08 at the open, then rallied all the way to a $147.99 close (+2.27%), a +4.16% rebound from the open; IGV, CRM, ADBE and INTU were all wrong on direction.
• The crack-spread falsification test the pre-market set for itself flipped over by the close. The 08:15 ET 3-2-1 was $62.39 (−$0.67), on which basis pre-market judged that "the refiner mechanism does not hold today, and VLO's move is front-running"; recomputed at the close it is $64.67 (+$1.61) — the mechanism did hold that day — but VLO opened at $359.96 and closed at $358.92, open→close −0.29%; the stock captured none of it.
• Tone for the next day: 9/1 is a binary event day, not a continuation day. As of this report's read moment, the last action on the California leginfo primary status page for SB 492 still stops at 8/30 "Ordered to third reading", with no 8/31 vote record whatsoever (see §6①). Today's −20%/−23% prices the already-finalized text, not a completed vote. Layered on top of PANW earnings after the close on 9/1 plus the August ISM manufacturing print at 10:00 ET, next-day risk is concentrated in those two points: after the close and in the morning.

---

1. Market Overview

Sentiment read: mild risk-off, and specifically a "duration and breadth both weak, energy the lone strong leg" combination.

1. Index declines were shallow, but breadth was weak. The Dow's −0.70% was clearly worse than the Nasdaq's −0.12%, with the Russell's −0.62% weak alongside; of the 11 sector ETFs, 9 fell and 2 rose. The Nasdaq's resilience came almost entirely from three names — TSLA (+5.51%), NVDA (+1.48%), MU (+2.77%) — while GOOGL −2.09%, AMZN −2.50%, META −0.98% and AAPL −0.89%, four heavyweights, fell together — calm at the index level concealed broad weakness in the large platform stocks.
2. The curve bear-steepened; it was not the bull-steepening the pre-market expected. The pre-market 08:08 ET readings were 2Y −2.1bp / 10Y +0.2bp / 30Y +1.1bp, judged as "partially unwinding Friday's bear flattening". Actual close: 2Y −0.6bp, 10Y +3.0bp, 30Y +3.5bp — the front end barely moved while the long end was sold all day. The direction is the opposite of the pre-market judgment, so this one the pre-market got wrong. Bear steepening moving in the same direction as oil's +3.19% reads more coherently as inflation expectations rather than policy expectations re-rating higher.
3. VIX +3.40% closing at 14.92 is still an absolute low. On a day when one S&P constituent fell 23% and another 20%, VIX rose only 0.49 point — that in itself is pricing evidence that "the market treats this as an isolated California regulatory event, not a systemic credit event", corroborating the control-group conclusion in §3.
4. Market breadth: per the Zacks/Yahoo Finance same-day close wrap, the NYSE decliner-to-advancer ratio was about 1.77:1. ⚠️ This item is a secondary source that this report could not independently verify; treat it as reference only. All sector and single-stock figures in this report are self-pulled and self-computed and do not depend on that source.

One convention correction that must be stated: several same-day close wraps said "10 of 11 sectors fell, 1 rose" and that "XLK rose 3.2%". On this report's own measurements, XLK closed at $186.50, +0.44%, and XLE +2.04% — that is 2 up and 9 down. These two figures from the secondary wraps do not match this report's measurements, and this report goes with self-pulled data throughout.

---

2. Pre-Market List Reconciliation

This section measures with two rulers at once, because they give completely different answers:

• "Day %" = close versus previous close, answering "was the pre-market judgment of market direction right";
• "vs pre-market price" = close versus the pre-market price recorded in the pre-market list, answering "if you executed this list at the pre-market price, do you make money by the close". On the long side these two rulers give nearly opposite conclusions, and that is the single most important thing about today.

2.1 Long Side (15 items)

Long side: direction on the day 15/15 = 100%, versus pre-market price 7/15 = 46.7%. That 53pp gap is today's core lesson: all 15 names rose, but for 8 of them the gain was already over before 09:30. The energy sector was a textbook post-geopolitical-gap flatline: XLE opened at $63.80 and closed at $63.96, moving just +0.25% across the whole day. Only one name on the pre-market list actually created excess return — SLB — and it happens to be the only one that did not ride sector β and had a primary-source company-specific catalyst of its own (the 8-K).

