persist
Crowding holds — positioning stays in its current band over the coming releases.
likely
Independent research · weekly · CFTC COT positioning
tiltline
A weekly read on whether crowded futures positioning is likely to persist, normalize, or shift — with the evidence attached.
Risk analytics, not a trading signal. It forecasts positioning regimes and uncertainty — no prices, no buy/sell, no guarantees.
01 / What tiltline monitors
Crowding in futures positioning.
Each week the U.S. CFTC publishes the Commitments of Traders report — how heavily each class of trader is positioned, long and short, in every major futures market: crude oil, gold, Treasuries, the grains, the currencies. When many traders crowd onto the same side, that positioning can persist, normalize, or shift. tiltline studies that crowding and forecasts which way it is likely to go.
02 / What it forecasts
The forward read the z-score cannot give you.
For each validated market on the watchlist, tiltline reports the likely positioning regime over the 8–13 week window — told by form and label, never by a buy/sell color.
normalize
Crowding unwinds gradually back toward typical levels.
mixed
shift
Crowding reverses — a regime change in positioning.
tentative
Horizon: the 8–13 week window. The edge is concentrated around two to three months out. There is no short-term edge, so tiltline does not sell a next-week call. Confidence is qualitative (likely / mixed / tentative), and each market is labeled robust, provisional, or coverage-only.
03 / What it is not
Stated plainly, and early.
not a buy/sell signal
not a price prediction
not “alpha”
not a charting tool or terminal
not a newsletter
not an “AI trading signal”
It forecasts positioning regimes and uncertainty — risk analytics, not investment advice. If a section of this site looks like it is trying to excite you, assume it is a draft we have not finished editing down.
04 / Data & sources
Public data, commercially clean, with a dated source trail.
Inputs
- CFTC Commitments of Traders — weekly positioning by trader class, legacy and disaggregated.
- U.S. Treasury covariates — yield-curve and liquidity/stress series used as context.
- Public, commercially clean sources — no scraped or licensed-for-display-only data.
How it is recorded
- Every input is tied to its official source date and release reference.
- Each weekly delivery includes a timestamped forecast record and source manifest.
- Stale or missing data is not published — a market is shown unrated, never guessed.
05 / Horizons & definitions
The defensible band is the long one.
tiltline is deliberately scoped to where the evidence is — and silent where it isn't.
What we forecast over
- 8–13 weeks — the window where the backtest evidence is currently strongest.
- 1–4 weeks: no demonstrated edge. We do not sell a short-term forecast for that window.
Regime definitions
- persist — crowding stays in its current band.
- normalize — crowding unwinds toward typical levels.
- shift — crowding reverses; a regime change.
06 / Validation method & backtest
Backtested against simple alternatives.
We tested the method on 17 years of historical futures-positioning data. For each market and time horizon, we compared it with two simple alternatives: assume positioning stays where it is, or assume it moves back toward its long-run average. tiltline only treats a market as forecastable when it beats the relevant alternative with statistical significance. Otherwise, that market is labeled coverage-only or left out.
Backtest — completed, scoped
What held up
+It was tested across 17 years and multiple market eras, not one lucky stretch.
+The strongest result is on a focused set of markets, concentrated in the 8–13 week window.
+The result is backed by a reproducible validation packet, not a black-box assertion.
+Every shipped forecast carries its source dates, validation status, and caveats.
+Methods that did not beat the simple alternatives were discarded, and the packet says so.
·Current gate: 7 robust and 8 provisional markets, with statistically significant horizons concentrated at 8w and 13w.
Limits — read with equal weight
Read this carefully
×No short-term edge. Inside ~4 weeks, the simple alternatives still win — and we don’t sell a forecast there.
×The edge is modest — an edge, not a crystal ball.
×Not every market. Coverage-only markets are monitored, not marketed as forecast-skill markets.
×The live record is being built weekly; current public evidence is the 17-year backtest.
×Any percentage is a rough guide, not an exact probability.
×Forecasts are risk context, not return guidance.
Live record is being built weekly.
Each weekly forecast is saved at the time it is made. Once enough real-time observations accumulate, this page will add the live track record alongside the historical backtest.
07 / What’s in every audit packet
Open it and check us.
Every weekly delivery ships with the data, the lineage, and the checks — so the reasoning is verifiable end to end, not asserted.
Data sourceCFTC Commitments of Traders (legacy + disaggregated)
CovariatesU.S. Treasury yield-curve / liquidity series
Source recordofficial source date and release reference
Package manifestarchive contents, report hashes, and source references
Validation statustested against last value and long-run average
Forecast recordtimestamped delivery record with source and validation references
delivery-package.zip
▪report.pdfreport
▪email.txtemail
▪report.jsonreport data
▪cot-source.jsonCFTC source
▪treasury-source.jsoncovariates
▪evidence.jsonevidence
▪muginn-analysis.jsonanalysis
▪source-audit.jsonaudit
▪viability-status.jsonstatus
▪delivery-summary.txtsummary
▪package-manifest.jsonmanifest
It reads as evidence, not a lead magnet.
The sample packet is a real, downloadable artifact: the report, email text, source extracts, evidence records, source audit, delivery summary, and package manifest. You can inspect the source dates and delivery contents yourself.
08 / Known limitations
Designed in, not buried.
Forecast & data
- No short-term edge inside ~4 COT releases.
- The edge is modest and market-limited; many covered markets are monitoring-only.
- Percentages are rough guides, not exact probabilities.
Maturity & operations
- Live record is accumulating weekly.
- COT data carries reporting latency.
- Missing data is not published — markets go unrated, not guessed.
09 / How tiltline differs
Against the four things it gets mistaken for.
| Compared with | What it gives you | What tiltline does instead |
|---|---|---|
| Raw COT data | The numbers — you compute meaning yourself. | Forecasts the positioning regime and ships the lineage to check it. |
| Charting platforms / terminals | Visualize the present and the past. | Forecasts validated forward regimes in the 8–13 week window, with confidence and agreement status. |
| Newsletters | An opinion, with no provenance to audit. | A reproducible packet: sources, dates, validation status, delivery manifest. |
| Signal products | Buy/sell calls, rarely with evidence. | Regimes and uncertainty — never direction — fully auditable. |
10 / The pilot
A paid evaluation, kill-gated, sold one to one.
No checkout and no self-serve. A pilot is a scoped, time-boxed paid evaluation — you are buying a real, computed, auditable forecast plus the published backtest evidence, with a gate that protects you if it isn’t useful.
- Scope5–10 markets from your watchlist, focused on validated 8–13 week horizons.
- FormatWeekly PDF + email report and the machine-readable table.
- TermA fixed four-week evaluation.
- CreditPilot fee is creditable toward a later subscription.
- GuaranteeRefunded if the first report isn’t useful to you.
- ClaimsBacktest evidence now; live record accumulating weekly.
Request received.
We saved your request. Denis will reply personally with the current sample audit packet and next steps.
Couldn’t submit the request.
Please try again or email pilot@tiltline.com.
11 / Who makes this
DR
Denis Redozubov
Founder · serial CTO & entrepreneur
Denis is a serial CTO and entrepreneur with a background in rigorous engineering and substantial exposure to finance and markets. He has built and led technical teams, researched market structure for years, and traded options firsthand. tiltline grew out of that overlap: an engineer’s bias for evidence and a market participant’s respect for uncertainty, applied to futures positioning risk.