Flows, Not Fundamentals

The rotation into safety

To understand last Thursday you have to start a month earlier. Through October the dominant flow in US equities was rotation, not selling. A heavy earnings season, a packed catalyst calendar and the corporate buyback blackout, the weeks before earnings when companies cannot buy back their own stock, all landed at once. With the largest structural buyer sidelined, investors took profits in semiconductors and AI infrastructure and moved into sectors that hold up when uncertainty rises: financials, utilities, consumer staples and healthcare.

We positioned for exactly that, with a pair trade long the defensive sectors and short semiconductors. It worked. Semiconductors (SOXX) fell roughly a third from their early-October high and still sit well below their 200-day average, while healthcare, staples, financials and utilities rose 10–13% from their autumn lows. Utilities stand out after months of heavy selling, and even the cyclical laggards, discretionary and materials, joined in.

The equal-weight to cap-weight ratio (RSP/SPY) shows the rotation in one line: it has climbed from its October low to the top of its range for the year. It also explains why the index barely moved in October. The S&P 500 is weighted by market value, so gains in hundreds of smaller stocks were offset by losses in the few AI giants that drive it. The reverse is also true: when money rotates back into the mega caps, the S&P 500 rises much faster than the average stock. That is the move our NVIDIA trade is positioned for.

By early November, then, the market had already been de-risking for weeks. It was lighter in AI, but still sitting on a large pool of systematic positioning that responds to price and volatility, not to views.

Thursday 5 November: anatomy of a mechanical sell-off

A one-day post-election rally

US stocks tend to rally after midterms, once the uncertainty clears, and on Wednesday they did: Democrats won both chambers of Congress and the market posted its best session in three months. By Thursday the reading had changed. Divided government means slower legislation and less fiscal support, and Democratic gains in states where opposition to data centres has been building raised fears that policy could slow the AI build-out. Investors read the result as disinflationary but growth-negative, and a market sitting on large AI gains began to take profits.

NVIDIA then took a second hit. Over the weekend, reports linked an executive at a Singapore-based customer under US investigation to Chinese intelligence, drawing in CEO Jensen Huang, and the stock fell again on Monday. Nobody has accused NVIDIA or Huang of wrongdoing, but in a shaken market a headline like that hits the most crowded stock first.

An election result does not change a single company’s earnings this quarter. A 4.4% drop needs a bigger explanation: who was forced to sell, and when.

The mechanics: how a sell-off feeds itself

The first group of forced sellers is CTAs (Commodity Trading Advisors), systematic trend-following funds. They add exposure when prices trend higher with low volatility and cut it when the trend breaks or volatility rises. Because they follow similar rules, they sell together, in size, and their selling can be estimated in advance.

This is why we turned cautious before the sell-off. Early last week, Goldman Sachs’ CTA model showed flows close to zero in a flat market, but in a falling market about $40 billion of selling within a week and $164 billion over a month, roughly $69 billion of it in US equities. A modest decline would unlock a very large seller, and on that basis we flagged that we would be looking for protection into the week. On Thursday, that seller arrived.

The same model projects about $120 billion of buying over a month if the market keeps rising: the force that accelerated the sell-off can just as easily fuel the recovery. Goldman’s key thresholds are 7,713, 7,544 and 7,082, with the medium-term 7,544 the most important. Holding above it is the line between CTAs as sellers and CTAs as buyers.

CTAs explain where the selling came from; dealers explain why it accelerated. Options dealers, the banks and market makers on the other side of investors’ trades, stay neutral by hedging in futures and stock. When they are short gamma, as they were last week, that hedging works with the move: they sell as the market falls and buy as it rises, amplifying moves instead of absorbing them.

Put the two together and you get Thursday. Profit-taking pushed the market lower, tripping CTA sell signals. CTA selling pushed it lower again, forcing dealers to sell more to stay hedged. Volatility-control funds, which cut exposure as volatility rises, added another layer. The loop ran into the close, which is why the index finished near its low on very heavy volume.