2.2 Short Side · Short Watch (2 items)

Short watch 2/2, and the only group on the whole list where both rulers — "direction on the day" and "vs pre-market price" — were strongly positive at the same time. Two details worth recording:

1. The three reasons behind the pre-market call that "EIX has better odds than PCG" — a smaller gap, roughly 14% of room still left to the 52-week low, and the unpriced independent risk of the adequacy of the Eaton provision — all delivered. EIX fell a further 11.29% versus its pre-market price, 2.6x that of PCG (−4.39%). EIX's intraday low of $53.11 leaves only 2.1% to the $52.00 52-week low cited pre-market, essentially covering that 14% of room.
2. But one pre-market reading was judged backwards. Pre-market observed PCG's decline narrowing from −18.4% at 08:02 to −16.39% at 08:17 while turnover expanded, and inferred "this looks more like buyers stepping in at the gap than one-way position cutting". It closed −20.06% on 138 million shares (above Friday's 114.2 million), so the decline widened rather than narrowed. Narrowing declines during the pre-market session cannot be extrapolated to the regular session — this was a case of mistaking pre-market noise for a flow signal.

2.3 Short Side · Watch Only / Directional Watch (11 items)

7/11 delivered. All four misses cluster on the single "month-end software give-back" leg; see theme four in §3. All four airlines were right on direction (4/4), and the pre-market handling — "four names agreeing on direction raises the credibility of the group, but thin volume in any single name cannot be treated as fact" — was appropriate. SRE was the finest call in this section: pre-market explicitly wrote "SRE is the wrong expression vehicle; 73.5% of its profit has nothing to do with this logic". Today SRE opened at $79.98 and closed at $81.70, a +2.15% rebound into the close and +1.29% versus its pre-market price — anyone shorting SRE to express this logic lost money today, while those short PCG/EIX made 4.4%/11.3%. Same logic; picking the wrong instrument alone determined whether you made or lost money.

2.4 Avoid List (5 items)

This block is the easiest thing to misread today, and it must be judged on OHLC rather than the close-to-close percentage change. Looking only at the day's percentage change, RDHL +36.16% and AEHL +83.33% look like two enormous missed moves. But bring the opening price in and the conclusion flips completely:

• RDHL opened at $1.36, high $1.60, closed at $0.90 — closing at the low of the day, −33.82% from the open and −43.75% from the high.
• AEHL opened at $6.76, high $8.48, closed at $6.49 — likewise below its open, −23.47% from the high; and it fell a further 16.03% after hours to $5.45.
• VVOS is more extreme still: −12.85% on the day, but from the $0.395 open to the $0.196 close is −50.38%. In other words, all of the gains in these three names occurred pre-market, and anyone executing the list after the regular-session open lost heavily across the board. On an executable basis, the pre-market handling of "avoid across the board, do not participate, and do not invent reasons for them" was 4/4 correct, not the 2/4 it appears to be on the surface. The only genuine miss is SMCI (+0.54%), but the magnitude is small and it underperformed SMH (+0.64%) on the day, so the "persistent weakness with no catalyst" characterization still holds on a relative-strength basis.

2.5 Hit Rate and Self-Critique

Three points of self-critique:

1. What deserves to be remembered is not the hit rate, it is that 53pp gap. Long-side direction was 100% right while the executable convention was only 46.7% — on the energy leg, the pre-market list was "confirming after the fact a gap that had already happened", not "discovering in advance a repricing that had not yet happened". The only name that broke that pattern was SLB, and its one distinguishing feature is that the catalyst was an 8-K the company itself filed that day, not sector β. The pre-market judgment that CVX's "$237.5 million of pre-market turnover is the best confirmation in the energy sector" turns out, in hindsight, to have confirmed the reality of the gap, not its subsequent persistence — the two were conflated.
2. The inference "narrowing decline + expanding turnover = buyers stepping in" was falsified on PCG. Pre-market observed −18.4% → −16.39% between 08:02 and 08:17 with turnover expanding from $36.6 million to $51.1 million, and inferred buying interest; it closed −20.06% with volume above Friday's. Price trajectories during the pre-market session cannot be extrapolated to the regular session — this is another face of the same class of error as the existing lesson that "pre-market volume and price are cumulative; a single snapshot is not a fact": a trajectory is likewise not a fact.
3. The whole software give-back leg was wrong, and the pre-market had in fact already written down the right answer without adopting it. Pre-market explicitly wrote that "this leg has an hourly shelf life, not a daily one" and that it "can be falsified within 30 minutes of the open" — and it did not even survive 30 minutes past the open, with NOW rallying from the $142.08 open all the way to $147.99. The qualifiers given pre-market were right, yet 6 names were still put on the list. When a piece of logic is itself annotated as "valid on an hourly basis", the correct action is to leave it off the day's list, not to include it and then append a disclaimer.

