The Iran ceasefire wiped out oil's war premium.
Once the truce broke, it came right back.
Oil for delivery now costs far more than oil six months out. Traders are paying up for barrels today and betting the squeeze eases later.
Desk notes, October 2026
The desk notes of Sophos Alpha, an independent quantitative research desk. Daily market notes across many topics, including CTAs, risk parity, vol control and other systematic funds, hedge fund activity and investor sentiment, dealer gamma and liquidity, volatility and tail risk, in equities, rates, currencies and commodities.
October 2026
157 notes
5 notes
Since 1989, oil for delivery now has only cost this much more than oil six months out during wars and supply shocks.
Every red stretch on the chart is one. Today sits level with the Iraq war and Russia's invasion of Ukraine.
The market is pricing in a major supply disruption, and systematic hedge funds have been positioned to capture it for months.
We warned you on October 14: Tech had no room left to the top of its channel and a long way down to the bottom.
Tech has underperformed sharply since then.
Investor positioning in Tech has also dropped sharply and is now in line with a significant deceleration in earnings growth.
The risk-reward begins to shift in favor of Tech again.
Brent six months out just hit its highest price since 2022.
The front month is back near its Iran war high, and the later contracts keep climbing. The market is pricing expensive oil well into next year.
Friday was the second worst day for S&P 500 futures since the Iran strikes, and oil barely explains it.
Since the strikes, a 2% crude jump has pointed to stocks down about 0.1%. Futures fell 1.8% with crude up 2%, one of the two biggest misses of the war.
4 notes
Trend following makes its money when equity sell offs persist.
That is what makes it a hedge.
A basic trend strategy was up in 9 of the 10 worst months for global equities since 1992.
Financial crisis: equities down 20%, trend up 12.8%
Covid crash: equities down 14.4%, trend up 6.6%
Russia default: equities down 14.2%, trend up 3.3%
Dot com bust: equities down 10.7%, trend up 4.0%
Those are single months. Over the whole dot com bear market, equities fell 47% and trend made 21%. The rest of the time it barely tracks stocks.
CTAs run far more refined versions of the same idea. That is why they belong next to an equity book.
The formula behind this strategy will be published in our next research piece.
Short government bonds.
Long energy.
That is the systematic hedge fund trade of 2026, and we’ve been flagging it since August, before it paid.
It is why CTAs are the best-performing hedge fund strategy of the year: up 16.6% through Q3, more than double the average fund and ahead of the S&P 500.
Update: vol of vol dipped early this week and is jumping again today, back at the edge of its Iran war range.
VIX is rising with it, but it is still below the bottom of its own war range. The gap is still there.
If your portfolio manager is still counting on the standard 60/40, you're suffering.
This year Treasuries lost money while energy doubled. In our latest research a few days ago, we made the case for "Hedging Portfolios With Commodities".
In that piece, we shared positioning intelligence on systematic trend following strategies, and the distinction matters as our models have been long oil since July and short gold and metals since September.
10 notes
The bond market stopped pushing rates higher once the Fed hiked.
The 2 year yield ran up into September’s hike, stalled, and slipped again after yesterday’s hold.
That is how 1997 and 2015 looked, when the Fed hiked once and stopped. In 1999 the 2 year kept climbing for a year.
The S&P 500 is priced about 14% above where real rates say it should be.
Valuations have peaked with real rates around 1% to 1.5%, eased slowly and fallen fast past 4%. At 2.3% today, we are just past the peak. Strong earnings growth is the likely reason for the premium.
Small Fed hiking cycles have ended three very different ways for stocks.
1997: inflation cooled, one hike, stocks ripped.
2015: growth scare, one hike, a long grind higher.
1999: hikes into a boom, stocks stalled, then the bubble burst.
Yesterday’s hold keeps all three in play.
The Fed paused yesterday.
The futures market says it is a pause, not the end.
A December hike is close to fully priced, a second by March and a third by July, which would take the funds rate to 4.50% to 4.75%. Only after that does the curve go flat.
Pensions and target date funds are set to buy $16bn of equities at month end, and none of it is US stocks.
International equities take $19.7bn of the demand. US equities see $3.7bn of selling.
The month end bid is international this time. The S&P 500 gets no help from it.
The Fed’s first hike has been a speed bump for stocks, not a wall.
In the last nine hiking cycles, the S&P 500 dipped in the first few months and was higher a year later in eight of them. The smallest cycles did best.
After yesterday’s hold, 2026 hasn’t even hit the bump.
US inflation just broke its summer cooldown, and energy is only part of the story.
Energy drove most of September’s jump, but core prices firmed too, and core services are still the biggest single driver. That makes it hard to dismiss this as a purely external shock.
The US economy is slowing, just not enough to give the Fed a reason to ease.
The Atlanta Fed’s GDP nowcast fell from around 5% in late September to about 3.3%. The core PCE nowcast crept up to 3.15%, well above the 2% target.
Q3 GDP and September PCE are out today.
Hiring has nearly stalled, but unemployment is barely moving.
Payrolls added just 29k in September and job openings fell again. Unemployment is 4.2%, inside its range of the past two years. Cooling, not cracking.
