Desk notes, September 2026

Don't put too much weight on any one report.

Last week, Goldman's prime book had hedge funds selling US equities. Morgan Stanley's had them net buyers, led by Tech, with US L/S net leverage back up to 54% from 50%.

Health Care is the clearest gap: sold 6 weeks running at Goldman, large buying at Morgan Stanley, driven by short covering in high short interest biotech.

2026 is already ahead of the pattern: up 0.9% since June 1, while the average midterm year is down 1.6% here.

The strongest stretch of the cycle starts in 5 weeks.

Real US investment in tech equipment and software is up 43% since late 2022. At the same point in the 1990s it was up 95%, and it nearly tripled by the peak.

Room left to grow.

The Nasdaq rallied over 3% anyway.

Agentic AI optimism carried it through last week, one more sign that macro faces a high bar to disrupt the AI trade.

IWM and XLY are two of the most rate sensitive ETFs on the board, yet their puts cost less than that sensitivity implies.

Other speculators provided the offset, helping stabilize price in the face of resilient real rates and a stronger dollar.

Then macro forces became too formidable. Longs liquidated. Option market dynamics point to hedging and/or outright shorting as well.

The presence of new downside positions and existing CTA shorts makes the tactical positioning debate less one sided. CTA net length already sits in negative territory, which limits how much selling more negative momentum can bring.

The VIX has typically peaked roughly one month before election day, then reset lower after the vote.

US equity volatility has tended to rise into the midterms, often trading above typical seasonality.

Since the end of July the Mag 7 is up 8.5% and the Nasdaq 100 7.1%. The average S&P 500 stock is down 2.3%, small caps 5.5%.

Yet the Mag 7 is still the weakest here since last October. Room to catch up if the rotation holds.

This selling is forced. After a strong quarter for stocks, they have to trim back to their target mix, whatever their market view.

Only 125 sessions since 1978 were worse.

That was the 7th drop this big in 2026, the most in any year since 2008. And 44 sessions landed outside the channel over the last 12 months.

Gold hasn't traded this wild since 2009.

They added $15.8bn of Nasdaq futures shorts in the week to September 22, the day the index closed at a record, right after weeks of covering.

If the rally keeps going, those shorts become fuel.

Big moves in single names cancel each other out, so the index looks calm while stocks whip around below. It's why so many stock pickers trail the benchmark this year.

AI strength has kept the index near its highs, but the typical stock inside it now trades about 16% below its own 52 week high. The index hasn't run this far ahead of the median stock since 2000.

That gap is also the setup. With so many stocks left behind, there is room for broad market upside and a catch up in the laggards, as long as macro uncertainty doesn't accelerate from here.

Over 70% of members trade below their 50 day average, yet the S&P 500 is still up over that stretch. A handful of giants carry everything. Under the surface this market is weaker than the headline.

Its average up day is now bigger than its average down day, a full flip from last spring when selloffs did the damage. Investors are still underweight, so when the market rallies, they chase.

Earlier this year its swings hit the most extreme level in at least five years, moving more like a meme stock than a metal. It has since cooled all the way back to its normal range.

Far out of the money NDX calls are priced near their richest in 3 years. Last week also brought the 3rd largest 1 month put call skew drop on record and the first record close since June.

Over the last six months it moved against 16 of the other 18 sectors and against the STOXX 600 itself. Since 2008 it had never moved against more than 5.

As long as oil drives this market, Energy offsets the rest.

The stock market isn't.

Treasury volatility sits near its highest reading of the past year while the VIX sits near its lowest. Bonds are pricing trouble that stocks are shrugging off. That gap is the one to watch into year end.

It still sits 30bp above the line it held against the lowest rate futures price for the Fed from 2023 until April 2025.

The index moved 0.9%.

That gap ranks 94th percentile since 1996. It has held above 5 points on 58% of sessions this year, versus 14% over the prior 30 years.

Stocks are cancelling each other out. The index looks calm.

September is tilting the other way. Long buying is outpacing new shorts again, with Europe leading the buying while Asia is still being sold.

With net exposure still light, there is room for more.

