@fxengineer.bsky.social - FX Engineer

17 May 2026 21:52 +0000

The FX market operates in geographic sessions. London opens. New York opens. Each session has a characteristic volatility signature. The 4h model captures the entire signature. The 1h model sees only a snapshot. That's why 4h wins.

16 May 2026 14:01 +0000

Statistical significance should not be conflated with practical significance. A p-value of 10^-136 sounds earth-shattering. A Cohen’s d of -0.294 means roughly 12% non-overlap between the two distributions. The shift is real. The magnitude is small.

15 May 2026 17:04 +0000

That result is published prominently, not buried. It is the kind of finding that most vendors building on sentiment data prefer not to test, because testing it seriously risks undermining their product. This publication tests it...

15 May 2026 12:03 +0000

The collapse from 81% to 50.3% measures the exact contribution of look-ahead bias. One contaminated input turned a coin flip into what looked like a near-certainty.

14 May 2026 17:01 +0000

A trader who executes at 21:00 UTC faces a market with roughly half the range activity of the same trade at 14:00 UTC. The choice of execution hour has a measurable and repeatable effect on the price-discovery environment.

14 May 2026 12:00 +0000

The source of predictive power rotates completely across the surface. Microstructure at 1h. Session timing at 4h. Calendar effects at 12h. Macro regime at 24h. These are different information sources, not the same signal at different...

13 May 2026 17:03 +0000

Institutional splits sharpen on AUDUSD. Retail 68% short (1st percentile). Leveraged funds +56.6K net long. Asset managers +44.9K net long. Consensus against retail drives 0.4% price gain. https://fxeresearch.substack.com

13 May 2026 12:03 +0000

CADCHF tops crowded longs: 93.3% long (SSI 13.8268), 70th percentile. Biggest 7-day shift tracked for AUDCAD to 36% long (29th percentile). Yet CADCHF price falls 0.9% (54 pips) this week.

12 May 2026 17:03 +0000

JPY crosses spiked RV last week: NZDJPY 55%, USDJPY 53%, AUDJPY 50%, CADJPY 49%, CHFJPY 44%, GBPJPY 39%. All peaked 2026-05-06 05:02. No pairs start week above 30% threshold.

12 May 2026 12:02 +0000

AUDUSD hits 1st percentile short bias. Retail traders 32% long, more short than 99% of readings since 2002. Leveraged funds hold +56.6K net long. Price up 0.4% (27 pips) this week. https://fxeresearch.substack.com

11 May 2026 17:01 +0000

GBPJPY marks most crowded short: 32.2% long (SSI -2.1094). 25th percentile, more short than 75% of readings since 2002. Price rises 0.2% (36 pips) this week.

11 May 2026 12:01 +0000

CHFJPY at 4th percentile. Retail traders 5% long, more short than 96% of readings since 2002. Extreme short crowding marks volatility conditions. A large price move follows 91% of the time, direction unknown. https://fxeresearch.substack.com

10 May 2026 14:01 +0000

Despite near-unit-root persistence, SSI z-scores exceed 1.5 standard deviations in 25% of 2-hour windows. The crowd barely moves — until it does. Then the displacement is fat-tailed.

09 May 2026 14:03 +0000

It forced a different question: not 'does this strategy work' but 'what does the data actually show when you stop trying to trade it and start trying to understand it.' https://fxeresearch.substack.com/p/retail-positioning-what-it-tells-you

09 May 2026 00:41 +0000

Institutions strongly long EUR: Leveraged funds +20,317 contracts. Asset managers +286,391 contracts. Splits from retail shorts on EURUSD (33.4% long). Next COT reveals JPY pain reaction.

09 May 2026 00:41 +0000

At 1 hour, the model relies on what price just did and how tight the market is. At 24 hours, microstructure is irrelevant. The model is driven by month-end proximity — importance 509, nearly double any other feature.

07 May 2026 17:03 +0000

Leveraged funds held -68,497 contracts short JPY into last week. Sharp USDJPY drop from 160 to 157 punished the crowded position. CAD shorts at -64,215 contracts.

07 May 2026 12:02 +0000

CADCHF z_2h at -2.29. Extreme unwind from 91.4% retail long crowding. Thin market amplifies ratios over conviction. NZDJPY z_2h +1.93 approaches extreme longs at 87.1%.

06 May 2026 17:00 +0000

EURCAD sees 88.9% of retail traders short (11.1% long). Most crowded short position this week. SSI -8.03. Crowd built shorts as the pair traded sideways. https://fxeresearch.substack.com

06 May 2026 12:04 +0000

EURJPY SWIFT generated +488.5 pip MFE. The largest single move from a positioning event in recent memory. JPY selloff drove the extraordinary excursion. https://fxeresearch.substack.com

05 May 2026 17:01 +0000

11 positioning events last week. 8 of 11 produced a 15+ pip move (73%). Matches the long-run 91% rate where extremes mark volatility conditions. Average peak excursion hit 62.8 pips.

05 May 2026 12:00 +0000

The 24-hour FX volatility profile is structurally stable across two decades. Hourly pattern correlation: r=0.987 between 2005-2009 and 2020-2024. Electronic trading was expected to homogenize intraday volatility. The data do not support that...

04 May 2026 17:03 +0000

This is the third instance in this research program where a forward-looking data element produced inflated results. The discipline of replacing every proxy with a strictly pre-entry measure is non-negotiable.

04 May 2026 12:02 +0000

In 0 of 7 major pairs does SSI Granger-cause price at p < 0.05. Retail positioning does not lead price in a linear, statistically detectable sense. Zero of seven. https://fxeresearch.substack.com/p/the-anatomy-of-retail-fx-positioning

03 May 2026 14:05 +0000

Eight consecutive horizons (9-16h) produce nearly identical AUC. That’s not what any simple model of information decay would predict. Something maintains constant predictive power from 9 hours out to 16 hours out.

02 May 2026 14:02 +0000

When the crowd panics short, they are reacting to a price decline that may be exhausting. When they pile in long, they are chasing a move that still has room to continue.

01 May 2026 17:04 +0000

Session patterns are driven by institutional coordination norms — when large participants are staffed, when risk limits reset, when key fixings occur — not by the technical cost of trading across time zones. That's why they haven't changed in 20 years.

01 May 2026 12:03 +0000

Earlier internal analysis reported 81% MAE-first rates and 97% favorable resolution. Those findings were wrong. The entry logic included forward-looking data. The corrected figure: 50.3%. We are publishing this correction with the same prominence as the original...

30 Apr 2026 17:01 +0000

SSI extremes do not predict directional price paths. But they do mark elevated volatility conditions. The crowd reaches extreme positioning during volatile markets, and volatile markets produce large moves in both directions. An SSI extreme tells you something is happening. It does not tell you...

30 Apr 2026 12:01 +0000

The central paradox of this dataset: SSI is persistent, episodically extreme, and informationally reactive to price, but it does not feed forward into price. The crowd has a measurable footprint. That footprint does not move the...