The week leveraged funds bought back their rate shorts
A 99th-percentile short-covering move in long bonds, a tape that cooled on two independent measures, an index that hid a 31-point spread between its best and worst large stock, and a nine-session slide nobody indexed.
The index rose +0.5% while participation fell on every horizon and sentiment gave up 11.6 points. The consequential change was in positioning: leveraged funds covered long-bond shorts at a 99th percentile pace and cut rate and Nasdaq shorts together, while adding to their S&P short in the same report.
What changed
Leveraged funds cut long-bond shorts at a 99th percentile pace
Weekly change in leveraged-fund net position
Contracts · week to the Tuesday observation, released Friday
The largest positioning change of the week was in the long bond. Leveraged funds bought back 58,389 contracts of 30-Year T-Bond exposure, cutting a net short of 361,383 to 302,994. Measured against 1,053 weekly changes in the record, that is the 99th percentile by absolute size.
It was not alone. The same group cut 94,674 contracts of 10-Year T-Note shorts, 20,539 of E-mini Nasdaq-100 shorts and 19,357 of Euro FX shorts. Four instruments moving the same direction in one week is what separates a position adjustment from noise.
The reading that keeps this from being a simple risk-on story is what they did elsewhere in the same report: they added 33,802 contracts to their S&P 500 net short and 11,050 to their VIX short. Shorts were covered in rates and in the Nasdaq, and rebuilt in the broad index.
Boundary: Commitments of Traders reports the Tuesday position and publishes it the following Friday. Everything here is dated 2026-08-25 and says nothing about Wednesday to Friday.
The tape cooled on two measures that do not share an input
Share of stocks above their own moving averages
Friday to Friday
SPY closed the week at 769.35, +0.5% higher. Underneath it, participation fell on every horizon we track: the share of stocks above their ten-day average went 45.56% to 38.88%, and the fifty-day reading went 53.02% to 48.29%.
Sentiment moved with it. Fear & Greed fell from 63.27 to 51.64, a 11.63-point weekly change that took it out of its prior zone and into neutral. The two measures are built from different inputs, which is the reason to report them together rather than separately.
A cooling tape under a rising index is not a warning on its own. It is the condition under which an index gain depends on fewer and fewer names, which is the thing the next story measures directly.
The index moved +0.5% and hid a 31-point spread
Weekly return, largest movers inside the large-cap set
Five best and five worst · Friday to Friday
CRM rose +22.4% on the week. MRVL fell −8.6%. The average absolute move across the large-cap set was 2.39%, and the index that contains both finished +0.5%.
An index return is a weighted average, and a small one is compatible with almost any amount of disagreement underneath. This week the disagreement was concentrated in one sector rather than spread evenly: enterprise software led the upside while semiconductor and energy names led the fall.
Dycom fell 25% and closed lower nine sessions running
A nine-session decline, with the week shaded
Daily bars and volume · 90 calendar days through August 28
DY closed at $294.34 on 2026-08-28. On the event date, 2026-08-28, volume was 933,238shares and the close was $294.34.
Dycom closed at 294.34 against 392.95 a week earlier, −25.1% on the week, and it has now fallen for 9 consecutive sessions.
A streak of that length is a different object from a single shock: it is a repricing that nobody stepped in front of for nearly two weeks. We are not attaching a cause. No company release in the week explains it, and inventing one would be worse than leaving the observation where it is.
Boundary: No verified company filing or release is attached to this move. The streak is the finding; the reason is not.
Solana rose +13.4% and the rest of crypto did not follow
Weekly return by crypto product
Friday to Friday
The Solana product gained +13.4% while the broad benchmark managed +1.1% and the Litecoin product fell −5.0%. Two weeks ago every product on this board rose at least 19% together; this week one of them moved and the others did not.
That is the difference between a rebound and a rotation, and it is the reason this appears as a small item rather than a headline. One product moving is a position rather than a market.
What did not confirm
A weekly record should preserve the evidence that refused to join the headline. These three gauges kept this issue from becoming a broad risk-off call.
Volatility did not confirm the cooling
The VIX/VIX3M ratio ended at 0.826, well below 1.00, with VIX itself at 14.43. Falling breadth and falling sentiment did not reach the volatility curve. Friday's index closes stay provisional until the official Cboe settle audit.
Open live gauge →The average stock did not break down
The Hidden Bear Index still reads mixed: the average large stock sits 0.9% below its own high against 1.1% for the index, a gap of 0.2 points. A cooling tape is not yet a damaged one.
Open live gauge →Three candidates were not this week
The scanner nominated volume spikes in NSA, WBS and SEM whose triggering sessions fall on July 21, August 19 and June 30. They are real anomalies from other weeks, admitted by a rolling window that is not clipped to the issue. They are excluded here and the scanner needs the filter.
What carries into next Friday
Does the rate covering continue?
The 30-Year position moved from a COT index of 74 to 92 in one week, which is a crowded reading. Next Friday shows whether the covering extended or reversed.
Watch live →Does short-term breadth recover 50%?
The share of stocks above their ten-day average is the fastest of the four horizons to turn. Back above 50% and this week reads as a pause; a second week lower and the index is carrying more of the load alone.
Watch live →Does the S&P short keep growing?
Leveraged funds added to their S&P net short in the same report where they covered rates. Whether that divergence widens or closes is the cleanest read on whether this was a rates trade or a risk trade.
Watch live →Method and source trail
The issue is a frozen Friday-close document. The Rust scanner combines liquid stock anomalies, historical series extremes, horizon-score changes, breadth, sector and cross-asset returns, release-aware economic data, and CFTC positioning. It emits evidence and caveats; it does not emit publishable prose.
Editorial selection narrowed the receipt to 5 changes. Company and government claims were checked against primary releases, and each dated observation links back to its live page. Historical percentiles use each series' available record. Current-universe cross-sections remain labeled point-in-time exhibits.
Figures above will not update. Live tools may change with later closes, revisions, or source corrections. Material corrections to this issue will move its modified date and be stated on the page.
Get next Friday's link by email.
Substack is the delivery rail. The long issue, live gauges, and frozen evidence stay here. Founding Market Changelog editions remain free.