Four weeks on, not one of those shorts has flipped, and all 19 are in profit since.

https://substack.com/@sophosalpha/note/c-325437995

Options price a 0.57% move in SPX today. Every session next week is priced higher, with GDP and PCE on Wednesday and NFP on Friday carrying the biggest move of the week at 0.82%.

The US 10Y is up 93bp this year and only 10bp of that is breakevens. The rest is real yields, now the highest since 2008, while breakevens have barely moved in three years.

This is growth and a higher neutral rate being priced.

On the historical odds, yields were more likely than not to rise over the coming year.

The 10Y is up 16bp since, already at the top of the range of every tightening cycle since 1963 at this stage.

https://substack.com/@sophosalpha/note/c-339471905

Net exposure sits in the lower half of the past year. Lots of bets on both sides, little conviction on direction.

Room to chase if this rally keeps going.

The index is up, but 9 of its 11 sectors fell.

Tech and communication services carried the whole move while everything from financials to utilities slid.

It is a double move: equities grinding higher while the bond selloff pushes the 10Y to its highest since 2007.

That has been the case for most of September, and the gap just hit its widest in a year. A narrow group is holding the index up while the average stock slips.

This is a market digesting an oil and rates shock, not repricing a collapse in earnings.

The S&P 500 is 1.2% below its August record close. The Euro Stoxx 50 is 3.8% below its own.

Brent's break above $100 pushed yields up and weighed on stocks, but the pullback is modest. Corporates are strong on profitability and balance sheets across the US, Europe and Japan.

Oil fell 10% in a week, yet our model of trend following funds (CTAs) is still long every energy market. Crude is up about 20% since summer and gasoline more than doubled since January.

We update this book every week. Subscribe to follow it: https://sophosalpha.substack.com/subscribe

The gap between them has rarely been this wide since 2010, and most of those times came this year.

Normally they rise and fall together. Since the March strikes, when crude jumps, stocks drop, and when crude eases, stocks bounce.

That is what a supply shock looks like. Until it fades, crude is the number to watch.

On real yields, 4.2%.

Today that is a move of about 50bp. The 10 year is up 36bp so far this month.

Less than 20bp more by month end puts September in that bucket.

35% of its stocks have a negative 3 month beta, down from 42% in early August. In 2001 it never got above 20%.

When the index rallies, a third of its stocks tend to go the other way.

Sixty days out, it did in 5 of the last 6. 2026 has held to that pattern so far, through the ongoing conflict between the US and Iran.

Then came regulated futures, options and US spot ETFs. Its yearly volatility fell from about 97% to 45%.

Calmer than it was. Still 3.5 times the S&P 500.

Its volatility over the past 12 months was 45%. Gold's was 27%, the S&P 500's 13%, US bonds 4%.

Nothing else on the board comes close, gold included.

That's 46% off the June low.

It needs another 46% to get back to the October high of $126,315.

The drop took nine months. The rebound so far has taken twelve weeks.

The index is calm, the stocks aren't. VIX closed at 14.2, lowest since December. Single stock vol is 39.

While that gap holds, a quiet VIX says little about single names.

Their three month correlation hit 65% in July, within a point of the record, and has held above 40% since April. It's 54% now.

While that holds, a big oil move is a rates move too.

But volatility measures magnitude, not direction, so that compression dampens rallies just as much as selloffs.

Yet Europe's net share of the global prime book sits near its lowest in 5 years while gross is mid range.

The S&P 500 is up 392% since it regained its 2007 peak in 2013. Thirteen years of one way traffic.

Seven times since 1871 a peak took years to see again: 25 after 1929, 20 after 1881, 18 after 1906, 8 after 1973, 5 after 2007. Only 1987 was quick, back at the high inside 2 years.

The drawdown is never the painful part. The waiting is.

IB desk flows tell the same story: asset managers sold on the week, with supply concentrated in macro products, communication services, industrials, consumer discretionary, and tech.

See the full table: https://sophosalpha.substack.com/p/institutional-monitor-00c

Less than 30% of S&P stocks are now above the 50-day moving average.

