The Elasticity-Reversal (ER) system, also referred to as the MA Flow model, is a medium-frequency FX strategy that detects and trades accumulated capital flow by measuring the divergence between FX pair returns and their equity driver. The signal is constructed from the cumulative beta-adjusted residual of each FX pair against the S&P 500 futures contract, with entry timing determined by an exponential moving average crossover on that residual. The system operates on 1-minute data during US trading hours across 10 currency pairs, with regime-conditional entry gating validated through Monte Carlo simulation. A five-year portfolio backtest (2021--2026) across all pairs produced a Sharpe ratio of 4.18 on fixed-size allocation, with statistical significance confirmed by permutation testing on the live driver period.
FX pairs, particularly emerging market and commodity-linked currencies, maintain a measurable relationship with equity markets through risk appetite and capital flow channels. When equities rise, risk-on currencies (AUD, NZD, MXN) tend to appreciate while safe havens (JPY, CHF) depreciate, and vice versa. This relationship is captured by a rolling beta coefficient.
However, the relationship is not constant. When the FX pair's return deviates persistently from what the equity driver predicts --- when the cumulative residual accumulates in one direction --- it signals that autonomous FX flow is occurring: capital movement driven by factors independent of the equity driver. These accumulated flows tend to reverse as positioning becomes crowded and mean-reverts.
The ER system detects this accumulation by tracking the difference between fast and slow exponential moving averages of the cumulative residual. When the fast EMA diverges sufficiently from the slow EMA, the system enters a position in the direction of the flow, expecting reversion.
For each currency pair, a rolling beta is computed against the equity driver using 720 bars of 1-minute data during US session hours (14:00--20:00 UTC):
The beta-adjusted residual at each bar represents the FX return unexplained by the equity driver:
Two exponential moving averages are applied to the cumulative residual:
Each pair has individually validated parameters from walk-forward optimisation:
| Pair | Fast EMA | Slow EMA | Entry Z | Flip Z |
|---|---|---|---|---|
| USDMXN | 15 | 60 | 1.5 | 2.0 |
| USDJPY | 30 | 120 | 0.5 | 1.0 |
| GBPUSD | 15 | 90 | 1.5 | 1.5 |
| EURUSD | 15 | 60 | 1.0 | 1.0 |
| AUDUSD | 20 | 60 | 1.0 | 1.0 |
| USDCAD | 15 | 60 | 1.0 | 1.0 |
| USDPLN | 30 | 60 | 1.5 | 1.5 |
| USDCHF | 20 | 60 | 1.0 | 1.0 |
| USDCNH | 20 | 60 | 1.0 | 1.0 |
| NZDUSD | 15 | 60 | 1.0 | 1.0 |
The variation in parameters reflects genuine differences in how each pair responds to equity driver dynamics. USDJPY, a primary safe haven, requires wider EMA windows (30/120) and a lower entry threshold (0.5) to capture its slower, more persistent flow patterns. USDMXN, with higher carry and volatility, uses tighter windows (15/60) but a higher entry threshold (1.5) to filter noise.
Each position is protected by a 5x ATR initial stop, providing wide room for the thesis to develop. Once unrealised profit reaches 1R (the initial risk amount), a trailing stop activates at 3x ATR from the best price achieved. The trailing stop only tightens --- it never widens --- locking in gains as the position moves in favour.
When the signal z-score reverses sharply past the flip threshold in the opposite direction while a position is held, the system exits and immediately re-enters in the new direction. This captures turning points efficiently, as the same flow dynamic that closes one trade often initiates the next.
Across the five-year backtest, exits distribute as follows:
The win rate of approximately 46% combined with an average win roughly 1.6 times the average loss produces positive expectancy through asymmetric payoffs.
The system uses a hybrid equity driver: S&P 500 CFD data (SPX_DUKA) provides continuous coverage from 2021 through November 2025, and ES futures (via Interactive Brokers) provide the live driver from December 2025 onward. This approach was validated through a driver comparison study on the four-month overlap period.
On the December 2025 to March 2026 overlap period, ES performance matched or exceeded SPX_DUKA on seven of nine testable pairs. Total portfolio P&L on the overlap: $365k using ES versus $327k using SPX_DUKA. The ES driver produces somewhat fewer trades due to coarser tick granularity at 1-minute resolution, but with no degradation in signal quality.
Each pair's volatility environment is classified hourly using a smoothed ATR percentile rank into four levels: QUIET, NORMAL, ELEVATED, and STRESS. Additionally, the direction of volatility (falling, flat, rising) is tracked via the slope of the smoothed ATR series.
For each combination of pair, regime level, and direction, the system's historical trades in that slice are subjected to Monte Carlo permutation testing. The test shuffles trade outcomes across 10,000 iterations to estimate the probability distribution of returns under the null hypothesis of no edge. Slices are classified as:
This produces a deployable configuration specifying, for each pair-regime-direction combination, whether to trade and at what position size multiplier.
On fixed-size allocation with $250,000 initial capital and a maximum of 10 concurrent positions:
A Monte Carlo permutation test on the ES-period trades (December 2025 to April 2026, 869 trades over 61 trading days) confirms the edge is statistically significant:
The ER system is fundamentally independent from the Fair Value system. FV trades hourly OLS residuals from multi-driver regression; ER trades 1-minute beta-adjusted cumulative flow against a single equity driver. The signals are constructed from different data, different timeframes, and different mathematical frameworks. Daily return correlation between the two systems is near zero, making them highly complementary in portfolio allocation.