How ADX Regime Classification Filters 22% of Losing Trades

ADX as a Regime Gate

ADX below 20 means the market is directionless. It does not matter how strong your signal is. If there is no trend to ride, momentum-based entries will chop you out. S04 uses ADX as a binary gate: below 20, no new entries for momentum signals. Above 20, proceed with the L1 pipeline as normal.

This sounds simplistic because it is. But simplicity is a feature when you are building a system with 43 interacting components. ADX has been around since 1978. It survived every regime change, every market structure evolution, and every quant revolution. That staying power tells you something about its signal quality.

The 22% Filter Effect

In the V7 backtest, ADX filtering removed 22% of trades that would have been losers. Of course, it also removed some winners. The net effect was a win rate improvement from 55% to 59.2% on the trades that survived the filter. The system takes fewer trades but better ones.

The beauty of ADX filtering is its interaction with L1 confidence. High-confidence signals in low-ADX environments are the most dangerous kind: the model is sure about a direction that the market is not committed to. Filtering these out prevents the worst category of losses, high-conviction losers.

Why This Works for FTMO

FTMO does not reward you for trading more. It rewards you for not losing. A system that takes 4,505 trades over 7.5 years with 59.2% win rate is vastly preferable to one that takes 6,000 trades with 54% win rate, even if the total R is similar. Fewer trades means less exposure to gap risk, overnight events, and execution variance. S04's contribution is not adding alpha. It is removing opportunities for the system to hurt itself. In a constrained-loss environment like FTMO, the best trade is often the one you do not take.