Next tests: job openings on 3 November, payrolls on 6 November.
Gold dealers are near their most net short on this chart.
That tells you how crowded the long side is.
Every futures contract has a buyer and a seller. Dealers are net short 40.4% of open interest and have sat below their normal range in 19 of the last 21 weeks.
8 notes
Every Fed hiking cycle since 1994 opened with an S&P 500 stumble.
Every one that kept going ended higher than that low.
The typical fall was around 10%. This cycle's dip so far is a fraction of that.
The S&P 500 tends to stumble once the Fed starts hiking, then recover.
In the last six cycles it fell around 10% from its peak, with the low coming within a few months of the first hike. 2022 went much deeper.
Six weeks into this cycle, the stumble hasn't shown up, yet.
Gold has broken its old link to real yields.
For nearly two decades, higher real yields meant lower gold. On that old relationship, today's real yield would put gold near $500. It's above $4,000.
Since 2022, gold has climbed no matter where real yields went
History says the hikes being priced are the kind the S&P 500 can live with.
Since 1990, years when the Fed hiked by a point or less still saw stocks gain double digits on average. It's the big hiking years that hurt.
Oil vol now rises with oil.
Since the strikes on Iran in late February, crude and its implied vol have moved together, a link that was absent before. Front month vol is still under half its March peak.
The S&P 500 shrugged off the hike pricing.
Small caps haven't.
The Russell 2000 has lost more than 10% against the S&P 500 since June, sliding back toward its 2025 lows as Fed hike pricing kept building.
The market is pricing Fed hikes again, and stocks have rallied right through it.
Hike pricing turned back on in late March, the first time since 2023. The last big build, in 2022, came with a rough year for stocks.
The average S&P 500 year since 1928 puts in its final dip on 27 October.
Midterm years run the same script harder: flat into late October, then the strongest finish. 2026 is a midterm year and well ahead of the script.
7 notes
Since 2021 the dollar (DXY) fell on 12 of the 14 FOMC days with a Fed hike priced.
Tomorrow only 3bps is priced, leaving room for a hawkish surprise.
The stock market tends to stumble when a new era of leadership at the Fed begins.
The S&P 500 traded sideways after Warsh took over in late May, through early August, before rebounding. Bernanke and Powell had choppy first years too.
Tomorrow's FOMC is Warsh's next test.
Quantitative investing meets multi agent AI: Alpha factors built for every market regime.
Most factor generators train a single model, so the factors it finds end up looking alike.
Today's pick from our upcoming open access library trains five AI agents together instead, each specializing in one regime: calm bull, volatile bull, sideways, slow bear and tail risk. A diversity penalty stops them converging on the same factor.
In backtests on the CSI 300 and S&P 500, it beat every model it was compared against.
Get notified when the library is live:
Nasdaq non dealer net positioning just hit a record $55.5bn, and they got there buying into a falling market.
Non dealers bought $9.5bn in the week to October 20 while the index fell 3.5%. Third straight week of buying, $21.8bn since September 29.
Meta's earnings sent the Nasdaq down 2.9% on October 22 and futures open interest jumped $10.6bn, the largest one day rise this year outside roll periods. Shorts were put on.
From October 22 to 26 the Nasdaq recouped 1.1% and open interest gave back $9bn. Shorts covered, and with funding cheapening, longs cut too. CTAs were likely a key driver after a momentum break.
Length is still elevated into Microsoft and Alphabet results on Wednesday and Amazon on Thursday. If they fail to impress, meaningful futures selling could follow.
The AI selloff is being treated as a growth scare, not an inflation one.
Last Thursday the Nasdaq fell almost 3% and the 10y yield had its biggest one day drop since the 2023 banking scare.
In the spring selloff, yields rose as tech fell. This time they are falling together.
The Nasdaq is swinging much harder than the S&P 500 again.
The volatility gap between them, both what options price and what the market delivers, has roughly doubled since September.
As long as AI drives this market, the Nasdaq moves further than the S&P in both directions.
Brazil's stock market just had an 11.4 sigma day, its biggest shock since at least 1998.
The Ibovespa, Brazil's benchmark stock index, fell 13.9% on Monday after Lula's reelection. It was down as much as 17.3% intraday and tripped the circuit breaker for the first time since March 2020.
The Asian crisis, Russia's default and Covid all brought drops as deep or deeper, but in markets already in panic. Measured against how calm the market had been, Monday was the bigger shock. Covid's worst was 9.5 sigma.
12 notes
The average S&P 500 stock is now as far behind the index as at the height of the dot com bubble.
Equal weight slid through the AI boom just as it did into 2000. After that bubble burst, it beat the index by nearly 50% over five years.
The better opportunities now sit outside the megacaps, which favours portfolios spread across sectors, regions and company sizes in the years ahead.
Systematic managers lost 1.4% in a single day, a 3.7 sigma move and their worst since August 19.
Last Thursday, all three major hedge fund groups lost more than 1% on the same day, a first since COVID.
Bonds are now more volatile than stocks.
Long dated Treasuries have swung more than the S&P 500 over the past 30 days, by the widest margin in over a year.
This AI sell off is tracking July's almost step for step.