---

3. Theme Verification

③④ Crack Spread: The Pre-Market's Own Falsification Test Flipped by the Close

Pre-market wrote this up as the sharpest self-falsification of the day: "Recompute the 3-2-1 (using second-month contracts) once after 09:30 and again at 14:00. If RBOB/ULSD gains catch up to and exceed WTI → the crack widens and the refiner logic holds for the day". The answer is: they caught up.

The key is that the two legs diverged: ULSD (heating oil/diesel) +4.43% outran WTI's +3.19%, while RBOB (gasoline) +1.86% still lagged. In the 08:15 pre-market reading both legs were lagging, which is why it computed as narrowing; by the close the diesel leg had overtaken, and the 3-2-1 widened by a net $1.61/bbl. This is fully consistent with the structural narrative cited pre-market — Jefferies' "the Hormuz shock shows up in crack spreads rather than crude" plus distillate inventories at their lowest for the period since 1996: what is tight is distillate, not gasoline. ⚠️ Convention qualifier: 16:50 ET is an electronic-session reading while the Friday base is a settlement price, so a slight convention mismatch exists; but the magnitude of +$1.61 is far larger than any error that mismatch could introduce, so the directional conclusion holds. Yet none of this helped the refiner stocks on the day: VLO opened $359.96, closed $358.92 (open→close −0.29%); MPC opened $374.24, closed $373.32 (−0.25%); PSX opened $248.12, closed $246.58 (−0.62%). All three refiners closed below their opening price, on a day when their profit function was improving. The correct reading: pre-market judged that "VLO's move this morning is front-running and the mechanism does not hold today" — at the mechanism level that judgment was overturned by the close, but the trading-level conclusion (reduce position; the gain has no same-day basis) was right after all. This is a case of "wrong reason, right conclusion" and deserves its own marker: the pre-market position guidance holds for a different real reason — crowding (YTD +116.5%, P/B 4.06) and thin pre-market volume ($7.3 million) — not the day's crack direction.

① California Wildfires: The Control-Group Test Passed Cleanly — the Most Solid Piece of Pre-Market Work

The overturn condition set pre-market was: "if DUK also falls more than 1.5% after the open, the "pure California regulatory event" characterization must be overturned".

DUK −0.28%, AEP +0.10%, NEE +0.61% — all five non-California utilities were inside ±0.6%; the verification point passed cleanly. Declines ranked strictly by California power exposure: pure California −23%/−20% ≫ mixed −3.1% ≫ no California ≈ 0. The characterization that "this is a purely California regulatory event, not a utilities sector event" holds completely. And the quantitative check on SRE remains self-consistent: pre-market used segment profit to derive SRE's true wildfire exposure of 26.5% and predicted a reaction of roughly 1/3.8 that of the core names. Today SRE −3.10% versus a −21.6% average for the two core names is a ratio of 1:7.0 — the direction of exposure dilution was right, but the magnitude was underestimated: SRE was even more resilient than a linear extrapolation from exposure would imply (which would be about −5.7%). The likely reason is its 2.15% intraday rebound (open $79.98 → close $81.70), i.e. the market actively removed SRE from this piece of logic during the session.

⑤ Software vs Semiconductors: Half Right, Half Wrong

The three semiconductor equipment names diverged: AMAT −0.71%, LRCX −0.14%, KLAC −0.05% all closed lower, while NVDA/MU/AMD rose — the "semiconductor re-add" actually happened only on the compute-chip side; the equipment side did not participate. Pre-market put AMAT/LRCX/KLAC on the same leg as NVDA, and that granularity was not fine enough. NOW's reversal is the most worth recording: it was deemed pre-market the purest representative, ticking all three boxes of "August's big winner (+30.10%) + up again on Friday + leading the decline this morning". It did gap down to $142.08 at the open (delivering the pre-market direction), then rose one-way all day to $147.99, with an intraday high of $149.60 on 23.9 million shares. A delivered gap-down followed by an all-day reversal shows that the pre-market selling pressure was a one-off release of overnight orders, and that the real regular-session flow ran the other way.

