COT Index Explained: The Williams Formula, and Why 26 Weeks Is Too Short

Key takeaways

• The COT Index rescales a trader group's net position onto a 0 to 100 range: 0 is the most bearish it has been in the lookback window, 100 the most bullish.
• The formula is Larry Williams': 100 × (net − lowest net) / (highest net − lowest net).
• The lookback window is the whole argument. On 25 August 2026, gold's Managed Money COT Index read 100.0 on a 26-week window and 61.6 on a 3-year window. Cocoa read 65.4 and 11.9. Same market, same data, different answer.
• We measured it: across 5,304 market-weeks of the 35 major markets, the 26-week Index pinned at 0 or 100 in 19.9% of readings. The 156-week Index pinned in 6.2%.
• The Index tells you where positioning sits in its range. The z-score tells you how unusual it is to be there. You want both.

Introduction

The COT Index is the most quoted number in Commitment of Traders analysis and the least examined. Traders say things like "commercials are at a COT Index of 12, that is a buy" without ever asking the obvious follow-up: twelve out of what, measured over how long?

That question is not pedantic. It is the entire signal. This article covers what the COT Index is, the formula behind it, what the choice of lookback does to the answer on real markets, and how to read the Index next to the z-score without fooling yourself.

What Is the COT Index?

The COT Index expresses a trader group's current net futures position as a percentage of its own historical range. A reading of 100 means this group is the most net long it has been in the entire lookback window. A reading of 0 means the most net short. Fifty is dead centre.

Larry Williams popularised it because he needed to compare positioning across markets of wildly different sizes. A net long of 376,513 contracts in corn and a net long of 10,475 in platinum tell you nothing side by side. Converted to 0 to 100, they are directly comparable.

That is the whole job of the COT Index: normalisation. It turns an unbounded contract count into a bounded, rankable reading you can screen and alert on.

The Formula

COT Index = 100 × (current net − lowest net in window) / (highest net in window − lowest net in window)

Where net = group longs − group shorts for whichever cohort you track (Managed Money, Commercials, Non-Commercials, Leveraged Funds).

A worked example with the real figures. Sugar No. 11, Managed Money, CFTC report dated 25 August 2026, on our 156-week window running from 5 September 2023:

• Lowest net in the window: −248,296 (week of 3 March 2026)
• Highest net in the window: +213,492 (week of 19 September 2023)
• Current net: +207,082

100 × (207,082 − (−248,296)) / (213,492 − (−248,296))
= 100 × 455,378 / 461,788
= 98.6

Two properties fall straight out of the formula, and both matter:

1. It is bounded. The Index cannot exceed 100 or fall below 0. A new record long does not push it to 130, it pins it at 100 and holds there while positioning keeps building. That is the Index's central weakness.
2. Two data points decide everything. Only the window's minimum and maximum enter the calculation. Every observation between them is discarded. One freak week can set the denominator and flatten every reading afterwards.

We Measured What the Window Does

Williams originally used a three-year lookback, then moved to 26 weeks, and 26 weeks is the value most free indicators and spreadsheet templates shipped with and never revisited.

We ran the same Williams formula over the same Managed Money series at three windows, on the 35 major markets, as of the CFTC report dated 25 August 2026. Where the windows disagree most:

Gold is the one to sit with. On a 26-week lookback, Managed Money in gold reads 100.0, the absolute ceiling, the most crowded long the window has ever seen. On a three-year lookback the same position reads 61.6, moderately above average and nowhere near an extreme. A trader acting on the first number is fading a record. A trader acting on the second is not doing anything at all. Both are reading "the COT Index".

Cocoa inverts it: 65.4 on 26 weeks looks unremarkable, 11.9 on three years is close to the floor of the range.

How often the short window saturates

Spot readings could be cherry-picked, so we measured saturation across time. Taking the 35 major markets and the last 156 weekly reports, 5,304 market-week observations:

One reading in five from a 26-week COT Index is stuck against a boundary, conveying only "at least as extreme as anything in the last six months", which after a quiet six months is not a high bar. At three years it is roughly one in sixteen, and when it happens it means something.

The trade-off, stated plainly:

• Short window (26 weeks): responsive, tracks the current regime, saturates constantly, produces many false extremes.
• Long window (156 weeks): stable, survives regime change, a reading of 100 is genuinely rare, but slower to acknowledge a market that has structurally repriced.