Ten trading days in, semis are down 14% and the S&P 500 just 2.4%. In July, both bottomed right at day 10, and semis were back to flat ten days later.
Friday was day 10. The next two weeks decide whether the pattern holds.
The S&P 500 has closed down 1% or more 21 times this year.
That is exactly the median for this point in the year since 1928, and the average full year brings 29.
Down days like these are the price of admission for stock market returns.
For the first time in history, Apple is moving against the Nasdaq 100.
Their 3 month correlation turned negative in August and is still below zero. In four decades it had never even come close.
The stock that helped build the index is now trading on its own.
Our Vol Stress Index jumped from 4.49 on Wednesday to 7.80 on Friday, the 81st percentile of the past year.
Two weeks ago it sat at 1.68.
That is a sharp change in hedging behavior for an S&P 500 sitting just 2.5% below its record close.
The Fed, GDP, PCE and three trillion dollar tech giants, all inside 48 hours.
Options are pricing Wednesday and Thursday as the two biggest SPX sessions, at 1.03% and 1.05%.
FOMC lands Wednesday with Microsoft and Alphabet earnings. GDP, PCE and Amazon follow Thursday.
November SPX options are priced for a big move either way.
Crash protection is priced far richer than options at the market, and calls well above it are priced up too.
The Fed and big tech earnings both land before November expiry. The market is pricing both tails.
The VIX is having a perfectly ordinary year.
Its 2026 average sits right around the typical year since 1999, and nowhere near the crisis years of 2008, 2009 and 2020.
At its March 30 low, the S&P 500 was down 7% on the year, its fifth worst start since 1989.
As of Friday it is up 12%, well ahead of the 8% an average year had gained by this date.
Last year's tariff low ran deeper, down 15% in April, and was up 16% by this point.
The average S&P 500 stock is beating the index by the widest margin since April 2009.
Leadership is shifting from the biggest names to the rest of the market.
2 notes
Gold futures fell over 2% on October 22 while the 10Y yield dropped almost 20bp, their worst day on a big bond rally since early 2020.
Over 20 years, gold rose on about 7 in 10 of those big yield drops. This time it traded with the Nasdaq 100, down nearly 3%, not with bonds.
The S&P 500 fell over 1% on October 22 while the average stock rose.
Equal weight beat the index by about 2 points, a gap seen only a handful of times since 2006, mostly in 2009 and 2020.
In October equal weight is up 4%, the S&P 500 roughly flat.
3 notes
New academic research in quantitative finance, curated every single day as it becomes available in our upcoming open access library.
Today's pick from the library asks how much history a return forecaster should use. Older observations may help reveal a weak predictive relationship, but they may also point to a relationship that no longer holds.
In simulations, old data pointing the wrong way made all five AI models tested worse. But cutting the history made every model less accurate on 56 real assets.
Get notified when the library is live: https://sophosalpha.substack.com/subscribe
Vol of vol is flashing a warning that VIX hasn't confirmed yet.
It has climbed back into the range it held during the Iran war, while VIX is still below the bottom of its own war range.
Measured against the VIX level, vol of vol sits in the 78th percentile of the last 3 years.
The options market is bracing before the index has moved.
We warned that asset managers were holding a record short in VIX futures, and that extremes like this have coincided with strong VIX forward returns.
Since then VIX dropped into Wednesday's expiry and bounced right back by Friday.
Vol of vol is back near its highest since early April, in the 85th percentile on a 3 year lookback.
The latest data shows asset managers trimmed that short, but it is still one of the largest on record.
Not an extreme yet, but there is a lot of short left to cover if anything breaks.
7 notes
Books used to be where you learned the latest in machine learning.
In reinforcement learning and LLMs, a book is out of date the month it comes out.
To keep up with these fields, read the research papers. That is where new work shows up first.
Today's pick in our free index of 10,000+ papers asks what an AI portfolio should do with old news.
On ten US stocks over 2023 to 2025, an LLM that simply ignored SEC filings older than eight weeks earned 18.43% a year after costs, with the news arriving four weeks late.
That beat every reinforcement learning model that used news. Keeping old news and telling the models its age did no better.
Get notified when our free library is live: https://sophosalpha.substack.com/subscribe
Retail and hedge funds just took opposite sides of the AI trade.
Hedge funds sold tech at the fastest pace in a decade and cut risk to near three year lows.
Retail bought the dip: tech ETFs had their biggest inflow day on record, and buying spread to chip names beyond the Mag 7.
Brent just hit a five month high.
A few days ago we shared that trend followers have been well positioned to capture this move for weeks.
One of our in house trend models has been in these longs since July and is well into profit on the energy front.
The S&P 500 fell 1.21% yesterday, yet 67% of its stocks closed higher.
The average member actually gained 0.74%.
Six of the Magnificent 7 fell, led by META down 7%, and their weight was enough to drag the whole index into the red.
Over 150 S&P 500 stocks made a 2 sigma move yesterday, the most since the May 2025 US China tariff truce.
The index closed red, yet two thirds of its stocks finished higher and most of the big moves were to the upside.
Just under 60% of S&P 500 stocks are 10% off their highs while the index sits within 2% of a record.
That is not a first.
