Reference 03 November 2026 · 7 min

A plain-English guide to the Traders in Financial Futures report: leveraged funds, asset managers, and why a record short is not a forecast.

Written Tuesday November 3, accompanying the November 3 analysis: in the CFTC book for October 27, leveraged funds were net short 2,077,860 contracts of the 5-year and 2,227,275 of the 10-year, asset managers net long 2,938,901 and 3,000,164, and the bond short of 150,393 was the smallest of 52 weeks, down from 415,340 a year earlier (CFTC, via bin/tff_report.py). What the report is, the four categories, why the Treasury short is mostly the cash-futures basis trade (about $830 billion as of September 2025, per the Fed), how the percentile and noise floor are built, and what the report cannot see.

By Igor Komolov

Written Tuesday November 3, 2026, accompanying the Tuesday November 3 analysis note, which recorded the latest leveraged-fund and asset-manager positions in Treasury futures alongside a JOLTS day that graded two hits. In the CFTC's Traders in Financial Futures report for positions as of October 27, leveraged funds were net short 2,077,860 contracts of the 5-year note and 2,227,275 of the 10-year, while asset managers were net long 2,938,901 and 3,000,164; the leveraged-fund short in the bond contract, 150,393, was the smallest of the past 52 weeks, down from 415,340 the previous November. Those numbers look like a verdict on yields. Mostly they are not. This guide explains what the report is, who the four categories are, why the leveraged-fund short in Treasuries is largely a financing trade rather than a rate view, how we normalise the numbers so that "record" means something, and what the report cannot tell you.

What the report is

The Commitments of Traders reports are the CFTC's weekly breakdown of open interest in US futures by type of trader. The legacy report splits positions into commercial and non-commercial. In 2010 the CFTC added the Traders in Financial Futures (TFF) report for financial contracts (Treasuries, currencies, equity indices, VIX) with four categories that are more useful for those markets: Dealer/Intermediary, Asset Manager/Institutional, Leveraged Funds, and Other Reportables (CFTC explanatory notes). Positions are as of the Tuesday close and are published the following Friday at 3:30 PM ET, so by the time you read a report the book is three days old, and by the time the next Tuesday's data is in your hands it is a week old. Our plans cite both dates for that reason. There is a futures-only version and a futures-and-options-combined version; the figures on this site are futures only, pulled from the CFTC's public API (dataset gpe5-46if) by bin/tff_report.py.

The four categories

Dealer/Intermediary. Sell-side: the large banks and dealers that make markets and warehouse the other side of client flow. In Treasuries they are net short in the October 27 book, as they usually are.
Asset Manager/Institutional. Pension funds, insurers, endowments, mutual funds. In Treasury futures they are the long side: 2.9 million contracts of 5-year and 3.0 million of 10-year in the latest report. A fund that wants duration exposure without holding the bond buys the future.
Leveraged Funds. Hedge funds and commodity trading advisers, including the relative-value funds that run the basis trade below. In Treasury futures they are the short side, and have been for years.
Other Reportables. Everything reportable that is none of the above, including corporate treasuries and smaller participants above the reporting threshold.

The CFTC assigns a trader to a category by its predominant business, declared on its registration, and a single firm can appear in more than one category if it has separately managed desks. The categories are not pure, and the CFTC says so in the notes.

Why leveraged funds are short Treasuries

The intuitive reading of a 2.2 million contract hedge-fund short in 10-year futures is "hedge funds think yields are going up." The larger part of it is something else. The Treasury cash-futures basis trade buys a cash Treasury, sells the futures contract against it and finances the bond in repo, collecting the gap between the futures price and the price of the cheapest bond deliverable into it. It is a convergence trade, and it shows up in the TFF data as a leveraged-fund short in futures, matched by a long in cash bonds that the report cannot see. The Federal Reserve's own research says that most market participants use the leveraged-fund short in Treasury futures from this report as their gauge of basis trading, and warns that it is an imperfect gauge because leveraged investors can also be short for directional or other relative-value reasons (FEDS Notes, March 8, 2024). A later note, using hedge funds' confidential Form PF filings rather than futures data, puts the trade at "approximately $830 billion as of September 2025" (FEDS Notes, June 22, 2026).

