Phase K: Volume filters killed every oversold and breakout signal

We tested 224 candidates pairing oversold, breakout, and cross signals with a volume confirmation filter. Zero graduated. The hypothesis was that requiring above-average volume at entry might filter out low-conviction signals and improve edge. The result: every strategy that cleared backtest gates died in forward testing.

Phase K took seven entry patterns (RSI oversold, Stoch crosses, MACD breakouts, and an OBV-trend variant) and added a simple rule: volume must exceed the 20-bar average to trigger a trade. We ran these on 1h and 4h timeframes with profit targets between 1.0% and 2.5%, testing the full pool across our standard coin universe. All 224 candidates that survived initial backtest validation collapsed when exposed to live market conditions.

This is not subtle decay. A 0.00% pass rate across 224 trials suggests the volume filter systematically damaged the edge these patterns originally had. One explanation: by the time volume spikes enough to confirm a signal, the move has already started and the entry is late. Another: volume surges often accompany volatility that invalidates the mean-reversion or breakout logic the underlying pattern depends on. Either way, the filter that was supposed to increase conviction instead removed tradeable opportunities and left only noise.

What this does not mean: volume is useless in all contexts. Our primitive system encodes volume as a binary gate at entry. We cannot express stateful patterns like "volume rising over three bars" or "volume divergence with price". A more sophisticated encoding might find edge. What this does mean: the specific, simple volume confirmation we tested here does not add value to these entry types on this data.

This phase joins a growing body of evidence that obvious, intuitive filters often fail when tested rigorously. The best-performing strategies in our registry tend to be simple, stateless patterns with tight risk controls, not complex multi-condition setups.

See the full survivor registry at stratproof.com/survivors, or test your own hypothesis at stratproof.com/prove.