On 30 Sep it was about 80%, with the index just as close to its record. The dot com bubble ran higher still.
The Nasdaq 100 fell 2.9% yesterday, its biggest drop since June and its 4th move of 1% or more in five sessions, the first time since early August.
Two of those big days were gains of about 1%. Two were losses of 2.5% or more. The selling days are the heavier ones.
4 notes
10,000+ quant finance papers in one FREE index organised in 12 categories and 74 topics.
Every day we add new research as it comes out and pick 3 papers worth your time.
Today: what an AI trained on order book data actually learns, American option pricing beyond Heston, and the volatility at which providing liquidity on DeFi exchanges stops paying.
No login, no paywall.
Get notified when it’s live: https://sophosalpha.substack.com/subscribe
Hedge funds aggressively net sold global Info Tech stocks at the fastest pace in the past decade (-3.3 SDs) over the past two weeks, driven by long-and-short sales.
Net allocation to Global AI stocks now sits in the 38th percentile while aggregate long/short ratio is in the 6 th percentile vs. the past year.
Hedge funds have heavily net sold global equities so far in October, driven by short sales far outpacing long buys.
Overall Long leverage now sits in the 73 rd percentile vs. the past year (94 th percentile 5-year), while Short leverage has risen sharply and is approaching 5-year highs in the 99 th percentile.
A narrow stock rally props up the dollar.
When equal weight beats the S&P 500, the dollar weakens on top of what stocks and rates do. NOK, AUD and NZD react the most. The yen barely reacts and trades on US 2 year yields.
A broader rally takes that support away.
6 notes
Catastrophic losses are the hardest to train an AI against: only a small number of observations ever reach the far tail.
An AI trained to protect against rare disasters held up better in them when it estimated how bad things can get, rather than learning only from the few disasters it had seen.
Today's pick trains a neural network hedger to model that tail with extreme value theory (EVT), tested on simulated markets with rare price jumps.
On 100,000 normal paths, EVT and the standard empirical approach hedge almost identically.
On 3,000 paths from the most extreme 0.1% of outcomes, EVT cuts the average hedging error from 17.49 to 10.03 and the worst from 114.27 to 84.36.
Not better everywhere. Better where it counts.
We choose 3 new papers a day in our upcoming free index of 10,000+ papers. Get notified when it’s live: https://sophosalpha.substack.com/subscribe
Trend followers are back in the red zone on the inflation trade.
They have been this long oil and this short government bonds only twice in 20 years: the 2022 inflation shock, and now.
Debt worries and sticky inflation are pushing bonds lower, and the unresolved US Iran war is keeping oil supported. If that conflict escalates, this trade has more room to run.
Your bonds only protect you from one kind of tech selloff.
This year most bad Nasdaq days came with rising yields, so bonds fell too.
The AI scares broke that. On DeepSeek and the 8 October OpenAI revenue scare, yields fell with tech, and DeepSeek’s drop beat almost every bad day of the dotcom unwind.
Will my bonds hold up when my tech stocks fall?
For two decades the answer was mostly yes. When tech sold off, Treasury yields usually fell with it, so bond prices rose and cushioned the drop. The dotcom unwind played out that way.
That broke in 2021. In 2022 the Nasdaq 100 fell by a third while yields climbed, and investors lost on both sides. In June the link hit its weakest since 1994.
It’s been climbing back since, and in the recent AI scares, DeepSeek and 8 October, tech and yields fell together again.
If the AI trade ever unwinds like dotcom, this is the chart to watch.
The S&P 500 has gone nowhere this month, but breadth underneath is starting to improve.
To start October all three gauges were oversold at once, like April 2025 and March 2026. All three are back in range.
With volatility rising too, it looks more like rotation than selling.
Bonds are pricing a lot more risk than stocks.
Treasury volatility is up about a third since late July. US stock volatility is below where it started.
7 notes
Researchers go incredibly deep on markets, even Reddit.
WallStreetBets has been tied to extreme volatility since GameStop. Today's pick from our library finds most of the signal in its posts is just which stock is named.
We choose 3 new papers a day in our upcoming free index of 10,000+ papers. Swap doomscrolling for research by people who study markets.
Get notified when it’s live: https://sophosalpha.substack.com/subscribe
Hedging the stocks everyone just ran to is still cheap.
Staples and health care options price barely above what those sectors actually move, among the cheapest in the S&P 500. Tech options carry a bigger premium.
If the defensive rush reverses, that is the cheap place to hedge.
Money is rushing out of tech and into safe stocks at dotcom bubble extremes.
Last week health care, staples and utilities beat tech by a margin seen only a handful of times since 1998.
The S&P 500 isn't breaking, it's rotating.
Last week's selloff wasn't about megacap tech.
It hit the high beta fringe.
The Mag 8 landed almost exactly where their beta implied. The 8 worst, CRWV, SMCI, DELL and RKLB among them, fell 18% to 26%, about 16 points worse.
A positioning flush, not a crack in the AI core.
Technology was the worst S&P 500 sector over the past week, while Energy, Staples and Materials moved higher.
The market now wants proof that AI spending pays. Meta reports tomorrow, and the biggest spenders are still on track for vast outlays this year.