There is no universally correct choice. There is a correct disclosure, and most tools do not make it. A COT Index reading without its window is not interpretable. COTInsight computes the Index on 156 weeks and the z-score on 52, and says so on every reading.

COT Index vs Z-Score: Not Redundant

This is the confusion that costs traders the most.

The difference shows up when positioning keeps going. Corn's Index reads 100.0 on every window we tested. It has nothing left to say. The z-score reads +2.19, and if funds add another 100,000 contracts next week the z-score will register it while the Index stays pinned.

The reverse case is more instructive. Live Cattle: Managed Money is net long 57,441 contracts. Net long. Yet the z-score is −2.42, deep in crowded-short territory, and the 156-week Index is 30.9. Nothing is broken. Cattle funds are structurally net long nearly all the time, so the 52-week average net sits far above 57,441. Being long 57,441 is, for this market, an unusually bearish stance, and it ranks in the bottom few percent of every weekly observation since 2006.

Crude oil makes the same point from the other side: Managed Money net −1,495 contracts, net short, with a 156-week COT Index of 94.4. Over three years WTI funds have been much more heavily short much of the time, so barely short is near the top of the range.

The lesson generalises. Absolute direction tells you very little. Position relative to a market's own history tells you almost everything. A number that reads "long" can be bearish.

How to Read It Without Fooling Yourself

Use the conventional thresholds, but know what they are. Above 80 is "crowded long", below 20 "crowded short". That is convention, not statistics. Unlike the z-score, the Index carries no distributional meaning, so 80 corresponds to no particular rarity.

Treat 0 and 100 as censored, not extreme. A pinned Index has stopped conveying information. Switch to the z-score or the raw weekly change to see what is still happening.

Check the cohort. A Commercials Index and a Managed Money Index on the same market move roughly opposite, because the two groups take opposite sides. Sources rarely label it. We compute on Managed Money for commodities and Leveraged Funds for financials, the speculative cohorts, because those are the positions that get liquidated.

Do not use it to time. Positioning can sit at 95 for months. The Index says the crowd is fully committed and the marginal buyer is scarce. It does not say when the unwind starts. We measured how long extremes actually persist here.

Read it with open interest. An Index of 98 on rising open interest means new money is still arriving. The same 98 on falling open interest means the position is already unwinding. Identical reading, opposite implication.

Where COTInsight Fits

Every instrument on COTInsight carries its 156-week COT Index and its 52-week z-score side by side, with the raw net position and the open-interest trend beside both, so the contradictions above are visible instead of hidden. The board covers around 350 markets that report weekly, refreshed as soon as the CFTC file drops.

Current readings sit on the market pages, for example gold, sugar, corn, crude oil and live cattle. The full ranked board is in the app, and pricing is here.

Pro carries the full board with its 9 signal layers and ten years of weekly history, plus CSV export. Ultimate adds twenty years of archive, the historical outcome statistics that tell you what actually followed each z-score bucket, the AI write-up on every flagged market, and the TradingView indicator, whose second panel plots this exact 156-week COT Index on your own chart, so you are not switching between a dashboard and a candle.

Frequently Asked Questions

What is a good COT Index reading? There is no good reading, only a crowded one. Above 80 conventionally marks crowded long, below 20 crowded short. Neither is a trade on its own.

Is the COT Index the same as the COT z-score? No. The Index is a min-max rescale onto 0 to 100 and saturates at the edges. The z-score counts standard deviations from the mean and keeps counting past any extreme. They frequently disagree, and the disagreement is informative.

What lookback should I use? Twenty-six weeks if you want responsiveness and can tolerate frequent false extremes, since one reading in five pins at a boundary. Three years if you want a reading of 100 to mean something. What matters more than the choice is knowing which one your tool used.

Why can the COT Index be high when the group is net short? Because it is measured against that market's own range, not against zero. Where the speculative cohort is usually heavily net short, a small net short sits near the top of the range. Crude oil on 25 August 2026 is a live example: Managed Money net −1,495 contracts, 156-week COT Index 94.4.

Where does the data come from? The CFTC publishes the Commitments of Traders report every Friday at 3:30pm Eastern, reflecting positions as of the preceding Tuesday. The COT Index is derived from that public file. What differs between tools is the cohort, the window, and whether they tell you.