A plain-English guide to COT positioning, with corn as the case.

The Commitments of Traders report is the closest thing to a public leaderboard for the futures market. It tells you who owns the trade and how long they have owned it. Most of the value in it shows up when a number gets extreme, which is why one of the first skills a commodities desk teaches is converting a raw net-position number into something comparable, like a percentile or a z-score. This is the explainer on what the report actually is, the math that turns it into something useful, and a worked example using corn, which is the most informative commodity print on the desk right now.

What the CFTC actually publishes

Every Friday at 3:30pm Eastern, the Commodity Futures Trading Commission releases the Commitments of Traders report. The data is a snapshot of open interest as of the close of the prior Tuesday. There is a three-day reporting lag and a one-day publication delay; if a federal holiday lands on Monday, the release moves to Monday.

For each contract, the report breaks open interest into trader categories. The classifications a working desk uses come from the Disaggregated Report:

• Producer / Merchant / Processor / User. The commercial hedge book. Farmers, grain elevators, ethanol plants, food companies. They are usually short the futures (to hedge a long cash position) and their net position is the slowest-moving series in the report.
• Swap Dealers. Banks and dealers running the other side of OTC commodity-index swaps. Their book is largely a reflection of index-rebalancing flows.
• Managed Money. CTAs, macro funds, commodity hedge funds. This is the bucket retail desks watch because it moves with conviction and it is the bucket that gets crowded.
• Other Reportables. Reportable accounts that do not fit the three categories above. Often family offices and proprietary commodity traders.

The net position in any category is just longs minus shorts. A managed-money net long of +200,000 contracts means the bucket as a whole is long 200k more contracts than it is short.

Why the raw number is misleading

A raw net-position figure cannot be read across contracts. Two hundred thousand contracts long in corn is a much busier book than two hundred thousand contracts long in soybean oil, because the contracts differ in size and open interest. The raw number also cannot be read across time: open interest grows and shrinks with market participation, so 200k today is not the same pressure as 200k a decade ago.

The right input to any positioning analysis is the normalised reading. The two normalisations a desk actually uses are the rolling percentile and the rolling z-score.

Percentile and z-score, in five minutes

The percentile rank of today's reading is the share of past observations that are at or below today's number. Compute it against a defined trailing window, usually one or three years. The Bloomberg COT functions default to a one-year window; most third-party data providers default to three.

Percentile tells you "how high is today, historically?" It is bounded between 0 and 100, easy to read, and insensitive to outliers.

The z-score asks a related but different question: "how many standard deviations above the trailing mean is today?" The formula is the one from any statistics textbook:

z = (x − μ) / σ

Where x is the latest reading, μ is the trailing-window mean, and σ is the trailing-window standard deviation. The same number that sits in the 95th percentile might be 2.0 standard deviations rich on a noisy series and 2.6 on a quiet one. Use both, not one.

Five-line worked example, by hand

To make the math concrete, here is a five-week window of corn managed-money net positions (in thousands of contracts). The same procedure scales to the 52- or 156-week windows a real desk uses.

Week	Net (k contracts)	Step
t-4	170	—
t-3	185	—
t-2	198	—
t-1	210	—
t	219	latest

Sum the five values: 982. Divide by 5: μ = 196.4. For each observation, take the squared deviation from the mean: (170-196.4)² = 696.96, then 130.41, 2.56, 184.96, and 510.76. Sum these: 1,525.65. Divide by 5: σ² = 305.13. Take the square root: σ ≈ 17.47. The z-score of the latest reading is (219 − 196.4) / 17.47 ≈ +1.29.

Five observations is not enough for a real read; the sampling error on the standard deviation is huge. But the arithmetic is the arithmetic. Expand the window to 52 or 156 weeks and the same operations apply.

The same calculation, on a year of real corn data

Here is a 52-week trailing chart of managed-money net longs in corn futures and options, in thousands of contracts. The dashed line marks zero; the shaded band is one standard deviation either side of the trailing mean.

Running the same arithmetic against the full year:

• Latest: 219k contracts net long
• 52-week mean (μ): 91k
• 52-week SD (σ): 77k
• Z-score: +1.66
• Percentile rank: 100th
• Range: -84 to 219

Two standard deviations rich and at the top of the trailing one-year range. That is the textbook definition of a crowded long.

Why crowded matters: reflexivity

Above the 90th percentile, positioning is the catalyst. A trade gets to that bucket because the macro story has been one-way; the risk is that the next datapoint going the other direction triggers a stop-cluster unwind that the spot can absorb only by moving multiple percent in a session. The pattern on the chart is the diagnostic: corn made a one-month high on the bullish USDA WASDE adjustments, but the spot price fell anyway over the last two weeks. Long, but falling is the regime that precedes washouts.

The reflexivity is not symmetric. Crowded shorts can wash, but short-side unwinds are usually faster and end with the asset closer to where it started. Crowded longs in commodities have a longer half-life and a worse recovery, because the cash buyer (the commercial book) usually rebuilds slowly.

Today's corn read, in numbers

Three facts the desk should know going into Friday's release:

1. The July 2026 contract closed at 463.25 cents per bushel on 22 May, down from 475.25 cents on 19 May. A twelve-cent drop in three sessions, on no new bearish headline, is exactly the kind of move that suggests positioning is doing the work.
2. The 18 March release showed speculators adding 35,533 contracts to net longs in a single week. That was the inflection week from the year's washed-out base toward the current crowded read.
3. The May WASDE projects 5.6 billion bushels of corn use for ethanol in 2026/27, which is roughly flat year-on-year. The bullish demand story relies almost entirely on the ethanol export book, which is already at a record pace (over one billion gallons exported in MY 2025/26 through April, up 13% YoY).

None of that is an argument that corn must fall. It is an argument that the trade is no longer where the easy carry is. If the next USDA print does not deliver a fresh bullish supply or demand surprise, the long positioning is the variable that moves first.

What this measures, and what it does not

A COT positioning read tells you who owns the trade and how extended that ownership is. It does not tell you the price target, the timing, or the trigger. It tells you which side of the next bad print will be more painful. That is enough to be useful, and it is much less than the chart-on-Twitter version of COT usually promises.

• Classification slippage. Traders sometimes file into a category that does not match their economic exposure (the Form 40 self-classification is imperfect). The CFTC has cleaned this up materially since the financial crisis, but it is not zero.
• Snapshot, not flow. The report is a Tuesday close-of-business position, not a rolling tape. Two weeks of sideways net positioning can mask big intra-week swings that cancel out.
• Window matters. A 52-week percentile and a three-year percentile can give different reads. Use both when they disagree.
• Survivorship. The trader population in a contract changes over time. Today's managed-money bucket in corn is not the same set of funds as the bucket five years ago.

Related reading

• The cross-asset positioning framework, applied to EUR/USD: EUR positioning has caught up to the dollar's slide
• The correlation toolkit that pairs with COT: A plain-English guide to currency correlations
• The dollar story shaping the cross-asset board: A new Chair, a record-low Michigan print, and the dollar still hasn't moved