The last four meetings all traded the same way: red on decision day, green the next session, and the bounce more than paid back the drop each time.

Readings this negative are rare. In the months after them the S&P 500 did roughly double what it does in an average stretch, and finished higher most of the time.

The AAII bull bear spread is at a reading seen a handful of times in a decade. The S&P 500 has barely moved in a month.

The macro is not.

Central banks still tightening, the ten year at five, the AI trade under attack, midterms weeks away. Small cap vol is cheaper still.

Buy protection when you can, not when you have to.

Midterm years have been the best fourth quarter of the cycle, with most of the gain after election day. Presidential years are the ones that dip first in October.

We enter that window with far less crowding than in July.

It jumped into September expiry and gave it back on the roll, back in the lower third of its one year range.

Stocks are trading their own stories, not one macro story. That keeps index vol contained while single names get hit.

That is what less embedded single name leverage looks like. Fewer crowded upside bets, fewer forced unwinds.

September weakness sits in the back half, more so in midterm years.

Since 1928 the average midterm path drops about 1% from here to September 30, recovers through October and runs past Election Day, up 5.6% off the low.

Industrials, Financials and Consumer Discretionary account for the entire decline. Only Energy and Communication Services added against them.

7 of 11 sectors net bought.

Info Tech and Comm Svcs led for a third straight week on long buys, Comm Svcs flow near 2.5 standard deviations above its 1 year norm. Health Care paid for it, net sold a fifth week running and the biggest outflow on the board.

TMT cumulative flow still sits 1.5% of US gross market value below where it began 2026. Three weeks of buying has only walked it up from a July trough near 3.6%.

However, current flow estimates are dramatic given that exposure: Wednesday's session was enough to turn their fastest signals.

We anticipate this cohort sells on a flat tape, and on a down tape their selling runs at more than twice their buying potential on a rally. Expect the market to feel “heavy” on rallies while these flows work through.

Every time since the 1960s that the 10 year yield sat right of that line, stocks and bonds moved together and the hedge was gone. Yields are just under it.

If the energy shock pushes them through, 60/40 has nowhere to hide.

For nearly four decades the two moved as one line. Since June, diesel has been trading far above anything crude can explain. The shortage is in refining, not in oil.

That is why a US diesel export ban is now a real possibility.

That panic has since completely unwound.

Our Vol Stress Index climbed all week into the meeting, then collapsed the session after the decision to the calm end of its range.

Once the outcome is known, gold can sell off sharply.

COMEX Managed Money net long peaked near 890 tonnes at the 2016 Brexit vote and near 800 into the 2024 US election. Both unwound.

We modeled every tightening cycle since 1963.

With rates on the move like this we looked at what the 10Y does around the start of a hiking cycle. The chart shows the average across every cycle since the early 1960s, with the range of the individual cycles behind it.

A year after the first hike the 10Y is up around 100bp on average. More telling, only one cycle, 2004, saw yields fall at any point in that first year.

On the historical odds, yields are more likely than not to rise over the coming year.

Near its 3 month low the 10Y and stocks move together as often as not. Two quarterly moves above it and the correlation is negative 4 times in 5.

Today the 10Y sits two moves above its low.

We modeled every cycle since 1988.

The index is usually red 3 months on but the dip is uneven. Energy, Materials and Tech beat it in most cycles, Discretionary and Health Care lag. A year on the index is up 9% on average.

The 10Y averaged over 10% in the 80s and the S&P tripled.

It averaged near 7% in the 90s and the NDX went up 15x. Since the 2020 low it has averaged 3% and the S&P more than tripled again.

The one losing decade, the 2000s, had yields at 4%.

7 cycles since 1988. The 3 that started with the 10Y under 5% (2004, 2015, 2022) were all red 3 months later. The 4 that started above 5% held, and were all up a year on.

Yesterday's hike came with the 10Y at 5%.

Bond yields do not.

60 years of monthly data and the S&P has one losing setup, the month the 10Y jumps more than 2sd. Sort the same months by yield level and stocks made money at every level, 3% or 8%.

September so far is a move, not a jump.

