When Speculators All Bet One Way

Sources & method

Every position is from the Commodity Futures Trading Commission's weekly Commitments of Traders report, read in its raw form from the CFTC's public data service and checked against its yearly archive files. Oil prices are the EIA's daily close of the NYMEX front-month crude contract; the 10-year yield is the daily series DGS10 from FRED, the Federal Reserve Bank of St. Louis database. The reasons given for each move are quoted or paraphrased from the EIA, the Federal Reserve and the studies named below; where a source hedges (for example "probably"), the text keeps the hedge. All figures were checked on October 2–3, 2026; nothing is estimated unless it says so.

The report.

The CFTC publishes positions as of Tuesday (in a holiday week, Monday) every Friday at 3:30 p.m. Eastern time, for markets in which 20 or more traders hold positions at or above the reporting levels. The report has been published since 1962, monthly at first; the futures-only history the CFTC publishes online begins in 1986. A new high is a week whose net bet is larger than in any earlier week; the record is the last new high before the bet stops setting them, counting a gap of up to eight weeks between new highs as the same stretch. This essay uses the legacy futures-only report, with three groups (non-commercial, commercial, non-reportable), from 1986, and for the split of the commercials the disaggregated futures-only report, published since September 4, 2009, with data from June 2006. Non-reportable positions are what remains of open interest after the reportable positions are subtracted, so the number and classification of those traders are unknown. The legacy non-commercial net in oil equals the disaggregated "managed money" net plus "other reportables" net (on January 30, 2018: 482,780 + 251,778 = 734,558), so about a third of what this essay calls speculators are not managed-money funds. Press figures often quote managed money only, futures and options combined and sometimes both WTI contracts (NYMEX and ICE Futures Europe); on January 30, 2018 that measure was 531,235, which is why it differs from the figures here. In the record week 411 traders were reportable in this market, of whom 184 held non-commercial long positions; a trader is counted in every category in which it holds a position, so these are not added up. Classification follows the entity: one organization can have a commercial unit and a separate non-commercial unit.

Net bet.

Long minus short; spread positions (equal long and short in different months) are not part of it. The three legacy nets add up to zero in every one of the 1,934 weekly rows for WTI since 1986, because all long open interest equals all short open interest; the legacy commercial net equals producer/merchant net plus swap-dealer net in every one of the 1,059 weeks since 2006. The report shows each group's total long and short positions and their weekly change, not the trades behind them; a fall on the long side can be a closed position, a trader dropping below the reporting level, or a change of category. Non-commercial long and short: 862,102 and 123,005 on February 6, 2018; 653,899 and 104,990 on October 2, 2018; 498,716 and 221,505 on January 8, 2019. From February to October the long side fell by 208,203 and the short side by 18,015; from October to January the long side fell by 155,183 and the short side rose by 116,515, which is 42.9% of the net decline of 271,698.

Week-by-week comparison (our calculation).

Weekly change in the front-month close on report days against the change in the non-commercial net, simple correlation. February 13 to October 2, 2018 (34 weeks): +0.63 in the same week, +0.20 with the following week; after 8 of 22 rising weeks the net rose the next week. During the build-up, June 20, 2017 to February 6, 2018: +0.52 and +0.30; 18 of 24. For 2010 to March 2019: +0.50 and +0.27; 133 of 241. Within the same week, from February 13 to October 2, 2018, the net rose in 12 of the 22 rising weeks and fell in all 12 falling weeks, so in 24 of the 34 weeks it moved the same way as the price. No significance test was run, the weekly price changes include contract-switch jumps, and nothing here is evidence of cause or a test of the studies cited.

What "record" means.

A new high is a week whose net bet was larger (or, for a bet on a fall, more negative) than in any earlier week since 1986, counted in contracts; the record is the last new high of a run. The run that ended with the record of February 6, 2018 began on November 14, 2017 and held ten new highs. February 6, 2018 is also the largest value in the series' full history to the report of September 22, 2026, and so is the bond short of October 1, 2024. Open interest, all contracts outstanding, grew in WTI from 74,334 in January 1986 to 2,589,270 on February 6, 2018, more than thirtyfold. As a share of open interest the February 6 record was 28.5%, the third-highest week; the highest share was 29.4% on March 20, 2018, and the second was 28.6% on February 27. In the 10-year note, open interest grew from 75,839 to 4,969,440 by October 1, 2024; the record net short of that day was 23.0% of open interest, the most negative share since February 1987, when a net short of 14,985 contracts was 25.5% of a much smaller market.