The key level is the dealer ‘flip’, where gamma turns positive. On SPY it sits at about $760, only 0.8% above spot at $754, roughly 7,600–7,650 on the S&P 500. Above it, dealers buy dips and sell rallies, and ranges hold. Below it, negative gamma deepens quickly towards $720, about 4–5% lower, with the largest put concentration at $729. Call gamma builds steadily towards $800, and heavy put positioning at $750–760 expires on 20 November, removing a source of downside pressure.

The volatility tell: panic, but short-dated

Thursday produced the third-largest one-day jump in the VIX since 1990, from 16.6 to 28.2. About 800,000 VIX calls traded by midday, three and a half times normal, and demand for S&P 500 puts surged. But the fear sat almost entirely at the front: nine-day implied volatility hit 31.4 while six-month and one-year barely moved, at around 25. The market priced a dangerous few weeks, not a bear market. In a genuine regime change, the back of the curve moves too.

VIX futures agreed: with spot at 28.2, November traded at 24.6 and December at 22.2. On Friday the unwind began, with the VIX down 9% and the VVIX down from about 135 to 122. As the VIX falls, the puts dealers are short lose sensitivity and dealers buy back their hedges, a flow known as vanna that can fuel a rebound on its own. After the last ten one-day VIX jumps of 50% or more, the S&P 500 was higher a month later in nine, according to Yahoo Finance.

The rebound: do we believe it?

The S&P 500 found its floor almost exactly at the 200-day moving average, and the next two sessions delivered a strong rebound on solid volume. Futures are back around 7,609, above Goldman’s 7,544 CTA threshold. Because so much of the decline was mechanical, we think most of the recovery is real, and its breadth supports that:

• It is broad. Nearly every sector is bouncing, equal-weight RSP has recovered almost all of the decline, and the under-owned Russell 2000 is recovering as yields fall.
• Credit agrees. High-yield and investment-grade credit ETFs are being bought back.
• Fewer forced sellers, more structural buyers. Leveraged and inverse ETFs, which must rebalance into the close and amplify big moves, have shrunk by $55 billion in 20 days, with gross exposure down $135 billion (Goldman Sachs). At the same time, companies are leaving their buyback blackouts.
• The weak link is AI. The Nasdaq 100 is still below its 200-day average and semiconductors well below theirs. That unrepaired gap is what our NVIDIA trade is designed to capture.

The signals we are watching

Our view rests on the move having been mechanical. If it was not, the evidence will show up first in credit, rates and volatility.

Credit is the line in the sand. Spreads are at their highest since the spring, though far from April 2025, when high-yield spreads blew out above 4.5%. More importantly, HYG and LQD, the ETFs institutions use to trade credit risk quickly, are already back at their 200-day averages. ETF prices tend to lead the spread indices, so this is an early sign that spreads will follow.

Macro: divided government is quietly disinflationary

With Congress in Democratic hands and the White House Republican, legislation slows, fiscal support becomes harder to deliver and new constraints on data centres add friction to one of the economy’s largest sources of investment. Over the medium term, slower growth and slower policy are disinflationary, and the dollar, which we believe has been trading rich, is rolling over. For equities that removes pressure on valuations. For AI it cuts both ways, which is why we use defined-risk structures.

The near term is less comfortable. After falling from around 5.3% in October to about 4.86% last week, the 10-year yield has jumped back to 5.0%, and the 2-year rose 18 bp in a single session to 4.75% as markets priced more hikes. The reason is the diesel problem described below: expensive refined products are keeping inflation sticky, and central banks on both sides of the Atlantic are leaning hawkish. 5% on the 10-year is now the key level for the whole letter. Holding below it keeps our medium-term view intact; a sustained break above it would pressure long-duration assets, AI stocks and gold alike.

October CPI: slightly hot, but no new acceleration

Today’s CPI is a reminder that disinflation is a forward-looking story. Prices came in slightly above expectations: headline up 0.4% on the month against 0.3% expected, core up 0.3% against 0.2%. But headline slowed from September’s 0.6%, and core matched the 0.3% pace of August and September: firm, not accelerating. Energy rose 0.7%, the component we expect to reverse as November’s oil drop feeds through. Year-on-year rates were not published, because data collection lapsed during last year’s government shutdown.