We model their behavior closely in our Institutional Monitor reports and share weekly intelligence on where these funds are positioned.

Latest: https://sophosalpha.substack.com/p/institutional-monitor-b49

Hedge funds made money in August but still trailed a rallying S&P 500, and even the top strategy, systematic macro, posted only modest gains. Year to date, the average fund is up about half as much as the index.

That's with the smartest algorithms and brightest PhDs money can buy. The reason is that most funds aren't chasing the index. They're chasing steady returns while limiting risk, hedging their bets so they can protect capital when markets turn.

They give up some upside in a rally to avoid the big drawdowns that wipe out accounts. That's the part the screenshots never show.

We modeled every cycle since 1988.

The first three months are usually red across the S&P, Nasdaq, Russell and Dow. By six months it flips green, and a year on each index is up in most cycles. Only 2022 stayed red.

The S&P 500's implied move for today is the highest in the window, with Friday's triple-witching expiry close behind.

What usually happens next? Stocks have often slipped on FOMC day before recovering by Friday. The exception was 2022's hiking cycle, which makes today's expected hike worth watching closely.

https://substack.com/@sophosalpha/note/c-337498758

A modest rally into the decision, a selloff on the day, a recovery by Friday. The one stretch that broke the pattern was the aggressive hiking of 2022, when the pace of hiking drove the vol.

Brent has returned to the range last seen in the spring, and 10 year breakeven inflation has followed it almost step for step since the war began.

What crude does next is now a rates story as much as an energy one.

Real final sales of domestic product still expand at a healthy pace while the 10 year real yield has climbed to its highest since the early 2010s.

The two have moved together for decades. Growth can carry higher rates.

Breakevens have swung with oil but sit near where they began the year. Real yields have climbed since the Iran war as markets price Fed hikes.

In recent quarters the S&P 500 rallied through the first weeks of reporting, went quiet for the next four as the focus shifted to sustainability, then rallied again into the next season as guidance took over.

We are entering that stretch.

About 40% of S&P 500 companies are now in their quiet period, and by early October nearly the whole index will be.

That sidelines the market's most reliable buyer until earnings reopen the window, so dips get less support.

It still sits below the highs seen earlier this year. Hedging, not capitulation.

Last week they gave all of it back in a matter of days.

Single stock buying kept grinding higher through it, led by tech. Not a de-risking, a rotation into one theme with the index hedged.

Ordinary grinds higher, like this month so far, have been fine. The 1990s ran at far higher yields and stocks tripled.

Between 1996 and 2016 a run this long never happened once.

Bearish sentiment in the AAII survey is running at levels that historically came with a VIX near 25, not the 15 we have today. The gap between how investors feel and what vol charges is still wide open.

The VIX sits in the mid teens, yet the cost of downside protection keeps climbing. Calm on the surface, and paying up underneath.

Since that peak they have fallen by roughly a third, the kind of drop last seen when the post covid frenzy faded.

Investor anxiety is still high, the tape is just thinner.

The last time it traded here in the spring, oil implied vol was roughly double where it sits today.

The price has returned, the fear has not.

The stock bond relationship cannot be taken for granted when inflation, fiscal pressure and rising yields weigh on both markets at once. Yields sit near levels where that hedge has historically been least reliable.

Does that hurt stocks?

Since 1962 each 10bp the 10 year rises in a month has cost the S&P about 15bp the month after, and the losses cluster in the fastest rises. This move, 26bp in a month, is not the fast kind yet.

The Nasdaq 100 tripled as the 30 year yield added 200bps. Over an overlapping window it doubled again while the Fed hiked 125bps.

AAII bulls minus bears sits in the bottom quarter of readings since 1995, even with the S&P within 3% of a record set last month.

Nobody got euphoric on this rally, so there is no crowd to unwind.

Its deepest fall this year is about 8%, against 25% in 2022.

The S&P 500 has given up 9%, against 25%. Yields at multi year highs are a real fiscal signal, but the tightening carried with them has been far gentler.

Shorts now sit 29% below where the build started. The squeeze fuel everyone was pointing to has already been spent.

