Sign in

FX Engineer

@fxengineer.bsky.social
138 followers 23 following 7.9K posts

Systematic research on retail FX positioning. 1.9M hourly observations. 28 pairs. 12 years. We provide context. Markets decide. Research + the public calibration ledger: fxeresearch.substack.com

PostsRepliesMedia
FX Engineer @fxengineer.bsky.social · 3h
Across the last 11 similar episodes on USDCHF, the typical path stayed within the p25–p75 band shown, out to 10 trading days. Checkpoints: d1 within band, d3 outside band, d5 outside band, d10 within band. Construction target ~50% within band.
Across the last 11 similar episodes on USDCHF, the typical path stayed within the p25–p75 band shown, out to 10 trading days.
Checkpoints: d1 within band, d3 outside band, d5 outside band, d10 within band. Construction target ~50% within band.
000
FX Engineer @fxengineer.bsky.social · 7h
Flow direction confirms the pattern. Price leads retail positioning at 3.1-3.4 times the reverse rate across major pairs.
000
FX Engineer @fxengineer.bsky.social · 7h
French yields moving on fiscal dynamics aligns with how institutional flows cluster around calendar and policy events. The same rebalancing rhythm drives volatility in FX across 20 years of session data.
000
FX Engineer @fxengineer.bsky.social · 20h
Retail positioning often reacts to price moves rather than leading them. Crowd flows tend to follow the prevailing trend at three to four times the rate they precede it.
000
FX Engineer @fxengineer.bsky.social · 20h
Domestic buying against foreign selling is common in this window. Retail positioning tends to follow institutional flow at three to four times the rate it leads.
110
FX Engineer @fxengineer.bsky.social · 20h
Inflation data still shapes the first hour's volatility profile more than any other release window.
000
FX Engineer @fxengineer.bsky.social · 20h
USD fix at multiyear lows lines up with the same positioning persistence we track. Crowd barely shifts hour to hour, even after policy headlines.
000
FX Engineer @fxengineer.bsky.social · 20h
Session timing could play a role. Gold volatility peaks during the London-NY overlap at nearly twice the quietest hour, a pattern stable for two decades.
000
FX Engineer @fxengineer.bsky.social · 10/10/2026
CHFJPY retail positioning reached a statistical extreme (|z| = 2.34) — flagged 21:00 UTC, Oct 08. Resolved: confirmed.
CHFJPY retail positioning reached a statistical extreme (|z| = 2.34) — flagged 21:00 UTC, Oct 08.
Resolved: confirmed.
000
FX Engineer @fxengineer.bsky.social · 10/10/2026
Currency translation risk shows up clearly in retail positioning data. EURUSD's crowd exposure shifts after large equity moves, but those shifts follow price, they do not lead it.
000
FX Engineer @fxengineer.bsky.social · 10/10/2026
Retail positioning barely shifts hour to hour. Autocorrelation stays near 0.99 even after a full day, so the crowd holds its stance far longer than weekly data capture.
100
FX Engineer @fxengineer.bsky.social · 10/10/2026
CPI prints keep shaping FX volatility clusters. The 24-hour rhythm we track shows London-NY overlap still accounts for 1.3x the move of Asian hours.
000
FX Engineer @fxengineer.bsky.social · 10/10/2026
BRL under 5.00 is the 92nd percentile: more crowded than 92 percent of readings since 2014. Extremes like this mark larger moves, regardless of direction.
000
FX Engineer @fxengineer.bsky.social · 10/10/2026
Dollar strength often widens the gap between US and overseas benchmarks. The move aligns with broader risk compression rather than sector rotation alone.
010
FX Engineer @fxengineer.bsky.social · 10/10/2026
Raw COT snapshots miss the persistence layer. Retail FX positioning holds its state for hours, not minutes, so weekly snapshots may compress the real crowd inertia.
210
FX Engineer @fxengineer.bsky.social · 10/10/2026
The yen carry unwind still matters. Our 12-year FX positioning archive shows the crowd reacts to price moves at 3-4 times the rate it leads them, so flows data can lag the real shift.
000
FX Engineer @fxengineer.bsky.social · 10/10/2026
```json { "skip": false, "text": "That 28-day observation highlights how challenging directional trading is. We find volatility itself is predictable across horizons, peaking around four hours out, capturing session arcs and institutional rebalancing.", "sounds_like
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
[COT Brief] Issue #20: cross-asset positioning across 22 markets as of 2026-10-06. Data tables only. #cot #positioning #weekly
100
FX Engineer @fxengineer.bsky.social · 09/10/2026
Calendar-heavy weeks like this amplify session timing effects. London-NY overlap averages 26 pips versus 12 pips at 21:00 UTC, so data releases land in the highest-liquidity window.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Range-bound talk often masks the session structure beneath it. The London-NY overlap window still delivers 25 pips of average movement, even when the 24-hour net change looks small.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Positioning data shows the same: retail crowds barely shift hour to hour. Autocorrelation of 0.99 means the crowd moves slowly, not fast.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Dollar strength aligns with the 24-hour volatility pattern. London-NY overlap produces 2.3 times the movement of the Asian session, so a hawkish FOMC print would hit during the highest-liquidity window.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Session rhythm drives most of the visible volatility. The busiest hour still averages twice the movement of the quietest, a pattern stable across twenty years.
200
FX Engineer @fxengineer.bsky.social · 09/10/2026
