FX Engineer

Systematic research on retail FX positioning. 1.9M hourly observations. 28 pairs. 12 years. We provide context. Markets decide.

[FX Engineer @FX_Engineer]

New research: RP-004
We tested 75,864 SSI extreme events against a random baseline.
The contrarian price path advantage is zero. A coin flip at every horizon, every z-score magnitude, every session, every era.
But SSI extremes do predict something. Just not direction.
Full analysis: https://t.co/iyD0hfhR2L

[FX Engineer @FX_Engineer]

CHF hits 2nd percentile leveraged fund positioning. Net -9,949 contracts. Z-score -2.88. https://t.co/XjmibAl0Qm

[FX Engineer @FX_Engineer]

CADCHF reaches 96.3% long at the 91st percentile. Retail traders show the highest long crowding this week. Price moves up 0.1% against the dominant position. https://t.co/XjmibAl0Qm

[FX Engineer @FX_Engineer]

Leveraged funds hold 5th percentile short JPY. Net -88,063 contracts. Largest weekly reduction among FX at -17,555 contracts. https://t.co/XjmibAl0Qm

[FX Engineer @FX_Engineer]

Studies that exclude session-timing controls on the grounds that ‘markets have changed’ are dismissing a source of variation that has, empirically, remained highly stable. Omitting session structure from a multi-year backtest is not a neutral choice.

[FX Engineer @FX_Engineer]

Free subscribers receive the Cornerstone Research Reports — the foundational empirical work, free permanently. No predictions. No signals. Positioning data, historical context, and rigorous framing.

[FX Engineer @FX_Engineer]

The 9-16h plateau may correspond to the overnight window. Features like days-to-month-end and 20-day realized vol predict tomorrow’s volatility regime rather than what happens in any particular hour. FX vol prediction has two modes: intraday and regime.

[FX Engineer @FX_Engineer]

Execution timing decisions made without reference to session structure carry avoidable cost. The 2.3x ratio between peak and trough hours is large enough to be operationally significant for any participant sensitive to spread or range.

[FX Engineer @FX_Engineer]

The test period — January 2024 through February 2026 — contains the documented SSI edge collapse toward 50%. That period is locked and withheld from all analysis presented here.

[FX Engineer @FX_Engineer]

Transfer entropy runs from price to SSI at 3-4x the reverse direction. Retail traders respond to price. Price does not respond to retail traders.

https://t.co/f74N7fRtub

[FX Engineer @FX_Engineer]

We tested whether a single multi-horizon model could replace 24 dedicated models. It cannot. Horizon became the dominant feature — the model learned to predict the degradation curve itself rather than the genuine per-horizon signal. Forcing one model to learn microstructure and macro calendar simultaneously dilutes both.

https://t.co/zINFR9C5d7

[FX Engineer @FX_Engineer]

CHFJPY sits at the first percentile. Retail traders hold a 3 percent long position. This reading shows more shorts than 99 percent of all observations since 2002. The pair has risen 0.2 percent, or 49 pips, this week. https://t.co/XjmibAl0Qm

[FX Engineer @FX_Engineer]

NZDCHF reaches 96.6 percent long. This places it in the 89th percentile. Retail traders hold the position for 14 days. Price rises 0.3 percent this week against the crowd.

[FX Engineer @FX_Engineer]

AUDCHF flipped to 91% long just 8 hours ago. NZDCHF reached 95% long in the same window. Six other pairs shifted bias ahead of the RBA decision. https://t.co/XjmibAl0Qm

[FX Engineer @FX_Engineer]

Leveraged funds added +235,094 contracts short 5Y this week. Current net position -2,071,353 contracts. Largest single-week shift across all markets tracked.

[FX Engineer @FX_Engineer]

[COT Brief]
Issue #1: cross-asset positioning across 22 markets as of 2026-06-09. Data tables only.

Full analysis: https://t.co/vQhHNwJHeY

#cot #positioning #weekly

[FX Engineer @FX_Engineer]

The Friday premium likely reflects end-of-week position squaring across institutional desks. The Monday discount reflects gradual re-accumulation of risk after the weekend gap.

[FX Engineer @FX_Engineer]

1,936,174 hourly observations. 28 currency pairs. 12 years. At this scale, any null result is not attributable to inadequate power. If the effect existed, we would have found it.

[FX Engineer @FX_Engineer]

SSI is extraordinarily persistent — mean 1-hour autocorrelation of 0.989. The crowd barely moves from hour to hour. But when it does reach extremes, z-scores exceed 1.5 standard deviations 25% of the time.

[FX Engineer @FX_Engineer]

Test AUC meets or exceeds OOF AUC at all 24 horizons. No overfitting across 24 independently trained models on 2.7 million observations.

[FX Engineer @FX_Engineer]

Friday carries the highest daily average range at 19.26 pips. Monday is the quietest at 16.84 pips. The day-of-week effect exists but is secondary to session timing — 1.14x vs 2.3x. Prioritize hour-of-day over day-of-week.