Forward testing in small live size, done properly

[Amanhasnoname]

Skip the extra demo month and run the smallest lot size live for four to eight weeks. The goal is not profit. The goal is measurement: average slippage per entry, spread actually paid, fill and rejection rate, and the gap between what the test says should have happened and what did.

For that to work, record ticks live during the period. Afterwards rerun the backtest over exactly those dates using the recorded data, then compare trade by trade. Sort the differences into buckets: trades the test took that live missed, trades live took that the test missed, and trades both took with different results. Each bucket points at a different cause, from signal timing to rejected orders to cost errors.

Set stop criteria before starting, not after seeing numbers. For example, if live expectancy sits more than 0.3 pip below the rerun, do not scale up until the cause is found.

The size cost is small, but at minimum lots commission per trade can weigh more than at full size on some accounts, so check that before drawing conclusions.

The useful output is the test-versus-live gap on identical dates, not the live profit figure.

How long do you forward test live, and what size of gap made you scrap a system?