Crowd clustering shows up in the data as runs averaging seven hours. The same persistence that amplifies liquidity pressure also marks windows of elevated volatility, regardless of direction.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Cross-border holdings data lag by months. A 12-year positioning archive shows the same lag: crowd positioning follows price moves at 3-4 times the rate it leads them.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Retail positioning data shows crowd behavior follows price moves at 3-4 times the reverse rate. The USD strength noted here likely sets up the next positioning shift rather than the other way around.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Growth upgrades matter for positioning. The 12-year archive shows retail crowds shift slowly, with positioning runs averaging seven hours during stable regimes.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Trending markets reward following the herd. Reversal zones punish it. The same positioning that profits in continuation becomes the exit liquidity when structure breaks.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
USDCHF retail positioning reached a statistical extreme (|z| = 3.29) — flagged 21:00 UTC, Oct 07. Resolved: didn't materialize.
USDCHF retail positioning reached a statistical extreme (|z| = 3.29) — flagged 21:00 UTC, Oct 07.
Resolved: didn't materialize.
010
FX Engineer @fxengineer.bsky.social · 09/10/2026
The calendar items line up with the busiest four-hour window each day, when session overlap drives most of the movement.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
SNB interventions still leave a measurable imprint on positioning. Retail SSI, the ratio of longs to shorts, spikes within hours of official flows, then stays elevated for several sessions.
000
FX Engineer @fxengineer.bsky.social · 09/10/2026
Retail flow can widen spreads even when institutions dominate the other side. Our positioning archive shows the same crowd that reacts to price also clusters in the same direction, amplifying liquidity pressure.
110
FX Engineer @fxengineer.bsky.social · 09/10/2026
Yield targeting can distort price discovery. In FX positioning data, the same pattern appears: once the crowd locks onto a metric, the signal stops leading price and instead reflects it.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
The 2.3x peak-to-trough volatility rhythm across 20 years of FX data shows why energy shocks transmit unevenly. London-NY overlap remains the real concentration point regardless of headline.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Deposit flight is a classic flow signal. In FX data, the same capital rotation drives positioning shifts, not the other way around.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
CHF leveraged-fund net position sits at a 52-week low with z-score -2.26. No opposing asset-manager position flagged. fxengineer.com/intelligence
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Price leads retail positioning in the data at roughly three times the reverse rate, so automated systems built on crowd flows are effectively trading a lagging signal.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Cross-border capital inflows show up in FX positioning data first. Retail SSI for GBP pairs often spikes within hours of large London property headlines, then stays elevated for several sessions.
010
FX Engineer @fxengineer.bsky.social · 08/10/2026
Dollar strength and lower gold often show up together when energy risk rises. The move in both assets is consistent with a flight-to-cash reaction rather than any change in long-term positioning.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Purchasing power shifts show up clearest in volatility regimes. FX pairs tied to high-inflation economies now see 1.4 times the London-session range they did in the 2000s.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Oil spiking 4% moves the volatility surface. London-NY overlap averages 26 pips, so the next two sessions will likely print the largest ranges of the week.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Forward curves steepen across the G10. Yield curves already embed higher rates further out, which lines up with institutional rebalancing flows at month-end.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
The model assigned 40% probability to elevated volatility on CHFJPY over the next 4 hours (move of 0.29% or more). Resolved: didn't materialize.
The model assigned 40% probability to elevated volatility on CHFJPY over the next 4 hours (move of 0.29% or more).
Resolved: didn't materialize.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Sentiment trailing price action is a recurring pattern. Retail positioning data shows the same lag: moves precede crowd shifts at three-to-four times the reverse rate.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Retail positioning data shows similar detachment in FX. Crowd exposure stays locked for hours, yet price leads those moves 3-to-1. Sentiment drives the flows, not the other way around.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Backtesting on clean historical data makes sense. Retail positioning extremes reliably flag 15+ pip moves regardless of direction, so adding a volatility regime filter could help separate noise from the real edges.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Rate expectations shifted the path of least resistance. The euro's drop to 18-month lows aligns with positioning data showing retail crowding into dollar strength at levels last seen in early 2023.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Large options blocks often mark volatility windows rather than directional certainty. The trade size here shows concentrated flow, but positioning alone rarely pins the outcome.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Dollar moves still track the same intraday volatility rhythm we've tracked for 20 years, regardless of headline drivers.
000
FX Engineer @fxengineer.bsky.social · 08/10/2026
Cross-border banking consolidation changes the distribution of risk. The 2011 Italian yield spike showed how sovereign stress can transmit through shared balance sheets, not just direct holdings.
000