Trend following funds (CTAs) are still long global equities, but the breadth has thinned fast.
Since August they unwound most of their Europe and Asia longs, and several markets flipped short. US large caps now carry most of it.
If these funds return as buyers, the room to buy back sits outside the US.
US L/S net leverage: fell 3.2pts last week, the largest weekly decline since Liberation Day and 4th-largest over the past five years.
5 notes
The buyback bid is about to return.
About 1 in 8 S&P 500 companies is out of its buyback blackout. With earnings packed into late October, that climbs to nearly all by mid November.
Flows have run light, mostly preset plans. Discretionary buying should follow as windows open.
Hedge fund net leverage fell last week by the most since Liberation Day.
The performance environment has been extremely challenging. Both fundamental and systematic managers lost money last week while global equities went broadly nowhere.
The pressure has been building for weeks. Funds have been adding shorts, selling single stocks at the fastest pace in two months and hiding in defensives. The long/short ratio is back near its Liberation Day low, and they are now reaching for protection through options.
Gross leverage is still elevated, so there is room to cut further. But with net leverage this low, shorts near five year highs and demand for protection accelerating, a lot of caution is already reflected in positioning.
A different volatility regime, but not necessarily a more bearish one.
Friday was the largest SPX put volume session of all time.
Index puts set a record too, and VIX calls were the busiest since the Friday before VIX spiked above 60 in August 2024.
After weeks of damage in single stocks, investors are finally hedging the index.
Dealer gamma took its biggest OPEX day hit in nearly five years on Friday, leaving positioning close to neutral.
Single stocks swung hard all year while the index stayed calm. With less cushion, that calm looks fragile.
Remember, volatility cuts both ways, upside included.
We shared more information in our latest institutional monitor which you can read here: https://sophosalpha.substack.com/p/institutional-monitor-3ad
Our Vol Stress Index, which tracks stress in S&P 500 volatility, just had one of its sharpest weekly jumps on record, the biggest since October 2018.
A week earlier it sat near its calmest level of the year
3 notes
Concerns over AI valuations and financing, plus strong US inflation prints, weighed on stocks last week.
Investors rotated out of AI names into healthcare and staples.
Korea led Asia lower with a double digit fall. Safe haven demand lifted government bonds, and Europe held up.
Higher for longer rates have lifted expected returns on cash and bonds, and they are squeezing equity valuations.
The expected returns point to rotating toward international and emerging market stocks, where valuations are cheaper and investors get paid more for the risk.
QQQ fell five days in a row and the vol market did nothing.
Then on Friday afternoon it flinched all at once.
VVIX, which measures how hard VIX itself is expected to swing, had one of its biggest one day jumps in four years. VIX closed at its highest since July.
one note
5 notes
Funds have room to add equity exposure from here.
Market sentiment slid from stretched in August to washed out into early October, and S&P 500 futures positioning is still running light.
The summer froth is gone and sentiment has reset.
Our Vol Stress Index jumped 3 points on Tuesday, its biggest one day move since July.
It has already eased to 4.1, below its one year median. Real stress reads 9 to 10, last seen in April.
One outlier session, not a stress regime (yet).
Index vol is heading into the midterms well below past cycles.
The vote usually brings a quick vol crush, and S&P 500 vol already sits below where past midterms settled after it.
S&P500 earnings growth for Q3 is almost entirely a tech story.
The headline looks demanding, but strip out two chip names and the bar sits far lower. The typical stock only needs modest growth.
From COVID through 2025, the three big hedge fund strategies never all fell this much on the same day.
It has now happened 3 times in 2026, the latest on Thursday.
Fundamental, systematic and multi strat were all down close to 1%, led by momentum and crowding.
6 notes
The compute used to train AI models grew 1.5x a year for six decades.
Since 2010 it has grown 4.3x a year.
Across 25 orders of magnitude, more compute has meant more capable models. That is a trend, not a law, and the industry is betting trillions it holds.
Since ChatGPT launched, S&P 500 semiconductor stocks are up 6.2x.
Software is up 2x.
AI is sorting firms into adopters and casualties, with little in between.
When outcomes split that hard, owning more names does not average the risk away. It just owns more of it.
Yesterday the calm may have cracked, but zoom out.
The S&P 500 average intraday range over the last 10 sessions is 0.68%, lower than 89% of the past five years and well under the 1.18% five year average.
The S&P 500 fell 1.11% yesterday against a 1.06% move implied by VIX at the close.
Just over the line.
Normal for a 1% down day: 62% of them since 2000 beat the VIX implied move. In 2026 only 28 of 197 sessions did, 14% vs 20% long run. Options have overpriced moves all year.
Brent is back above $100, where it sat in mid 2022.
US CPI was running above 8% then. It is under 4% now.
Same oil, less than half the inflation. That spread is what to watch from here.
Why diesel matters for it, in our post below.
The S&P 500 closed down 1.11% yesterday, ending a 53 session run without a 1% decline.
Only 11 runs that long since 2000.
The streak was the outlier, not the drop: under one standard deviation, a 1.23% intraday range, the 19th 1% decline of 2026.
6 notes
Last week previewed how this market trades an AI slowdown.