The other side of that trade is the asset-manager long. A fund that wants Treasury duration and buys the future rather than the bond is, in effect, paying a leveraged fund to hold the bond for it. That is why the two columns move together, and why the asset-manager long is the number to read next to any leveraged-fund short before calling the short a view.

Normalising: percentile and noise floor

A net position in contracts is not comparable across markets (the 2-year contract has a $200,000 face, the others $100,000) or across years (open interest grows). The site's script reports two normalisations for the leveraged-fund net, the same approach as the COT positioning guide uses for the managed-money bucket in commodities:

Percentile within a trailing window. The share of the last 52 weekly readings at or below the latest net (the script's default window is 156 weeks; the plan quoted 52). Because a short is a negative net, a high percentile means a small short and a low one means a large short. The bond's 100.0 means the latest net is the highest of the 52, the smallest short of the year; the 10-year's 26.9 means the short is larger than it was in roughly three-quarters of the past year; the euro's 1.9 means the latest net is the lowest of the window, the largest short. A percentile is a rank, not a probability, and "record" always carries the window's length. It is also easy to read backwards, which we did in the first draft of this week's plan and corrected before the release.
Noise floor. 1.5 times the mean absolute weekly change over the trailing 12 weeks. A week-on-week change smaller than that is within the ordinary churn of the series and the plan writes it as noise. In the October 27 book the 10-year short grew by 120,051 contracts against a floor of 125,825, so even the largest Treasury move of the week was noise by this test; the bond short shrank by 12,364 against a floor of 52,878, which is how a year-long decline from 415,340 to 150,393 arrives: in steps that are each too small to call.

The two answer different questions. The percentile says whether the level is unusual; the floor says whether the week's change is. The dangerous combination in a directional market, positioning at the edge of its range while price has already turned, is described in the COT guide. In Treasury futures the leveraged short is structurally large for the basis reasons above, so a high percentile is a weaker signal than the same percentile would be for a currency.

The rest of the book, October 27

From the same report, leveraged funds were net short the euro at the 1.9th percentile of the year (71,574 contracts, the largest euro short of the 52 weeks), net short the yen at the 94th percentile (14,630, among the smallest yen shorts of the year), short the Canadian dollar at the 7.7th, long the pound at the 31st, short the S&P e-mini at the 50th and short VIX futures at the 52nd. Every one of those week-on-week changes was within its noise floor except the yen, where an 8,260 contract reduction in the short was also inside a 36,378 floor. The full table is in the November 4 plan (editorial/plans/2026-11-04.md). None of it is a call; it is where the book stood a week ago Tuesday.

What the report cannot tell you

Cash positions. The report sees futures. The cash Treasury held against a basis short, the swap against a futures long, the option against either: invisible.
Why. A category is a type of firm, not a motive. The same leveraged-fund short can be a basis trade, a curve trade, a hedge for a mortgage book or a view on the Fed, and the report does not distinguish.
Anything about this week. Positions are a week old on publication. A plan written on a Wednesday is quoting a Tuesday book that was published the previous Friday.
Gross exposure. Net long minus net short hides the size of the book. The report publishes gross longs and shorts and spreading separately; the net is a summary.
Whether an extreme matters. The euro short was the largest of 52 weeks on October 27 and the bond short the smallest; both weekly changes were within their noise floors. An extreme reached by small steps is not a washout setup, and the report has no way to say what would make it one.

A checklist for reading a Friday report

Cite both dates: positions as of Tuesday, released Friday.
Read the asset-manager column next to the leveraged-fund column before saying who "the market" is.
Quote the percentile with its window, and the week's change against the noise floor. Run the script; never quote positioning from memory.
Treat a Treasury leveraged short as a financing position until something in the cash market says otherwise.
If the API is down, say positioning was not checked. A plan with no positioning section is better than one with a remembered number.

Where this shows up in practice

November 3 (Tuesday): JOLTS 7,523 thousand, two hits, and the October 27 Treasury book for the record (the live example)
October 30 (Friday): the CFTC release recorded, not read, the practice this guide replaces
The COT positioning guide, where the percentile and noise-floor method is worked through on corn