Price series.

The essay uses the front-month NYMEX close (EIA series RCLC1), the market's usual reference price; the positions themselves are the sum of all delivery months, many of them later than the front month. The front-month series switches to the next contract at each expiry, so it can jump on the day of the switch; the percentages describe changes in the price level, not the return on a held position. The 2014 and 2017 high dates are confirmed by the spot price series; the 2018 end points are not near an expiry (the November contract expired on October 22, 2018, the January contract on December 19, 2018). June 20, 2014 was the expiry day of the July contract, so $107.26 was its last close (the August contract closed at $106.83 that day); February 21, 2017 was the expiry day of the March contract, and the high two days later is on the April contract, so the essay gives the date of that high and no percentage rise. On the spot series (EIA, Cushing WTI) the 2018 high was $77.41 on June 27, 20 weeks after the record, and the two series differed that day by more than $4 ($77.41 against $72.76); the essay uses one series throughout and does not mix them in a claim. The 2017 low of the front-month series was $42.53 on June 21, 2017, the same figure as its low of December 24, 2018. The front-month series ends on April 5, 2024, so this essay carries no live oil data.

How the lag is measured.

A new high is a week in which the net bet was larger than in any earlier week since 1986; a run is a series of new highs at most eight weeks apart; the record is the last new high of a run, that is, the last new high before the bet stops setting them. The lag runs from that week to the highest front-month close the price reached before it next fell by more than a fifth from that high. The table covers every run since 2010; earlier runs, among them those ending in 2005, 2006 and 2007, are not part of it.

Run ends	Net long	Close that day	Price high	Lag	Rise to the high	The fall that followed
Jan 12, 2010	135,669	$80.79	Apr 6, 2010 · $86.84	12.0 weeks	+7.5%	−21.7% (to May 20, 2010)
Dec 7, 2010	176,557	$88.69	Apr 29, 2011 · $113.93	20.4 weeks	+28.5%	−33.6% (to Oct 4, 2011)
Mar 8, 2011	275,582	$105.02	Apr 29, 2011 · $113.93	7.4 weeks	+8.5%	the same fall
Jul 30, 2013	362,941	$103.08	Sep 6, 2013 · $110.53	5.4 weeks	+7.2%	a fall of a fifth came only on Oct 8, 2014
Mar 4, 2014	425,818	$103.33	Jun 20, 2014 · $107.26	15.4 weeks*	+3.8%	−58.6% (to Jan 28, 2015)
Jun 24, 2014	458,969	$106.03	Jun 25, 2014 · $106.50 (forward); the year's high was Jun 20 · $107.26	+1 day (−4 days to the year's high)	+0.4%	the same fall
Feb 21, 2017	556,607	$54.06	Feb 23, 2017 · $54.45	+2 days	(not stated, see Price series)	−21.9% (to Jun 21, 2017)
Feb 6, 2018	739,097	$63.39	Oct 3, 2018 · $76.41	34.1 weeks	+20.5%	−44.3% (to Dec 24, 2018)

* 14.4 weeks (Jun 13, 2014 · $106.91) if the series switches contracts one day before expiry. The eight runs belong to five separate falls (2010, 2011, 2014–15, 2017, 2018). The 2017 fall is at the threshold: with a threshold of a quarter instead of a fifth, it would not count. Between the 2018 record and the October high the largest setback was −12.3% (August 15, 2018). The 2014 and 2017 records would normally have been published on June 27, 2014 and February 24, 2017, both after the price highs.

The 2014 and 2017 records.

The record of June 24, 2014 came four days after the year's highest closing price, June 20, which fell in the week before the record; after the record the price held close to that level for one more day (June 25, $106.50). It then fell by three-quarters, to its low in February 2016. The EIA's reason: "robust global production exceeded demand", with much of the extra oil coming from North America. The record of February 21, 2017 came two days before the price high of February 23 ($54.45); by June 21 the price had fallen by more than a fifth. According to the EIA, prices fell in early March as US crude inventories built to a multi-decade high and US production rose, and in June they fell again after reports of large inventory builds. One of the five earlier records, December 7, 2010, was followed by 20 more weeks of rising prices, to a high on April 29, 2011.

The split of the commercials.