With the Nasdaq near record highs, a weak OpenAI revenue report sent it sliding far more than the S&P 500. DeepSeek and the summer chip rout looked the same.
Tech is the biggest weight in the US and EM Asia.
In Europe it trails financials, industrials and health care.
That is why one sector keeps deciding who wins across regions. Tech earnings are still booming, but the worry is now future profitability. The bar elsewhere is lower.
Only 49% of S&P 500 stocks are above their 200 day average, yet the index is 7.5% above its own.
Breadth catches up or the index catches down. Financials, reporting this week, and bond volatility are the tells.
S&P 500 swings are near their quietest of the year, with dealers holding enough gamma to buy dips and sell rallies.
Demand for VIX calls keeps climbing.
Calm like this only lasts so long. One catalyst can thin that cushion fast.
The S&P 500 barely moved over the past week, but the leadership underneath it changed.
Tech is fading while Energy, Staples and the defensives pick up the slack. Tech positioning is starting to slip from overweight, and most other sectors are still light.
The rotation has room.
Tech has no room left to the top of its channel, and a long way down to the bottom.
Against the rest of the S&P 500, it is back at the edge where every recent run has turned, and the turns came as fast as the climbs.
The rotation has run far. The risk reward now points down.
2 notes
The S&P 500 is at all time highs while US fundamental long/short net exposure sits near 47%, the 2nd percentile of the past year.
Their long/short ratio is in the 1st percentile of five years.
Funds are not chasing this rally. That is fuel if earnings deliver.
Equity volatility tends to peak about a month before US midterms, then reverse sharply as stocks rally once the uncertainty clears.
The vote is three weeks away. Markets are likely to digest any electoral outcome and anchor back to fundamentals.
7 notes
Diesel is the fuel of inflation.
Since 1995, every month with US diesel up more than 30% on the year saw US inflation come in above forecast. Diesel is up 68% now. Yet August inflation beat forecast by only 0.6 points, the smallest beat with diesel up 40% since 2000.
Nasdaq 100 implied vol still runs 1.38 times the S&P 500's, higher than 94% of days since 2009.
The ratio is down from 1.73 in July, but the average since 2009 is 1.18. Hedging the Nasdaq 100 still costs more than usual next to the S&P 500.
Mag 7 stocks moved 10.9% on average on their earnings last season, matching early 2022 as the biggest reaction in 43 seasons since 2016.
Microsoft and Amazon each jumped more than 15% in a day and Tesla fell 14.5%. That is the bar the next round of prints walks into.
The last three weeks were the narrowest the S&P 500 has been near a record since the Dot com peak.
It closed Friday 0.1% from its all time high, yet 56% of its stocks sit more than 20% below their own.
Speculators hold the biggest Russell 2000 futures short since August 2025, 88k contracts net.
Upside is cheap against that: a 1 month 5% OTM call costs about 0.64% of spot, cheaper than 88% of days in the last 3 years.
If small caps rally, there is a lot of short to cover.
Investors are selling S&P 500 futures at a pace only seen in 2020 and 2022.
Asset managers and hedge funds did over 80% of the selling.
The difference this time: back then the index was below its highs, now it barely moved and set a new record on 6 Oct.
Single stocks are swinging far harder than the S&P 500 itself.
The average member's realized vol is running 23.1 points above the index's, on both a 3 and 9 month view.
On 9 months that is the widest gap since 2009, and 2026 is on track as the most dispersive year since 2009.
2 notes
Only 29% of S&P 500 companies have beaten the index over the last 3 years.
Since 1998, readings this low came only at the 2000 dot com peak and in the past year.
This rally is being carried by a few names. If they stumble, there is little underneath to catch the index.
Hedge funds rotated out of Info Tech and into Industrials last week, reversing the trend since early September.
Info Tech selling was heaviest Thursday.
Health Care was degrossed for a 4th straight week. Single stock shorts modestly outpaced long buys.
5 notes
The first nine months of a midterm year are usually dead money for stocks.
Not this time.
The average midterm year since 1934 is flat into October. The S&P 500 is up double digits, even after the Iran war selloff in March.
The strong part of a midterm year hasn't started yet.
The S&P 500 is following the midterm year script almost to the letter.
It dipped into September right on schedule, the same weeks the average of the last 23 midterms bottomed, and has already turned higher.
The best part of the script comes after the vote. That's November 3.
Betting against the S&P 500 around a midterm election has been a losing trade for almost a century.
It rose around 21 of the last 23 votes. The only losers: 1978 and 2018.
Markets read the midterms as a check on the White House, not a threat. This one is November 3.
The S&P 500 runs the same play almost every earnings season: rally when companies report, stall for a month, then grind higher.
This time the rally window has company. The Fed meets and America votes inside it.
Markets hedge into events like that, then rally once they pass.
The S&P 500 has been stuck in a box for two months.
Tuesday is when it usually breaks out.
For a year now it has rallied when companies report, then gone nowhere until the next season.
The street expects earnings growth to slow this quarter. The economy isn't showing it.
6 notes
The biggest moves in the S&P 500 this past month have been up, not down.
Up days have swung nearly twice as hard as down days in the S&P 500, and harder in the Nasdaq too. Small caps show the opposite.
