Short-Term Reversion Keeps Paying Under a Record S&P 500

• Short-Term Momentum's trailing 20-day return of -5.62% ranks in the bottom 0.2% of all 20-day windows since 2020; today added another -0.33%.
• Workday closed +17.57%, adding $7.6B of market value, after Reuters reported Silver Lake take-private talks.
• Cisco closed -8.37%, erasing $43.5B, as a 210 bp gross-margin contraction outweighed a record $4B of AI orders.
• Realized style-factor volatility is running at 27.7 vol points, the 93rd percentile since 2020, against 12.6 points on the index leg.

August 13, 2026 · 16:21 ET

Both ends of the sort contributed. The worst recent multi-day performers gained roughly +1.0% while the strongest lost about -0.5% — last week's laggards outrunning last week's leaders, for the fourth consecutive week. Over 60 sessions the factor has returned -7.51%, and its 20-day z-score of z=-4.0 is the most extreme cross-sectional reading on the board by a wide margin. The exception is this week itself: the five-day return is -0.10% at z=-0.1, so nothing has been paid since Friday. The case sits on the 20- and 63-day windows, not on the last five sessions.

What Makes the Snap-Back Stick

Three conditions are in place, and they are precisely the ones under which short-horizon reversion runs hardest. Single stocks are moving violently on their own stories: implied single-stock volatility (Cboe VIXEQ) sits at 37.75, the 80th percentile of its own history, and realized style-factor volatility at 27.7 vol points ranks in the 93rd percentile since 2020 against just 12.6 points on the index leg. Index volatility, meanwhile, is calming — VIX at 14.62 is 29% below its 20-session high of 20.66 set July 29. A market that churns underneath a quiet index is one where each week's leaders and laggards trade places.

The third leg is balance-sheet churn after a stress episode. Goldman Sachs Prime Services data covering the week ended August 12 showed a second consecutive week of global equity net buying, long purchases outpacing short sales 1.4-to-1, following a substantial reduction in systematic hedge-fund leverage; the same data set the episode's floor, with Citadel absorbing a roughly $16 billion block from the collapsed Situational Awareness fund on August 6 at a 10% discount after its 67% July drawdown, as Morningstar reported. Our own ownership lens shows no fresh pressure on that book today — hedge-fund-owned names lagged only modestly after style controls at z=-0.7, with the raw spread flat. A multi-week Short-Term Momentum run this deep, straddling a documented de-grossing and re-grossing episode, is indicative of hedge-fund deleveraging and releveraging churn rather than any directional view. In yesterday's session, the high-beta momentum cohort traded on light volume, with gains concentrated in the easiest-to-trade names.

Where the Risk Actually Sat

Almost every large move today was the stock's own. Workday's +17.57% close was 15.5 points of stock-specific move, the largest in its trailing year, after Reuters reported Silver Lake is in talks to take the company private; software rode it, with IGV up +2.75% on 2.5× the typical session pace Economic Times. Cisco fell -8.37% despite $17.3B of quarterly revenue and record $4.0B AI infrastructure orders, as non-GAAP gross margin contracted 210 bp to 66.3% MorningstarCNBC. Cellebrite closed -29.61%, $1.1B of market value gone, after cutting its 2026 revenue outlook and announcing an immediate CEO change the SEC; Tapestry -16.42% and Resideo -20.48% on soft forward guidance. Applied Materials closed the regular session -2.50% and fell a further -2.71% in late trading after guiding to a record $10.25B of fourth-quarter revenue Bloomberg.

The subtler read is which gains were bought and which were simply carried. Liquidity and Residual Volatility are one trade, not two — their exposure vectors correlate 0.66, so the highest-turnover names and the highest-volatility names are largely the same list. The Trade Desk's +7.30% was mostly that carry, +4.81% of it from the factor complex against +1.98% of its own move; AppLovin's +3.10% was entirely factor. Sandisk's +13.69% was its own move, and no driver for it appears anywhere in today's coverage.

China is the cleanest illustration of the day's shape. KWEB fell -2.47% on 3.0× the typical session pace, JD.com closed -7.56% after second-quarter revenue declined 2.9% year over year, and Fitch cut Baidu to 'A-' FITCHRATINGS — yet the China factor itself rose +0.33% at z=+0.8, meaning the most China-exposed names as a group outperformed. The e-commerce damage was name-specific, not an exposure move, which is why a China hedge would have done nothing for a book holding JD. Korea ran the other way, EWY +1.82% as the KOSPI entered a technical bull market on a 155% year-over-year jump in early-August semiconductor exports MarketWatch. Precious metals gave back a slice of a big month — GDX -2.87% on 2.4× typical pace after a +22.92% 20-day run, Pan American Silver closing -9.62%. And the Bitcoin factor's +0.45% came from a top bucket of high-volatility proxies, not the miners, which fell hardest of anything on the board.

The plateau underneath all of this: realized 20-day pairwise correlation is 8.43%, the 4.1st percentile since 2020, but the 3-month clock has it at the 54th — the extreme was set earlier and is holding, not deepening. Implied single-stock correlation (Cboe COR1M) at 7.71 remains in the lowest 1.3% of its 20-year history, so the options market is priced for stocks to keep going their own way. That is the environment a book gets paid in for name selection and punished in for carrying last week's winners.

Economic Context

The Producer Price Index the BLS, released at 8:30 AM ET, was flat month over month in July against a 0.2% consensus, with core up 0.2% versus 0.3% expected; energy prices fell 3.1%, led by a 5.7% drop in gasoline. Initial Claims the BLS, also at 8:30 AM ET, rose 9,000 to 209,000 for the week ended August 8, above forecasts of 202,000–204,000 and a third consecutive weekly increase, while continuing claims fell to 1.78 million. The curve bull-flattened modestly in response: the two-year yield fell 4.8 bp to 4.151% and the ten-year 4.3 bp to 4.649% CNBC.

Data compiled by FactorPulse AI; edited and verified by Jeff Klein. For informational purposes only. Does not constitute financial advice, an investment recommendation, or an offer to buy or sell any securities. Always consult a qualified financial professional before making investment decisions.

For more on factor construction methodology, see www.factorpulse.com/glossary.

← FactorPulse  ·  All Digests