The disaggregated report splits the legacy commercials into producers and merchants, firms that produce, process or handle physical oil and use futures to hedge the risks of that business, and swap dealers, entities that deal mainly in swaps (private agreements with clients that fix a price for a later date) and use futures to hedge the risk those agreements leave with them. Their clients, in the CFTC's words, may be "speculative traders, like hedge funds, or traditional commercial clients". A group's net bet is its bets on a rise minus its bets on a fall, so it shows which way the group as a whole stood to gain or lose; that is why the essay uses it. Counted before that subtraction, producers and merchants held about a third of all the commercials' bets on a fall, but most of those were balanced by bets on a rise within the same group, so they do not show in its net. One kind of client the CFTC's definition describes is an oil producer that wants to fix the price of oil it will pump later: under such a swap the dealer pays the producer if the price falls, so the dealer stands to lose from a fall and can offset that by selling futures. On January 30, 2018: swap dealers −680,564 net, producers and merchants −84,233; together −764,797, so swap dealers held 88.99% of the commercial net short. Of the gross commercial short (1,491,842, including 165,110 swap-dealer spread positions), swap dealers held 68.2% and producers and merchants 31.8% (473,963 short against 389,730 long). The share is not drawn as a line over time, because it is meaningful only when both nets have the same sign: from 2006 to 2009 swap dealers were net long every week, and in later years the weekly share ran from below zero to above 100%. On October 11, 2016 it was 27%, when producers' gross short (540,757) was the largest since 2007; that week falls in 2015–2016, when the share was unusually low compared with 2013–2014 and 2017–2018, so it is not the start of a trend toward 2018. The swap-dealer logic above is a conditional reading of the CFTC's definition: if the dealers hedged their swaps in futures, their net short faced a swap book that gains from a rise, and their clients stood on the side that gains from a fall or is protected against one. The report shows neither the clients, nor how fully the dealers hedged, nor whether they also hedged elsewhere.

Bond futures and yields.

The 10-year Treasury note future is traded on the Chicago Board of Trade, in contracts of $100,000 face value. The record net short of October 1, 2024 was 23.0% of open interest, the most negative share since February 1987 (see What "record" means). Long-term yields rose sharply after the Federal Reserve committee's December meeting and an "unexpectedly strong" December jobs report, and the 10-year yield fell back after the December consumer price figures were published on January 15, 2025. Part of the selling side of note futures belongs to the basis trade, in which a fund buys a note with borrowed money and sells a futures contract on it, earning the gap between the two prices rather than gaining from the direction of yields. Federal Reserve staff note that such short positions may be held "for a variety of reasons, including directional positioning or other relative value trades in addition to basis trades", and how much of the 2024 record was a basis trade, the data cannot show; so the essay does not say whether those sellers profited from the rise in yields. After the yield high of January 13, the net short grew again, to 881,374 contracts on March 18, while the yield fell to 4.16% on March 3. So in bonds only the first part of the oil sequence repeated. The minutes of the Federal Reserve's rate-setting committee record that yields rose on "stronger-than-expected data releases" and on communications read as signaling "a more gradual pace of policy easing", that is, slower cuts in interest rates. The price of the 10-year Treasury note future follows the notes deliverable into it, not the 10-year yield one for one, so the yield (DGS10) is used only to show direction. The contract is the standard 10-year note (CFTC code 043602); the Ultra 10-year note is a separate contract (043607) and is not included. For financial futures the CFTC also publishes a more detailed report (Traders in Financial Futures), which Federal Reserve studies use; its "leveraged funds" do not map one for one onto the legacy "non-commercial" group. The Federal Reserve statements quoted cover windows that start in mid-September 2024, not on October 1. The December jobs report was published on January 10, 2025 and the December consumer price index on January 15, 2025; on our yield series the 10-year yield peaked at 4.79% on January 13 and moved from 4.78% to 4.66% on January 15. The net short was 672,446 contracts (58.8% of the record) in the report of January 7, 2025, whose week contains the yield high, and 567,935 (49.6%) in the report of January 14.

Late reports.

Readers at the time saw the last part of the 2018 fall in the bet only later: the reports stopped in late December during a shutdown of the US government, and the January low in the bet became public only in February, about a month late; Figure 2 marks the late reports with a grey band. The report published on December 21, 2018 was the last before the shutdown. The CFTC then published the delayed reports in order, starting on February 1, 2019 with the report for Monday, December 24, 2018, and returned to its normal schedule with the report published on March 8, 2019. The report for January 8, 2019 appeared between February 8 and 12. The reports for December 24 and 31, 2018 describe Monday positions, because Christmas and New Year's Day fell on Tuesdays. Publication dates other than the CFTC's announced schedule are taken from file timestamps and are an upper bound.

Studies on timing.