Unexpected Themes: Two the Pre-Market Missed Entirely

1. TSLA +5.51%, the biggest gainer in the S&P 100 on the day, and not mentioned anywhere in the pre-market list. It closed at $367.95, opened at $347.21 (essentially equal to the previous close), so the entire gain happened inside the regular session, not in a gap — meaning this was a leading theme that a pre-market scan could not in principle catch, yet was fully tradable intraday. Volume 58.52 million shares.

   Catalyst attribution is questionable and this report offers no verdict: the explanations retrieved include Musk's weekend post about SpaceX making its own gas-turbine blades, Nevada approving Tesla Robotaxi as an autonomous-vehicle network company, and FSD regulatory progress, but they are mostly second-hand restatements that contradict one another and I could not verify them against a primary source. The fact that "TSLA was the strongest heavyweight of the day" is certain, however, and missing it is a genuine gap in the pre-market scan convention. The mechanism is consistent with an existing lesson: the pre-market scan convention is "names that are moving pre-market", and TSLA's opening price was essentially equal to its previous close (+(−0.44)%), so it was completely invisible in the pre-market screen. This is the same gap the pre-market itself admitted in its §1.1 — "systematically missing the entire layer of names that have a catalyst but do not move pre-market" — the pre-market had already identified this methodological hole and patched it with an analyst-rating layer, but never patched the theme layer.

2. The crypto chain was the second-strongest leg of the day, and pre-market gave it the lowest strength on the whole list, C+, plus "watch only". COIN +5.31% (open $178.51 → close $188.12, open→close +5.38%, likewise entirely intraday), MSTR +4.42%, IBIT +1.75%. The pre-market reasoning was "turnover is adequate, but I could not find any corresponding news catalyst", so it was classified as a technical rebound. In hindsight, "no catalyst found" was used as the basis for "low strength" — but those are not the same thing: finding no catalyst only means the evidence is insufficient, not that the momentum is weak. This is yet another recurrence of "cannot find news ≠ there is no news", and this time it led the market two days running.

Pre-Market §1.1 Rating Layer: It Said "Divergent, No Conclusion", and the Market Answered That Same Day

In the named rating layer the pre-market added, DE (PT $640→$800) and AGCO (PT $120→$150), both sharply upgraded by Baird, rose +3.90% and +4.34% today, both among the day's leading large-cap gainers.

The pre-market handling of this block was: "the Canadian retaliation list includes agricultural machinery vs Baird's same-day sharp upgrade — two pieces of evidence pointing in opposite directions coexist, so this report gives no conclusion and only flags the divergence". The market's answer today was unambiguous: the upgrade side completely overwhelmed the tariff negative. And the pre-market had in fact already found the decisive qualifier — a RealAgriculture check showed the list mainly covers components and does not broadly cover complete tractors, combines, planters and sprayers, so the actual damage from "tariffs on agricultural machinery" is far smaller than the phrase implies. In other words, the pre-market already held evidence sufficient to break the "divergence", and chose not to conclude. "Flagging divergence" is over-caution when the evidence is asymmetric: when one of two pieces of evidence has already been weakened by primary-source verification, a directional judgment should be given rather than presenting them side by side. Verification of the rest of the rating layer: LITE +2.21% (Evercore initiation, PT $1,100) ✅; FRO −0.93% (Nordea downgrade) ✅ — pre-market specifically flagged that "a downgrade of tanker stocks runs counter to the intuition that 'a blockade is bullish for oil shipping' and deserves separate follow-up"; today, against a backdrop of oil +3.19%, FRO closed lower against the tape, so that counter-intuitive judgment was right.

---

4. After-Hours Earnings Moves

There were no large-cap earnings after today's close, so this section is substantively empty — which is itself consistent with the pre-market judgment.

Pre-market had explicitly stated that the week's two tech earnings confirmed by primary company sources fall after the close on 9/1 (PANW) and 9/2 (AVGO), with nothing scheduled for 8/31 after the close. Checking the stockanalysis after-hours movers list (read moment 17:0x ET), the top five movers in both directions are all extremely small micro-caps, none of them earnings-driven:

The one item relevant to this list: AEHL fell a further 16.03% after hours to $5.45. It is precisely one of the micro-caps named on the pre-market avoid list (pre-market +67%); it had already fallen from its $8.48 high to $6.49 by the close and kept sliding after hours — the "avoid across the board, do not participate" handling was validated once more after the close.