The risk that matters in this tape is a tech rally running away from you.
Bonds stopped protecting stocks in 2022.
For two decades they rose when stocks fell. Since the inflation shock, stocks and Treasuries move together again, and the link is now the tightest since the late 1990s.
That is why income money went looking beyond bonds.
A handful of megacaps is carrying the whole S&P 500.
The index is flat on the month, but fewer than 1 in 4 of its stocks are actually up. Normally it is about half.
Going back to the 90s, the market has almost never been this thin underneath.
Those double digit yields are not free.
The income ETFs paying the most swing about twice as hard as core bond funds. The extra income is payment for taking on stock market risk.
More yield has meant a bumpier ride.
A big yield is not a big return.
The income ETFs paying double digit yields have been shrinking since 2019. Spend the payout instead of reinvesting it and the price you hold keeps falling. The funds paying less grew.
Part of that check is your own capital coming back.
Volatility is calm versus policy uncertainty.
The VIX is sitting at calm levels while trade and fiscal risks just pushed the policy uncertainty index to one of its highest readings in three years.
Protection against the headlines is cheap right now.
2 notes
US equity vol has reset structurally higher since COVID, a clear break from the exceptionally low vol regime of the 2010s.
Single stocks lead, with AI driving a record number of outlier moves. FOMO and boom bust dynamics should keep it there, and keep option premiums rich.
US fundamental long short net exposure just fell to its lowest since the April 2025 lows, with the S&P 500 at all time highs.
Managers are underexposed into a rally that keeps grinding higher. Plenty of fuel if they have to chase.
7 notes
Correlation between S&P 500 sectors hit zero in early September, the lowest of the year, while correlation within sectors held its range.
Stocks moved with their sector and sectors went their own way. Dispersion works as long as that holds.
Stocks delivered even less correlation than options priced.
S&P 500 implied correlation is 8%, lower on 2% of days since 2006, and realized fell to 3% in early September, under it.
Q3 earnings season starts next week, and it is back loaded: 44% of S&P 500 market cap reports in the single week of Oct 26 to 30.
Consensus wants 27% EPS growth, led by Tech at 64% and Energy at 109%. With no help expected from multiples, earnings have to carry the tape.
Over $1.5 trillion has poured into US ETFs this year, already past last year's record annual haul with a quarter still to go.
That is about $167 billion a month, $8 billion every session. Hold that pace through a seasonally strong Q4 and 2026 ends above $2 trillion.
93% of S&P 500 companies are now in their buyback blackout window, 94% by the end of the week.
The market's biggest steady buyer is mostly sidelined until Q3 earnings start in mid October. Authorizations run 11% ahead of last year, so the bid returns once windows reopen.
Dealer gamma is the market's shock absorber, holding even with the 10Y at its highest since 2002.
The catch: the flip sits just below spot. A modest dip turns dealers from buying weakness to selling it, right where CTAs start trimming.
More in this week's Institutional Monitor: https://sophosalpha.substack.com/p/institutional-monitor-d84
We flagged on 17 September that, on the historical odds, the 10Y was more likely than not to rise over the year after the Fed's first hike.
19 days in, it is up 30bp from the first hike close, already running above the average cycle.
7 notes
The S&P 500 is correlated with almost nothing but itself.
9 of 10 sectors are tracking it less than in 95% of their history.
Hedge funds rotated into tech, media and telecom in September and sold nearly every other sector. When nothing else works, the AI trade adds passengers.
The S&P 500 is ~1% off its record high, yet net new 52 week highs in the index sank to negative 23 (5 day average) in late September, the lowest since April 2025.
Fundamental long/short net leverage is at its lowest since April 2025 too. A divided market with little net length.
The S&P 500's 3 month correlation with crude oil hit negative 0.61 in late September, a level last seen in October 1990.
In April 2025 it was positive 0.68. Oil is telling a different story than stocks, and the S&P 500 is trading like an asset class of its own.
The S&P 500 and its equal weight version are drifting apart.
Their 3 month correlation fell to 0.57 in September. Before 2024 it never dropped below 0.79, back to 2007.
The top 10 contributors are expected to deliver over two thirds of S&P 500 earnings growth this season.
Systematic L/S, a proxy for quant performance, rose an estimated 3.46% in September.
That is its best month of the year, and it came entirely from short side alpha.
It is now close to 20% YTD, well ahead of fundamental L/S and MSCI TR. It also stayed positive through the March and April drawdown that pushed both lower.
This is why we track quant flows so closely. Our latest Institutional Monitor is out later covering systematic funds and what’s their position like right now.
Subscribe to get it, free:
The AI trade has real revenue under it.
Cloud revenue at Amazon, Google and Microsoft is running at nearly double its pre boom average.
That backs the Nasdaq's lead over the S&P 500 this year, and the case that cash flow risk behind the capex is low.
Facing Chinese open source, US AI labs are now cutting prices.
Meta's and OpenAI's new cheap models even undercut some of the Chinese ones.
Early sign: falling costs are driving more AI spending. AI earnings can keep growing even if margins come under pressure.
3 notes
Sentiment is starting to show signs of bottoming, as investors add length ahead of earnings.
