COT reports

Definition

The CFTC’s Commitments of Traders reports publish weekly snapshots of futures positioning by trader category — commercials (hedgers), managed money/large speculators, and small traders — the only regulator-published positioning census in any major market. The classic folklore reads commercials as “smart money” and crowded speculators as a contrarian flag; the academic record is MIXED — category positioning describes market structure reliably and predicts returns unreliably.

How it works / structure

• The data (engine-executable): reported Friday for Tuesday’s positions (a 3-day lag); legacy format (commercial / non-commercial / non-reportable) and disaggregated format (producer, swap dealer, managed money, other); the standard normalization is the COT INDEX — each category’s net position percentiled against its own 1-3 year range.
• The structural read (reliable): who is long what, at what extreme — managed-money nets track trend-follower positioning almost by construction (strategy-futures-trend-following crowding gauge); commercial nets mirror them (hedgers take the other side); extremes describe HOW CROWDED a move is.
• The predictive read (unreliable): Sanders-Irwin-Merrin (2009) and related work found little forecasting value in category positioning for agricultural markets; other studies find modest effects in some markets/periods — the honest summary is “sometimes, somewhere, weakly.”
• Where it’s most watched: agriculture (the native habitat — instrument-agricultural-futures), FX futures (speculative positioning proxy for the spot market — ext-fx), VIX futures (the perennial “record short vol” headline — instrument-vix-futures).

When it applies

Crowding context on futures theses (a trend entry when managed money is at its 95th percentile net-long is a different risk than at the 40th); squeeze-fragility framing (extreme one-sided positioning + reversal catalyst); structure description in macro narratives — always as CONTEXT, never as a standalone signal (the evidence does not support more).

Risk profile & failure modes

• Smart-money folklore: “follow the commercials” as a trading rule fails the documented tests; commercials hedge business exposure — their nets are inventory management, not forecasts.
• Extremes that extend: record positioning regularly becomes more record; contrarian triggers on percentile alone churn — confirmation from price or a dated catalyst is structural, not optional.
• The 3-day lag: positioning as of Tuesday published Friday — event weeks make the snapshot archaeology.
• Category bleed: swap-dealer intermediation blurs who holds the economic exposure; disaggregated format helps, imperfectly.

Evidence & limits

Methodology is CFTC-documented. The mixed predictive record is the honest citation (Sanders-Irwin-Merrin for the null in ags; scattered positives elsewhere) — the platform grades COT as structure-descriptive, prediction-weak, and requires any COT-conditioned strategy to earn its keep in replay.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Managed-money net-long in crude at its 95th percentile will precede a negative 3-month return in most instances this decade (crowding-reversal thesis)” — falsified by the conditional tally.
• “COT-conditioning (skip entries above the 90th positioning percentile) improves trend-strategy replay Sharpe in ags” — falsified by the paired replay.

Cross-references

• The frame: lens-sentiment (positioning vs opinion)
• The crowd it measures: strategy-futures-trend-following
• Native habitats: instrument-agricultural-futures, ext-fx, instrument-vix-futures
• Flow siblings: sent-fund-flows, sent-short-interest