The next day's after-hours session is the real earnings node (see §6).

---

5. Flows and Sentiment

5.1 Sector Rotation (SPDR sector ETFs, 16:00 ET close, measured by this report)

Three readings:

1. Energy was the only sector with a meaningful gain, and it was a geopolitically driven one-legged move. XLE +2.04%, in line with WTI +3.19% and Brent +2.75%. But XLE opened at $63.80 and closed at $63.96, just +0.25% across the whole day — this was a "gap then flatline", not "sustained intraday buying". The positioning was completed before the open.
2. XLU's −1.17% is a badly misleading number. It came almost entirely from PCG and EIX; excluding those two, the five non-California utilities (DUK/AEP/SO/NEE/ED) averaged −0.06%. Treating XLU as evidence of "rates weighing on utilities" is a completely wrong reading today — precisely the error the pre-market control-group test was designed to prevent, and today re-confirmed its necessity.
3. XLC −1.35% at the bottom, dragged down by GOOGL (−2.09%) and META (−0.98%); XLK's +0.44%, meanwhile, rests almost entirely on semis. The two "technology" sectors moved in opposite directions, showing that there was no unified technology style today, only the single narrow leg of compute chips.

5.2 VIX and Treasuries

• VIX 14.92, +3.40%, still low in absolute terms. On a day when one S&P constituent fell 23% and another 20%, VIX rose only 0.49 point — the market clearly priced this as an isolated, containable regulatory event rather than a systemic credit event. This corroborates the §3 control-group conclusion (DUK −0.28%), two independent data sources pointing at the same conclusion.
• The curve bear-steepened: 2Y −0.6bp to 4.344%, 10Y +3.0bp to 4.752%, 30Y +3.5bp to 5.243%. The front end did not move while the long end was sold all day, in the same direction as oil's +3.19%, which reads more coherently as rising inflation expectations than as rising policy expectations.

  ⚠️ Contrary to the pre-market judgment: the pre-market 08:08 ET readings showed a slight bull steepening (2Y −2.1bp / 10Y +0.2bp / 30Y +1.1bp), judged as "partially unwinding Friday's bear flattening". The close was a bear steepening; the direction was called backwards.

• Dollar DXY 99.414 (−0.29%), gold −0.74% — on a day when oil surged 3.19% and the Middle East kept escalating, both the dollar and gold fell. This combination is consistent with the pre-market conclusion: safe-haven assets showed no panic-level displacement. Pre-market honestly flagged that it "cannot separate the two variables of geopolitical pricing and the Warsh hawkish repricing"; today's data still cannot separate them, that qualifier remains in force, and it should not be escalated into a stronger claim.

5.3 Risk-On / Risk-Off Characterization

Mild risk-off, but the structure is extremely uneven and should not be summarized with a single label.

The most coherent reading: today was not an across-the-board contraction of risk appetite but a rotation "out of large platform stocks and defensive sectors and into energy, compute chips and high-β speculative assets", plus a California regulatory event unrelated to the broad market that dragged the indices lower by roughly 0.1–0.2pp. Crypto and TSLA leading at the same time contradicts the "risk-off" label — so the characterization this report gives is "the index appearance of mild risk-off with a risk-on core in the flows", not simply a haven day.

---

6. Next-Day Outlook (Tuesday 9/1)

① Theme Continuation

California Wildfires: The Single Most Important Thing for the Next Day — the Vote Status Itself Is Still Unconfirmed

This is the passage in this report that requires the most careful wording, because the primary source and the media accounts do not agree.

• Primary source (leginfo.legislature.ca.gov, SB 492 status page and history page, this report's read moment about 17:0x ET / 14:0x PT): the bill's status is still "Active Bill – In Floor Process", and the last action in the history record stops at 8/30 "Ordered to third reading"; 8/31 contains only an Assembly third-reading file scheduling entry (3rd Reading File, Item 112). There is no record of any 8/31 vote, Senate concurrence, enrollment or transmittal to the governor.
• Media accounts: several financial media outlets reported on 8/31 that SB 492 had been "passed / approved", and explained the day's decline on that basis.
• This report's handling: the two are not necessarily contradictory — the media's "passed" mostly refers to the amended text finalized on 8/29 and the content of the bipartisan agreement, not a completed floor vote; and leginfo's history page does lag in updating, so "no record" does not equal "did not happen". But until the primary source confirms it, this report will write neither "SB 492 has passed" nor "SB 492 has not passed".