Sanders, Boris and Manfredo (2004) used weekly CFTC positions in crude oil, gasoline, heating oil and natural gas; Büyükşahin and Harris used non-public daily CFTC data from July 2000 to March 2009 (CFTC working paper). Both are cited for the years they cover and are not evidence about 2018.

The 2018 rise.

The EIA wrote mostly about crude oil in general or about Brent, the international benchmark, so its reasons describe oil prices broadly, not WTI alone. In May 2018 it said that prices had "probably been driven higher" by "falling global oil inventories, heightened market perceptions of geopolitical risks, and strong global economic growth signals". Its October outlook named WTI as well (quoted in the text).

The 2018 fall.

Besides the reasons in the text, the EIA's review of the year named production in the United States, Russia and Saudi Arabia at or near record highs. The price fell 44.3% from its high of October 3 to its low of December 24 (front-month close).

Same time, not cause.

The figures show positions and prices in the same weeks. Nothing in the data shows that the falling bet caused or deepened the fall in price.

Revisions.

The CFTC can revise past rows; one weekly change reported in the press in 2018 as 18,365 contracts appears in today's data as 18,367.

What we left out, and why

Gold: the 2018 case was not a record (the net short of October 2018 was the most negative since 2001, not since 1986), and for 2020 we have positions but no verified price series, so gold has no beat. The COT index (a position's place between its own high and low over a chosen window): it is our formula, not a CFTC measure, and the argument does not need it. Why speculators' net bet fell between February and October 2018 while the price rose: we found no source for 2018; the EIA's outlooks mention fund positions only in May 2018, while they were still growing. Why the price turned on October 3, 2018 rather than another day: the EIA explains the October decline, not its first day. How much of the 2024 bond record was a basis trade: the data cannot separate it. Any general rule that turning points cluster near records: eight runs in oil and one in bonds are not enough to state one. The managed-money and futures-and-options figures quoted in the press: they measure a different group in a different report (see The report).

References

1. U.S. Commodity Futures Trading Commission. Public Reporting Environment: Legacy Futures-Only (6dca-aqww), Disaggregated Futures-Only (72hh-3qpy), Disaggregated Futures-and-Options Combined (kh3c-gbw2); codes 067651, 067411, 043602. publicreporting.cftc.gov (legacy) · publicreporting.cftc.gov (disaggregated combined)
2. U.S. Commodity Futures Trading Commission. Historical Compressed files for 2018.
3. U.S. Commodity Futures Trading Commission. About the COT Reports; Explanatory Notes; Disaggregated Explanatory Notes; Traders in Financial Futures — Explanatory Notes; Glossary (Offset, Roll-Over, Net Position). cftc.gov
4. U.S. Commodity Futures Trading Commission. COT Historical Special Announcements (Dec 21 and 22, 2018; Jan 30, 2019) and press release 7864-19 (Jan 29, 2019). cftc.gov
5. U.S. Energy Information Administration. Series RCLC1, RCLC2 and RWTC; Futures Prices — Definitions. eia.gov (RCLC1) · eia.gov (RWTC)
6. U.S. Energy Information Administration. Today in Energy 19451 (Jan 6, 2015), 28472 (Oct 24, 2016), 31952 (Jul 7, 2017), 32032 (Jul 13, 2017), 36152 (May 10, 2018), 37832 (Jan 2, 2019), 37852 (Jan 3, 2019).
7. U.S. Energy Information Administration. Short-Term Energy Outlook, April 2017, October 2018 and November 2018.
8. Federal Reserve Bank of St. Louis. FRED DGS10. fred.stlouisfed.org
9. U.S. Bureau of Labor Statistics. The Employment Situation — December 2024 (Jan 10, 2025) and Consumer Price Index — December 2024 (Jan 15, 2025).
10. Board of Governors of the Federal Reserve System. FOMC minutes of Nov 6–7, 2024 and Jan 28–29, 2025; Financial Stability Report, November 2024; Glicoes, Iorio, Monin and Petrasek, FEDS Notes, March 8, 2024.
11. Sanders, Boris and Manfredo, Energy Economics 26(3), 2004.
12. Büyükşahin and Harris, "The Role of Speculators in the Crude Oil Futures Market", CFTC working paper.

@misc{tradiens2026herdsledger,
  author       = {{Tradiens Research}},
  title        = {When Speculators All Bet One Way},
  year         = {2026},
  version      = {2.0},
  howpublished = {Tradiens Data Essays},
  url          = {https://tradiens.com/essays/herds-ledger}
}