forex-scalping.com

Regime changes: the day the system stops matching the market

Volatility, trend persistence, and liquidity vary over time. A scalping system tuned to one regime may stumble when the market moves to another. A shift to lower volatility after an extended active period, or a change in central bank policy, can alter price behavior for months.

Detecting regimes involves tracking indicators like realized volatility, average range, spread levels, and autocorrelation. Compare current values with historical percentiles.

One approach is to test the strategy separately in each regime historically. If it performs well in high volatility and poorly in low, the system can be switched on or off by a volatility filter, or sized according to the regime.

The danger is overfitting to regimes, defining them in hindsight. Use simple, observable criteria that can be applied in real time.

Even a coarse classification, such as three levels of volatility, can improve robustness. The point is acknowledging that one set of parameters will not suit all conditions.

What regime indicators does the system monitor, and has any change in the market prompted adjustments?

The most useful change I made was a very simple volatility check. Every morning I take the average daily range of the last ten days and compare it with the same measure for the last sixty days. If the ten day figure is below about seventy percent of the sixty day figure, I call it a quiet regime and cut my targets and size accordingly.

I only found this after a bad summer on EURUSD. My system had done well in spring when daily ranges were around 85 pips, then the range dropped to the mid 50s and my win rate slid from about 58 percent to just over 50. The entries had not changed, but the market had no follow through, and the targets that used to be reachable were simply too wide.

I kept the rule coarse on purpose, just two states, because anything finer felt like fitting to the past. It has not improved my best months, but it has reduced the length of my bad ones.