The Fed: a December hike is fully priced

Markets price a 97% probability that the Fed raises rates on 9 December to 4.00–4.25%, up from 77% a week ago and 70% a month ago. That creates an asymmetry. Delivering a fully priced hike is not a new shock; the surprise can only run one way, and if the data give the Fed room to hold, pricing it out would be a strong tailwind for stocks and bonds. Today’s CPI did not provide that room, so the numbers to watch now are the odds of further hikes in January and March. For gold, the softness around the release has given us a better entry for Trade 01.

Oil: the problem isn’t crude anymore, it’s refined products

The oil story has changed. In the spring the problem was crude: a war premium priced on the risk of losing Hormuz. Today crude is flowing again. Brent has fallen nearly 25% in ten days to about $90 (Bloomberg). Middle East crude exports excluding Iran are back above 90% of pre-war levels, moving through Hormuz and the bypass pipelines and kept moving with workarounds such as ship-to-ship transfers, despite attacks on tankers. Many physical grades now trade at deep discounts to futures, and for the first time since the war began, global inventories are no longer drawing every week. Talks remain live, and a US–Iran deal could return up to about 2 million barrels a day of Iranian exports.

The futures curve confirms it. The Brent January–June 2027 spread collapsed from $11.71 to $6.58 in a week, the WTI front spread is back to about $1 from above $2, and Dated Brent’s premium over futures fell from $36.80 a month ago to $3.43. Buyers have stopped paying up for prompt crude. Freight is easing too: the Breakwave tanker shipping ETF, which tracks freight futures, has fallen roughly 45% from its October peak, and rates west of Suez are falling sharply, although a supertanker from the Gulf to China still costs about $1.4 million a day, against under $100,000 before the war. As Bloomberg’s Javier Blas argues, $90 is not the market pricing peace; it is the market pricing no immediate third round of fighting.

The bottleneck is now refined products. Gulf exports of diesel and jet fuel are running at only about 60% of pre-war levels, held back by a shortage of product tankers, and Ukrainian strikes on Russian refineries account for about a third of the rise in diesel prices. The diesel crack is around $84 a barrel and rose on the week while crude fell. US refineries are running at 94.5%, the highest for the week since 1997, and still cannot rebuild stocks: US distillate inventories are the lowest for the time of year since records began in 1982, and the SPR is about 35% below its five-year average.

That matters for inflation more than the crude price does. Expensive diesel feeds directly into freight, food and, over time, wages, which keeps inflation sticky even as crude falls. It is the main reason the Fed, the Bank of England and the ECB are all signalling more hikes: Fed Governor Waller has said a 50 bp hike in December may be needed, citing diesel. The oil shock’s transmission channel has moved from oil prices to interest rates.

For crude itself, the outlook is two-sided. About $90 is wrong either way: too low if fighting resumes, too high if a deal is signed, which would point towards $70. Around 1 billion barrels of inventory have been drained and will need to be rebuilt, but China has historically restocked only with Brent below about $70, so that support sits well below today’s price. Our base case is a slow drift lower as discounted physical barrels weigh on futures, which favours options over directional positions in crude. Product cracks should stay wide, which benefits refiners outside the Gulf that buy cheaper crude and sell expensive diesel.

The catalyst calendar

The single most important event ahead is NVIDIA’s earnings on 17 November. It is the largest weight in the index and the centre of the AI trade: a sign of slowing growth would hit the whole market, while a clean report would remove the biggest overhang left from last week.

Highlighted: the events that matter most for our positions. Times CET. US markets are closed for Thanksgiving on 26 November, with a half-day on the 27th.

Our view: constructive, with patience built in

We are constructive on a one-to-three-month view, and we expect noise along the way. CTAs would sell again if the market turns lower, NVIDIA’s earnings are a binary event and credit spreads have not yet confirmed. That is why the trades below are option structures: they cap the loss upfront and give the mechanical flows time to clear.

From the trading desk: this week’s positions

Our trading desk turns the research above into three defined-risk positions and one hedge. Each comes with its structure, cost, payoff and the conditions that would make us wrong.

Status of open positions

This is our first issue, so there are no open positions from previous letters.