⚠️ One search lead that must be retracted: during the search, a claim appeared that "SB 492 passed the Senate without objection on 8/24 and must be voted on before 9/1 (the 72-hour print rule)". After checking the original text, that report is confirmed to be about a different bill, AB 2700, not SB 492 — the search summary had stitched two bills together. This report does not adopt that timeline.

Practical implications for the next day:

1. Today's −20% / −23% prices the publicly available 8/29 amended text, not a completed vote. If a vote result, floor amendment or governor's statement emerges the next day, that is still incremental information.
2. The recess deadline for this session of the California legislature falls around 8/31, so the window is extremely narrow. Whatever the outcome, a definitive status update is likely around the next day's open — and that is the main source of PCG/EIX volatility the next day.
3. EIX's intraday low of $53.11 leaves only 2.1% to the $52.00 52-week low cited pre-market. That technical level will be tested directly the next day.

② Next-Day Earnings and Macro Calendar

③ Key Watch List (Ticker + Verification Point)

Per the existing lesson, write the verification point first, then discuss the conclusion.

④ What to Avoid

• Chasing shorts in PCG / EIX. The direction has been amply validated, but the two-day cumulative declines already reach PCG −26.1% (from $17.95 on 8/27) and EIX −26.7% (from $73.68 on 8/27), and an unsettled vote status means a binary outcome in the bullish direction is still possible. Most of the odds have already been eaten by the price.
• Chasing energy and refiners at the open. Today's lesson is extremely clear: 8 of the 15 long-side names closed below their pre-market price, and the entire return on the energy leg was in the gap. For a sector that gapped then went flat, the expected value of chasing at the next day's open is negative.
• Using XLU or SRE to express the California logic. Excluding PCG/EIX, today's XLU −1.17% leaves only −0.06%; SRE rebounded +2.15% into the close (from the open), so shorting it lost money today.
• The micro-cap gainers list. Re-validated today: RDHL open→close −33.82%, VVOS open→close −50.38%, AEHL a further −16.03% after hours; the gains were entirely concentrated pre-market, and executing in the regular session lost heavily across the board.
• Extrapolating "month-end rebalancing" into September. Pre-market had already flagged that this theme "expires at today's close", and today it did not even hold for 30 minutes past the open.
• Taking a large position in any direction ahead of ISM and the 9/4 payrolls. The rate-hike probability is 60.4% and still rising, so strong data is a negative.

⑤ Input Prompts for the Next Day's Pre-Market List

1. The first task is to confirm SB 492's final vote status, and it must be checked on the leginfo primary status page; procedural restatements from financial media must not be accepted. The pre-market list has already corrected a media error on this exact point once ("passed at the 8/29 adjournment"), and today the media again widely wrote "already passed" while leginfo has no record — the same pit two days running.
2. The scan convention must add the layer of "no pre-market move but leads intraday". Today's two strongest legs (TSLA +5.51%, COIN +5.31%) both opened essentially at their previous close, with 100% of the gain occurring inside the regular session, so they are invisible in any screen based on pre-market moves. The pre-market list has already added a rating layer for "has a catalyst but did not move pre-market", but the theme layer is still missing.
3. Long-side names must be labeled as to whether the catalyst is a primary company source or sector β. SLB, the only name that created excess return today, is the former, while all the other 14 items are the latter and on average closed below their pre-market price. Recommend adding this field directly to the list.
4. The handling of "no catalyst found" needs correcting. The crypto chain has led two days running with still no verifiable catalyst, and pre-market gave it the lowest strength, C+, on that basis. "Insufficient evidence" should lead to "do not bet big", not to "judged weak" — those two conclusions are identical in position size and completely different in attention.
5. Carry forward the two methods validated as effective today: ① the control-group test (DUK made the "pure California event" characterization falsifiable the same day); ② measuring pre-market liquidity by notional turnover rather than share count (PSX's $0.8 million was correctly flagged as weak evidence, and it was −0.82% versus its pre-market price on the day).
6. Add one hard rule: for any name whose gains are concentrated pre-market, the opening price must be given alongside, not just the pre-market price. The true conclusions on today's micro-caps and energy names are only visible once the opening price is brought in.

