Squeeze fuel
Hold 1–2 weeksSpeculators press new shorts into a crowded short while hedgers buy back their hedges. Go long every market listed.
2007–20never seen−0.24
2020–26built on+0.82
Livesince Jul 26−0.83
Signals this print
Four rules are live. Take one position per market, best signal first (S1, then S2, S3, S5). Enter at the Friday release, while the chart still looks broken: waiting for a technical confirmation has historically used up the edge. Each card shows its net return per week in three windows. The first is the one that matters most: 14 years the rules never saw.
Pressing shorts
Hold 2 weeks, exit around 30 OctThis week longs were cut and shorts were added, with the short side dominant. Fade that crowd: go long every market listed.
2007–20never seen−0.06
2020–26built on+0.40
Livesince Jul 26−0.17
Deep-short tilt
Hold while rel ≤ −90, at most 4 weeksSpeculators are at their most short in a year. Tilt long; the edge front-loads, so exit after four weeks in the state.
2007–20never seen−0.48
2020–26built on+0.89
Livesince Jul 26−1.01
Pressure spread
Hold 1 week, re-ranked every printRank every market by combined crowd flow. Long the three the crowds lean hardest against, short the three they pile into. Smallest sleeve, liquid names only.
LongShort
2007–20never seen−0.22
2020–26built on+0.18
Livesince Jul 26−0.49
Tracked, not traded: S4 and S6
S4Unconfirmed short crowdDemoted 4 Jul 2026 after failing the out-of-family test.
S6Cover-rally fadeThe only outright-short idea. Loses in every era tested, including live.
No market qualifies this print.
Every market
Rel is speculator net positioning against its own one-year range: −100 is the most short of the year, +100 the most long. The trail shows the last 52 weeks. Select a row for the flows behind the call.
| Market | Rel | 52 weeks | Speculators | Hedgers | Week | Since | Call |
|---|
Week is the price move over the COT week, into the release. Since runs from the release to the latest close (2026-10-12); a dash means no close has printed since the release yet. LME rows are as of Friday 2026-10-09.
Did it ever work before 2021?
The rules were built on October 2020 to June 2026. On 23 September 2026 they were replayed, unchanged, on every print back to 2007 (24 CFTC commodities, raw prices, the same costs). The bars are gross returns per week, before costs, by year. The shaded years are the window the rules were built on.
Net of costs, % per week
What this means
S1 to S3 are real inside the 2021–26 commodity regime and roughly flat before it. Treat them as a regime bet that pays while conditions like these hold, and size them as one.
The short side (the spread’s short leg and S6) loses in every window, including live. Short calls here are only the hedge leg of the spread.
The one use of this data that survived the 20-year test belongs to another book: breakout longs do better in markets where speculators are already crowded long.
Live record
The rules were frozen on 4 July 2026. Since then every print has been scored on what actually happened, after costs. Nothing here was tuned on these weeks.