The Framework
How we detect, validate, and surface positioning extremes.
Retail crowds. We quantify when they're wrong.
Most sentiment tools show you what retail is doing. That's useful — but it's not enough.
The question isn't "what is positioning?" The question is: when does positioning matter?
After 13 years of data and 3136+ occurrences, we know the answer: when positioning reaches statistical extremes and the crowd is trapped.
Proprietary Positioning Data
Real flow data. Not surveys.
Our patterns are built on actual retail positioning data — real Long% and Short% figures from live trading accounts, updated continuously. This isn't a poll. It's not self-reported. It's flow.
| Metric | What It Measures |
|---|---|
| Long% / Short% | Percentage of retail traders positioned long vs. short |
| Net Position | Directional imbalance of the retail crowd |
| Positioning Extreme | Statistical deviation from normal — how trapped is the crowd? |
34 instruments tracked.
We don't disclose our data source. Proprietary access is part of the edge.
Opportunity Profile
We measure setups, not outcomes.
Every pattern has two excursions: the maximum favorable move (MFE) and the maximum drawdown (MAE). The ratio between them tells you how much opportunity a setup offered relative to its risk — regardless of where price was at the benchmark.
Across 3,136 patterns, the median opportunity ratio is 5.64:1. What you capture is up to you.
Opportunity Ratio Distribution (3,136 patterns)
| Pattern Configuration | Median MFE | % Offering 2:1+ |
|---|---|---|
| SWIFT | 61.2 pips/pts | 76.9% |
| PULSE | 46.5 pips/pts | 74.2% |
We identify when crowds are trapped. The resulting setups historically offer 5.64:1 median opportunity-to-drawdown profiles. What you capture depends on your management — not our benchmark.
The Price Path
Mean-reversion has a signature.
When crowds are trapped, the initial move is often against the pattern direction — a final squeeze before the turn. We've validated this across 2,754 patterns:
| Price Path | % of Patterns | Favorable Rate | Median MFE |
|---|---|---|---|
| Drawdown first, then rally | 83% | 97.4% | 63.5 pips/pts |
| Rally first, then collapse | 17% | 23.7% | 13.8 pips/pts |
The spread: 73.7 percentage points.
When the worst drawdown occurs before the best favorable move, patterns resolve favorably 97.4% of the time. This is the mean-reversion mechanism at work: trapped traders add to losing positions, then capitulate. That's when price moves.
Pattern Detection
When multiple timeframes confirm, we pay attention.
A single extreme can be noise. When independent measures across different timeframes confirm each other — and the crowd is demonstrably trapped — that's structural. That's when we recognize a pattern.
Our detection requires:
- 1Short-term extreme — Positioning is statistically unusual right now
- 2Longer-term confirmation — The extreme is structural, not a blip
- 3Directional agreement — Multiple timeframes point the same way
- 4Fresh setup — The extreme just formed, not stale
When all conditions align, we fade the crowd. They're long? Pattern favors shorts. They're short? Pattern favors longs.
SWIFT / PULSE
Two configurations. Same edge. Different character.
SWIFT
Balanced frequency. Intraday resolution.
1,021 patterns · 87% favorable · +56 pips/pts
PULSE
Highest frequency. Fastest cycle.
2,115 patterns · 83.6% favorable · +45 pips/pts
Conviction Classification
The difference between 82.8% and 86%.
Not all extremes are equal. We evaluate whether recent price action contradicts the crowd's positioning — or aligns with it.
When the crowd is long but price is falling, those traders are trapped. When the crowd is short but price is rising, same dynamic. Trapped traders create urgency. Urgency accelerates mean reversion.
| Conviction | Condition | Favorable Rate |
|---|---|---|
| HIGH | Price contradicts crowd positioning | 86% |
| STANDARD | Price aligns with crowd positioning | 82.8% |
Delta: +3.2 percentage points. Every pattern is labeled with its conviction level.
Per-Pattern Performance Tiers
Not every instrument is equal. We track them separately.