---

Data-Pull and Pipeline Failure Record (Internal)

• yfinance was unusable throughout. The very first call threw YFRateLimitError: Too Many Requests (at the _get_crumb_basic stage); this report made no further retry attempts and switched directly to the dual channel of CNBC restQuote + stockanalysis, consistent with the existing conclusion (retrying is pointless under a pure IP ban). All market data in this report comes from CNBC restQuote, with key names independently cross-checked via stockanalysis /api/quotes/s/<sym>.
• CNBC restQuote works normally in the post-close session: with exthrs=1, single-stock last_time values all fall within 16:00–16:20 ET that day, consistent with the existing conclusion: CNBC is usable for the post-close slot, while the pre-market slot must go through stockanalysis.
• The CNBC Treasury change_pct field breakage recurs for the Nth time: US2Y shows +0.0156% while last 4.344% < prev 4.350%, self-contradictory. All Treasury changes in this report are self-computed as last − prev.
• CNBC single-stock extendedMarkPrice / extendedMarkChangePct all returned None, so after-hours moves were instead taken from the stockanalysis after-hours list, which works.
• The leginfo pages appear to be lagging in cache: both billStatusClient and billHistoryClient only go up to 8/30, and the status page footer date shows 08/29/26. It is impossible to distinguish "the vote has not yet happened" from "the page has not updated", and the main text is worded to reflect that uncertainty without drawing a binary conclusion.
• A search-stitching error that nearly made it into the main text (must be recorded): searching "SB 492 Assembly floor vote" returned a summary stating "passed the Senate without objection on 8/24, must be voted on before 9/1 due to the 72-hour print rule", which appeared to answer exactly the question of why leginfo has no 8/31 record and was highly persuasive. After WebFetching the original, that report (Press Democrat) turns out to be about AB 2700, not SB 492. Had it been accepted at face value, the main text would have stated an entirely fabricated "9/1 voting deadline" and built the whole next-day outlook on it. Lesson: search summaries stitch together facts about different bills/entities, and the stitched result often happens to fill exactly the hole you are looking to fill — the more a search result "explains things just right", the more it must be checked against the original for whether the subject is the same.
• Same-day close-wrap secondary sources conflict with measurements: one wrap said "10 of 11 sectors down, 1 up" and that "XLK rose 3.2%", whereas measurements show XLK +0.44% and XLE +2.04%, i.e. 2 up and 9 down; another said "the S&P closed at 7,711.76", which is actually the prior day's close. Consistent with the existing conclusion: same-day close-wrap summaries cannot be quoted directly, and all sector data in this report is self-pulled. The NYSE advance/decline ratio of 1.77:1 could not be independently verified and is flagged as secondary in the main text.
• PCG volume differs between two sources: CNBC reports 138,095,264 shares, stockanalysis reports 154,554,479 (EIX likewise differs, 23,841,723 vs 24,686,825), apparently different consolidated-exchange conventions. The main text quotes the CNBC convention and uses it only for the qualitative comparison "above Friday's 114.2 million shares"; that conclusion holds under both conventions, so it does not affect the judgment.
• Crack-spread convention flaw (disclosed in the main text): the closing reading is taken from the 16:50 ET electronic session while the Friday base is a settlement price, so the two conventions are not fully consistent. It has been confirmed that @CL.2/@RB.2/@HO.2 all have last != prev with last_time falling on the day, ruling out the stalled-contract pit (the @RB.1 expiry problem hit in the previous report). The magnitude of +$1.61 is far larger than the possible error from the convention mismatch, so the directional conclusion is robust.
• The TSLA catalyst could not be verified against a primary source: the search returned multiple mutually inconsistent claims — Musk's post about SpaceX gas-turbine blades, the Nevada Robotaxi approval, FSD progress — all second-hand restatements, and no single attribution was accepted; the main text states only the price facts and flags the attribution as questionable. This is a deliberate blank left in this report: better to admit that a leading theme was missed than to back into a reason.
• This report launched no sub-agents; all data was self-pulled and self-computed. The cost is that market breadth (advance/decline counts) can only be cited from a secondary source; if it is to be filled in later, breadth can be self-computed by pulling closing prices name by name for a fixed 100–500 name sample pool, reusing the same script as the pre-market list's 127-name scan pool.

⚠️ 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 convention; please defer to company disclosures/SEC filings, and do not use this directly as a basis for trading.