Average put call skew across NDX single stocks is near the flattest in 20 years, and implied vol has collapsed since the summer. Upside and downside cost roughly the same now. Convexity is cheap, not crowded.

Single stocks are moving on their own stories, and that is what keeps index vol pinned. History says correlation this low does not last, and the unwind tends to show up as a burst of index vol.

The part that gets missed is the rest of the market. Excluding energy, it grew 14%, its strongest in at least six quarters. The earnings base is broadening, not narrowing.

This quarter's de-grossing was one of the sharpest since 2020, and gross leverage now sits near the low end of the past year.

The median company isn't seeing it.

A growing share of index EPS growth is "other income," largely unrealized valuation gains on OpenAI and Anthropic stakes. Mark-to-market gains are not operating earnings.

The share of S&P 500 companies with positive forward EPS growth normally tracks ISM manufacturing with a 6 month lead. Since the AI capex boom it has run well ahead of the survey.

In 2026 they've been raising them.

This year's forecast path is tracking near the top of every year since 1996, alongside 2018 and 2021. Revisions this one-sided are rare, and leave little room for error.

It also happened in 37 of the other 72. The difference is depth, 9% vs 5%, then flat by 31 October.

VIX went in at 14.9. Only four midterm Augusts closed lower, and those ran to a median 1.8x.

In keeping with the pattern, the S&P 500 over the last month has chopped in a tight range of just over 2% despite several high-profile catalysts (Jackson Hole, geopolitical escalation, trade escalation, CPI, payrolls).

https://substack.com/@sophosalpha/note/c-321506892

Midterm years are tougher, but note that just before and after midterms the market fully recovers. Is that enough to justify lightening up?

After midterms the market tends to be strong regardless. 2026 initially followed the script but markets have been climbing since early August.

Excluding the pandemic anomaly, core PCE is only 12bps off its 34 year high seen in May.

Nominal GDP may be inflated by energy prices but overall it is hard to say US yields are too high based on this.

If nominal GDP settles at 5.5% then 4.8% would be historically average.

So even if not sustainable at these levels, this level of nominal growth has helped push UST yields higher.

10yr average NGDP is c.5.5%, so 4.8% seems broadly fair. But current NGDP is 6.6%. Do we factor that in?

The average down day moved 0.43%.

Rallies are running about 1.6 times as violent as sell offs. That is not what a market full on exposure looks like, it is one chasing strength it does not own.

The MOVE index closed Friday at 73, cheaper than roughly 85% of the last five years.

Odd place to sit with the long end back at multi decade average yields and a CPI print landing Friday.

Sticky inflation, heavy supply and rebuilding term premium can carry it higher with no Fed hike at all. That is the real risk to long duration assets.

It came from both sides of the book, new longs plus short covering, so this isn't just a squeeze. The only bigger week on record came less than a year ago.

(Y=axis what it means).

Our US 2Y and JGB shorts are 185 days old, the US belly 176, and all 19 markets are green since entry. This decline hasn't been a surprise to systematic money, it's been the trade.

Option market and general hedge unwinds fueled a large portion of the bounce.

Full update on institutional activity will be published on Sunday. Get the latest release in-time by subscribing for FREE: https://sophosalpha.substack.com/subscribe

Global funds returned +0.6% while global equities gained +2.7%.

What they made came from beta, not selection. Long-short spreads were negative globally at -0.3%, with North America weakest at -0.5%. Equity L/S was the one bright spot at +0.8%. YTD the gap is wider still, +8.9% vs +14.6%.

Beta is carrying the industry, not the stock picking.

Core is going sideways, the share of items running hot isn't widening, and expectations stay anchored. The ECB's risk isn't the rebound itself, it's how long the energy shock lasts.

They're selling because the macro mix is starting to smell like stagflation.

Non-dealers flipped back to net selling in S&P futures in the latest COT, ending a two-week buying streak. New shorts and long liquidation contributed about equally.

Hedge funds have now cut gross longs three weeks running, and each leg lower lines up with the stagflation basket ripping higher.