Each instrument earns its tier independently based on historical favorable rate and sample size. A symbol doesn't inherit credibility from the aggregate — it proves itself or stays in validation.
| Tier | Criteria | Patterns |
|---|---|---|
| ULTRA | ≥ 90% favorable rate, significant history | 15 |
| PREMIUM | 80–89% favorable rate, proven track record | 30 |
| STANDARD | 75–79% favorable rate, sufficient history | 6 |
| VALIDATING | Promising, building history | 18 |
69 active configurations across 34 instruments. Tiers reflect actual performance, not marketing.
Below the threshold. Symbols that don't meet our tier thresholds are tracked but not published. Their 64% favorable rate versus ULTRA's 94% demonstrates the filter works — inclusion is earned, not assumed.
13 Years of Data
We publish everything. Including the bad years.
| Year | Patterns | Favorable Rate | MFE:MAE |
|---|---|---|---|
| 2014 | 232 | 93.1% | 8.65:1 |
| 2015 | 238 | 94.1% | 8.57:1 |
| 2016 | 241 | 91.3% | 6.69:1 |
| 2017 | 254 | 89.8% | 5.53:1 |
| 2018 | 228 | 90.4% | 5.96:1 |
| 2019 | 226 | 87.6% | 7.03:1 |
| 2020 | 156 | 79.5% | 5.5:1 |
| 2021 | 208 | 88.9% | 4.56:1 |
| 2022 | 212 | 80.7% | 3.91:1 |
| 2023 | 204 | 82.4% | 4.16:1 |
| 2024 | 338 | 79.9% | 4.6:1 |
| 2025 | 381 | 84% | 8.58:1 |
| 2026 | 218 | 58.3% | 1.65:1 |
Combined: 3136+ patterns, 84.7% favorable rate. t-statistic: 16–19 (highly significant).
Methodology applied consistently since 2014 using historical and live price data.
What's not in the framework.
In earlier versions, we experimented with additional metrics: velocity, exhaustion, regime detection, cross-asset correlation. We removed them.
They didn't survive validation.
What survived validation is what you see. Everything else was noise.
We'd rather give you one thing that's validated than ten things that look impressive but don't hold up.
What You Get
Intelligence, not indicators.
Real-Time Dashboard
Live positioning data, active patterns, conviction levels, and per-pattern performance tiers — all in one interface. No spreadsheets. No manual calculations.
Pattern Monitoring
Hourly updates on active patterns — positioning trajectory, P/L movement, and combined signal strength. Know when conditions are improving or deteriorating.
Pattern Status Updates
Hourly updates on every active pattern — P/L trajectory, z-score movement, conviction shifts, and a 7-state quality assessment (STRONG, FAVORABLE, STABLE, MIXED, WEAK, ADVERSE). Know when conditions are improving or deteriorating — not just entry and exit.
Intelligence Reports
Daily PDF briefing before market open with active patterns, outcomes, and approaching setups. Friday weekly recap. Sunday market open preview. Monthly performance summary. Quarterly deep dive. 6 report types covering the full intelligence cycle.
Research Library
15 numbered quantitative research notes covering conviction dynamics, cross-tab stacking, session timing, MAE/MFE profiles, and more. 5 public, 10 subscriber-only.
Resolution Analytics
Full post-resolution data on every pattern: entry/exit prices, MAE/MFE, P/L in pips/pts, and mid-life signal trajectory. Learn from every outcome.
API Access
Programmatic access to patterns, positioning data, and historical performance for integration into your own systems. See the API Reference →
Tools for traders who've stopped looking for magic.
$499/month | $4,990/year — 2 months free
Get Started →How we report these numbers matters. For a detailed breakdown of our statistical methodology — why we use medians, how we compute ratios, and what we intentionally avoid — see Research Report #014: Statistical Transparency →
Past performance does not guarantee future results. All statistics reflect historical pattern outcomes across observed and live data. Favorable rate measures the percentage of patterns that resolved with positive P/L — it is not a guarantee of profitability. FX Engineer provides positioning intelligence, not trading advice. You are solely responsible for your trading decisions.