Our Market Sentiment Indicator has been washed out since late September, its gloomiest since June, while the S&P 500 sits barely off its highs.
Hedge funds and institutional money managers just dumped S&P 500 futures at the fastest pace since 2022.
They sold almost every week since late August, yet the index is barely off its highs. That leaves plenty of room to buy back if earnings deliver.
The bond and credit markets are nervous.
The stock market isn't.
Credit volatility just jumped and confidence in Fed policy is near record lows, yet the Vol Stress Index sits near the calm end of its range. If the stress spills into stocks, hedges are still cheap.
4 notes
The 10 year yield closed at 5.29% on 30 September, its highest since May 2002.
TLT hit a record low this week.
Bond ETFs now take about 27% of all ETF trading volume, their highest share since at least 2021.
When rates move this fast, the bond trade runs through ETFs.
Nvidia alone weighs more in the S&P 500 than the smallest 254 companies combined.
Of every $1 in the index, 35¢ goes to the Mag 7 and 41¢ to the top 10. The 400 smallest companies share 23.5¢.
Diversified by name, concentrated by dollar.
Four stocks carried the S&P 500 in Q3.
MSFT, NVDA, AAPL and META added 315 points, more than twice the index's entire gain. The other 496 took away 164, and 313 of the 500 fell.
The index rose 2%. The typical stock fell more than 4%.
NFP hits at 8:30am ET today.
Options price a 0.73% SPX move on the day, about 56 points either way.
Next week: Trade Balance Tuesday (0.65%), FOMC Minutes Wednesday (0.69%) and Jobless Claims Thursday (0.70%).
11 notes
The midterm story has historically improved around election day.
Since 1950 the S&P 500 has weakened into the vote, bottoming on average about 23 trading days before it, roughly mid October. 240 trading days after election day it stood 15.6% higher, almost double the 8.4% for all years.
2026 sits right where that pattern has typically started to improve.
The seasonal improvement in equities is not shared across asset classes.
Since 1984, October has been a difficult month for commodities, down 0.44% on average, and especially crude oil, down 2.23%. Gold has also struggled, down 0.34%.
Treasury futures have posted a modest +0.20%, while the dollar has been broadly flat at +0.08%.
Treasury added $168mn per bp of duration in September net of buybacks, down from $188mn in August.
After coupons, ageing and index extension, net supply was +$48mn.
The 30 year auction on 10 September was the biggest supply day, alongside the first enlarged long end buyback: $5.2bn vs $2bn in August.
The supply impact is small. The signal is that Treasury is watching long end yields.
The S&P 500 is calm.
Its stocks are not.
Index implied vol sits at 16.0 while the average S&P 500 stock carries 39.7, a 23.7 point gap wider than 93% of sessions since 2021.
Big single stock moves are cancelling each other out at the index level. That calm only holds while stocks keep moving in different directions.
Midterm years add an important layer to October's already elevated volatility.
Since 1950 the S&P 500's peak to trough drawdown in midterm years has averaged 17.5%, against 12.4% in other years. 12 of the 19 midterm lows came in the third quarter or in October.
In midterm years, October has frequently coincided with the latter stages of significant corrections. 2026's drawdown so far: 9.1%.
Bond volatility and equity volatility usually move together.
Right now they don't.
The MOVE is up 47% in September to 110, while the VIX sits at 16 with the S&P 500 within 2% of its record. MOVE only spikes come when the price of money itself is in question. When both spike, as with Covid and Liberation Day, trouble is more serious.
US Treasury net duration supply has almost halved, yet the 10 year yield keeps climbing.
Net supply over the past 12 months is down to $485mn per bp from $949mn in November 2024, while the 10 year has gone from 4.45% to 5.16%.
Supply is set to keep falling as reinvestment demand from ageing, index extension and coupons rises, but it is still above anything seen from 2015 to 2019.
Rising yields are hitting stocks again.
The 6 month correlation between the S&P 500 and the 10 year yield hit negative 0.54 this month, the lowest in at least five years. The 3 month sits at negative 0.44.
Yet 6 month implied equity/rates correlation is still priced near zero, leaving the stocks down, yields up outcome cheap as a hedge for sticky inflation, an energy shock or fiscal concerns.
Utilities just closed their worst quarter since the pandemic.
XLU fell 13.0% in Q3, the biggest quarterly drop since Q1 2020, after falling in each of the last three months, with September the worst at 6.6%.
Unfavorable legislation and the backup in rates weighed heavily. The 14 day RSI hit 22 last week, the most oversold since October 2023.
Gold enters one of its weakest months of the year.
Since 1984, gold has averaged a 0.3% decline in October. 2026 has been anything but typical: up 25.5% at the end of January, down 7.3% by mid July, down 3.4% now.
With the 10 year yield closing above 5% for the first time since 2007 and the dollar firming, gold faces an unusually challenging rates backdrop.
October marks the start of crude oil's weakest seasonal stretch.
Since 1984, oil has averaged a 2.2% decline in October, followed by an even weaker November at 2.8%.
In 2026 the seasonal signal has been overshadowed by the conflict between the US and Iran. Oil is up 68% this year, driven by the geopolitical risk premium rather than the calendar.
