Volume clocks: measuring time in activity, not minutes

[Amanhasnoname]

A minute during a data release contains much more market activity than a minute at lunchtime. Indicators calculated on fixed time bars treat them equally, which distorts their meaning.

A volume clock, or activity clock, defines bars by a fixed amount of activity, such as a fixed number of ticks or a fixed amount of volatility, instead of fixed time. In active periods, bars form quickly. In quiet periods, slowly.

Indicators based on such bars adapt automatically. A moving average spanning 20 activity bars covers a short time in a busy market and a long time in a quiet one, matching the information content.

Statistical properties of returns also improve. Returns sampled by activity tend to be closer to normal and more homogeneous in variance, which helps modeling.

The drawbacks are practical. Timing for orders and news is based on wall time, and tools for activity-based bars are limited on retail platforms.

Experiments can start in an analysis environment with recorded tick data before any live implementation.

Would a signal calculated on tick-count bars outperform the same signal on time bars in the data available?