COT and positioning analytics glossary
Definitions for reading CFTC positioning data, market maps and Futures Flow reports. These terms describe data and analytical methods; they do not constitute trading recommendations.
Reviewed
Commitments of Traders (COT)
The CFTC report of aggregate open positions broken into reporting categories. Check observation date and report format before comparing values.
Open interest
Outstanding contracts that remain open at the observation date. Every contract has both a long and a short side. Open interest differs from trading volume.
Net position
Long contracts minus short contracts for a defined participant category. A negative net value indicates more short than long contracts in that category.
Commercial
A Legacy COT reporting classification commonly associated with business hedging activity. It is not synonymous with informed or profitable trading.
Non-commercial
A Legacy category for reportable traders not classified as commercial. Large speculative positions may remain extreme during sustained trends.
Non-reportable
The difference between total open interest and reportable positions. It is an aggregate residual, not a direct survey of retail accounts.
Managed money
A Disaggregated participant category covering registered commodity trading advisors, commodity pool operators and other identified money managers.
Swap dealer
A Disaggregated category associated with entities dealing primarily in swaps. Futures positions can reflect hedging of swap exposure.
Asset manager
A TFF reporting category that includes institutional investment managers. Its exposure may reflect portfolio mandates rather than short-term price views.
Leveraged funds
A TFF category including hedge funds and other money managers whose strategies may use leverage. Category positions aggregate differing strategies.
Percentile rank
The share of sampled observations at or below a measured value. Always identify the input series, lookback and ranking convention.
Z-score
A value expressed relative to a sample mean and standard deviation. It normalizes scale; it does not turn a historical distribution into a forecast.
PCA
Principal component analysis transforms correlated features into components that explain sample variance. Component meaning depends on the input features and fitted loadings.
UMAP
A nonlinear dimensionality-reduction method used to visualize high-dimensional relationships. Local neighborhoods can be informative while global distances are distorted.
K-means
An algorithm that groups observations around a chosen number of centroids. Cluster numbers are identifiers and do not have permanent economic meanings.
Positioning regime
A descriptive interpretation of recurring exposure patterns across markets or time. Risk labels require context and should not be inferred from cluster membership alone.
Correlation
A statistic measuring association between specified series over a sample. Positioning correlation differs from correlation between price returns.
RAG
Retrieval-augmented generation supplies retrieved context to a language model. Retrieval can improve grounding but does not eliminate generation errors.
Observation date
The date the reported positions describe. It generally precedes the release date and should remain attached to historical analysis.
Release date
The date the CFTC makes a report public. Holidays and disruptions can alter the usual schedule.