Latency simulation changes the result of fast strategies

[Amanhasnoname]

Backtests assume an order is placed the instant the signal appears. In real trading, there is a delay from signal to fill, often 100 to 500 milliseconds or more.

For slower strategies, that delay is irrelevant. For scalpers entering on tick-level triggers, it can be the entire edge. By the time the order arrives, the price has moved.

Simulating latency is straightforward in custom backtests. Shift each entry by a fixed delay and use the price at that later time. Running the test with delays of 0, 200, and 500 milliseconds shows how sensitive the strategy is.

Many apparent edges vanish with even a small delay. Those that survive are more likely to work live.

It also helps with infrastructure decisions. If the strategy dies at 300 ms of delay, investing in a fast VPS makes sense. If it survives at 1 second, the extra spend may be unnecessary.

How sensitive is your strategy to execution delay, and has that been tested?