education, not advice, three honest workflows
How traders use this, without pretending it predicts.
Our own backtest says it plainly: a positioning extreme alone is a risk state, not a countdown. Most comparable setups did not turn. So the honest use of this site is not “trade the signal”, it is knowing where the crowd is, what has happened from here before, and what it cost to be wrong. Three ways readers actually put that to work:
workflow 1 · the defensive check
Before any trade: am I joining a crowded side late?
The ledger answers one question in ten seconds: where does your market sit in its own 3-year positioning range, and is it in the crowded zone. Historically, the crowded side still “won” more often than not, but when it lost, the worst-case runs in the files show what late joiners sat through. Knowing you’re entering at the 99th percentile doesn’t forbid the trade; it changes what a sensible stop and size look like.
workflow 2 · the patient fade file
If you’re a contrarian: wait for the tape, not the meter
Fading a crowd early has been the expensive version of being right. Each stretched market’s file lists what actually marked the past turns, first counter-crowd close, 10-day break, the week the specs themselves cut, with how much of the move was typically still left afterwards. The catch is printed with it: those triggers fire all year, so they only mean something while the extreme holds. And the risk panel exists for exactly one sentence: size for the worst case on record, not the median. We measured the alternative: a tight stop (the market’s median shakeout, floored at one ATR) was stopped out of 81% of all cases. The middle is not a stop.
workflow 3 · the ten-minute friday
Replace the raw CFTC release with one readable pass
Friday 21:30 CET the report drops; the Friday Letter and the refreshed ledger say what moved, what’s stretched, and what each file says, in plain language. Between reports, the daily nowcast estimates what the specs have likely done since Tuesday’s cut-off, so you’re not trading Wednesday on Tuesday’s picture. Its own hit rate is graded weekly, in public.
THE 60-SECOND GLOSSARY
Every word of jargon on this site, translated once. Boards link here.
the crowd (big speculators) · the funds betting on direction. When they all lean the same way, that’s the crowd, and you might be part of it.
the hedgers (commercials) · the businesses on the other side: producers and processors. They hedge; they don’t chase.
stretched / at an extreme · the crowd’s bet is in the top 5% of everything it has done in the last 3 years.
fade · a bet against the crowd.
the bar · the move that counts as a real turn: clearly bigger than the market’s normal weekly noise.
the shakeout · the move against you first. Most real turns hurt before they paid.
the fingerprint · the measurable conditions that were present when this exact market actually turned before.
the window · 13 weeks on a clock. Every trade is graded at fixed checkpoints; the exit is the clock, not a feeling.
pooled / the family · counted across all 44 markets. Used only when a market’s own record is too thin to be honest, and always labeled.
Start with the free file, judge it like a skeptic.
This week’s top file is open in full, every week. No account needed.
Common questions
How do I read the COT report?
Read it in three steps: how one-sided the crowd (managed money) is versus its own 3-year range, how long it has been there, and the measured base rate (how often a reading this stretched actually reversed). cotreport prints all three next to each other so you never confuse a crowded reading with a reversal signal.
What does a stretched or crowded COT reading mean?
It means speculative positioning has reached the top or bottom of its own 3-year range. The crowd is heavily on one side. It is a risk state, not a countdown: across 1,206 extremes since 2006, a stretched reading reversed only about 39.7% of the time within eight weeks, barely above the ~38% baseline.
Should I fade a crowded COT position?
Not automatically. cotreport's own tests found that fading every extreme did not beat the baseline once corrected for multiple testing. Use the reading as context and risk management, size for the worst case, and check the market's own base rate before acting. Every decision and its risk are yours.
How often does the COT report update?
Once a week. The CFTC measures positions as of Tuesday's close and publishes the report the following Friday at 15:30 ET. cotreport adds a daily nowcast because the report is already three days old when it lands.