Price is holding. Positioning is repricing growth vs inflation.

The GS sentiment indicator has drifted back to neutral after peaking in the spring, even with the index holding near its highs. Fundamental investors are hesitant here. That is not what a fragile tape looks like, it is what a cushion looks like.

Hedge funds have barely committed.

Long-only money is close to fully invested, while hedge fund net exposure sits near the low end of the past year. One crowd has nothing left to buy. The other has all the room.

S&P top of book has held up through the late summer stretch, and last Friday was the first session in almost three weeks that depth fell below its 1y rolling average. At $11.12mm as of last week's close, the ability to transfer risk remains robust, though we expect a moderate weakening into the official end of summer.

Our systematic trend following portfolio has been adding to gold longs since start of August, when momentum-driven conditions turned meaningfully better. Platinum moved the same way and copper stayed long throughout. Silver and palladium sit closest to the flip.

Gamma is mid-range versus the past year though, so treat it as a cushion, not a floor.

Average daily return for every calendar day of the year, 1928 to 2026.

We post dozens of charts like this in our daily notes, subscribe for FREE.👉 https://sophosalpha.substack.com/subscribe

How closely index stocks are expected to move together has fallen from a typical 44% before 2017 to 13% today. The index cancels itself out, until it stops.

Day to day it has been calmer than 71% of moves that size since 2003. Big moves usually come with a rough ride. This one arrived as a slow grind.

The S&P 500's day to day swings are in the calmest 12% of five years, yet protection is still priced at 16. That gap sits in the top fifth of history, and all of it goes to whoever sold you the hedge.

The S&P 500 moved 2.2x harder on its up days than its down days, the 14th highest reading in 488 months since 1986. The volatility now lives in the melt ups, not the selloffs.

The link between S&P 500 down days and a rising VIX normally runs near 81%. Over the past month it collapsed to 39%, looser than all but 5% of the last 36 years. The alarm system went quiet. The risk did not leave.

For 36 years it rose about 1 point for every 1% the S&P 500 fell. Right now it manages 0.45. Only 9% of history has seen a weaker reaction. Protection looks cheap because it has quietly stopped protecting.

Since 1928 the back half of the month has averaged 0.91% negative, the weakest two week stretch of the year, and has been higher in only 45% of years. The first half is barely negative. That window opens in two weeks.

What follows is a very different catalyst calendar: NFP Sep 4, PPI Sep 10, CPI Sep 11, FOMC Sep 16.

Unlike earnings, that is a far more two sided set. After a season dominated by positive surprises, the right tail is less obvious.

Long buys were roughly offset by short sales, leaving the book marginally net bought with overall dollar flows muted.

Gross positioning on both sides is large, but cumulative net flow has barely moved since the spring.

The buyback window is at its widest around September 12. From there an increasing share of companies enter pre earnings blackout ahead of Q3 reporting, and one of the market's largest and most consistent sources of structural demand gets progressively smaller as the month advances.

Buying the SPX on election day and holding: three months later you were up 7 times out of 9 (+5.5% average), six months 9 of 9 (+11.4%), nine months 9 of 9 (+12.3%), twelve months 9 of 9 (+14.7%).

The average path is flat for the twelve months into the vote, up just 2.8%, then gains 14.1% in the twelve months after it. 2026 is running well ahead of that, up 12.2% at the same point in the cycle.

Seasonal moves may be exacerbated by midterm election seasonality, which tends to be soft into the election and then strongly positive after it.

+13.3% in August, one of a handful of 10%+ months since 2000, and speculative demand hit a nominal record along the way.

The bearish share has averaged above 40% for four weeks, a level that has coincided with the S&P sitting 7 to 10% off its highs. Today it is less than 1% away.

Currently, a recession is highly unlikely to manifest over the next few quarters. What is unclear is if the Fed hikes, and what the scale and length of the hiking cycle would be.

Meanwhile, professional investors have not been chasing this last leg higher.

The cost of carrying S&P 500 exposure, which reflects how much leveraged longs are paying dealers to hold their positions, spiked into the July rally and has been easing since, even as the index printed fresh highs in August.