The Range Trading (RT) system is a mean reversion strategy designed to trade range-bound currency pairs using four complementary signal types: Bollinger Band fade, rolling range fade, RSI overbought/oversold, and session-based range breakout fade. Each pair is assigned its best-performing signal type through walk-forward validation. Trade management parameters --- take profit, stop distance, and maximum hold time --- are dynamically adapted per pair and volatility regime, with each parameter combination validated through expanding-window walk-forward testing and Monte Carlo simulation. The system trades 12 FX pairs with a portfolio Sharpe of 2.08 over five years, with particularly strong performance during ELEVATED volatility regimes where mean reversion is most pronounced.
Many FX pairs spend the majority of their time in range-bound conditions, oscillating around a central value rather than trending persistently. This is particularly true for pairs with strong carry dynamics (USDMXN, USDZAR), pairs anchored by monetary policy convergence (EURUSD, USDCHF), and pairs with seasonal flow patterns (Scandinavian currencies). During these range-bound periods, prices that reach the extremes of their recent distribution tend to revert toward the centre.
The challenge is that what constitutes the range, and how aggressively prices revert within it, varies dramatically with the volatility regime. A Bollinger Band fade that works well in quiet conditions may be catastrophically wrong during a volatility expansion. The RT system addresses this by adapting its trade management to the current regime, trading more aggressively when conditions favour mean reversion and stepping aside when they do not.
The most widely applicable signal. A simple moving average with standard deviation bands defines the range:
Assigned to USDSEK, AUDUSD, GBPUSD, USDCAD, and USDPLN with window and band width parameters validated per pair.
Suited to pairs with well-defined trading ranges. The rolling high and low over a lookback window define the range:
Assigned to USDNOK and USDZAR with a 48-hour lookback window.
A momentum oscillator identifying exhaustion at extremes:
Assigned to USDMXN, USDCHF, and EURUSD. Threshold levels vary by pair --- tighter thresholds (25/75) for less volatile pairs, standard thresholds (30/70) for higher-volatility pairs.
Exploits the tendency of Asian session ranges to contain the day's price action:
Assigned to NZDUSD and USDCNH, pairs with strong Asian session liquidity and predictable range behaviour.
| Pair | Signal | Key Parameters |
|---|---|---|
| USDSEK | Bollinger | window=50, std=2.0 |
| USDNOK | Range | window=48 |
| USDMXN | RSI | 14/30/70 |
| USDCHF | RSI | 14/25/75 |
| AUDUSD | Bollinger | window=50, std=2.0 |
| NZDUSD | Session | Asian 0--8 UTC |
| USDCNH | Session | Asian 0--8 UTC |
| USDZAR | Range | window=48 |
| USDCAD | Bollinger | window=20, std=1.5 |
| GBPUSD | Bollinger | window=50, std=2.0 |
| USDPLN | Bollinger | window=50, std=2.0 |
| EURUSD | RSI | 14/25/75 |
Signal assignment was determined through a comprehensive research phase that tested all four signal types on each pair across multiple parameter combinations and horizons. The assigned signal is the one that produced the highest walk-forward Sharpe ratio for that pair.
Each trade is governed by three parameters that determine its risk/reward profile:
These parameters are not fixed --- they vary by pair and volatility regime. During STRESS regimes, where ATR is three to four times wider than QUIET, the same 2x ATR stop represents a fundamentally different risk. The system adapts:
The parameter sweep uses three expanding walk-forward folds to prevent overfitting:
For each pair and regime level, the 80 parameter combinations (4 TP fractions x 5 stop multipliers x 4 hold times) are evaluated on training data, then tested on the out-of-sample fold.
A parameter combination must pass three gates to be selected:
1. Minimum trades: at least 15 trades in each out-of-sample fold
2. IS-OOS gap: out-of-sample Sharpe must not degrade more than 50% from in-sample
3. Consistency: out-of-sample Sharpe must be positive in at least two of three folds
If no combination passes all gates for a given pair-regime slice, the system falls back to default parameters (TP=0.5, stop=2.0, hold=12).
For each pair-regime-direction combination, historical trades are subjected to 1,000 permutation tests. The observed Sharpe is compared against the distribution of permuted Sharpes to assess statistical significance:
Direction gating further refines the configuration: within each regime level, certain volatility directions (falling, flat, rising) may be blocked entirely if they consistently produce negative returns.
The system's edge varies significantly by volatility regime:
The key insight is that regime-conditional parameters increased the portfolio Sharpe from 1.47 (fixed parameters) to 2.08 --- a 41% improvement from adapting trade management to the environment.
Range trading fills a specific gap in the multi-strategy portfolio. The Fair Value system trades OLS residual mean reversion on hourly data. The Elasticity-Reversal system trades beta-adjusted flow on 1-minute data. Range trading operates on a different signal class entirely --- price oscillation within technical bands --- and is strongest precisely when the other systems are weakest (QUIET regimes with low FV signal activity). The three systems together provide coverage across virtually all market environments.