Latency in backtests: the delay your strategy will actually experience
Posted: Sat Oct 31, 2026 7:54 pm
A signal fires at a certain tick in a backtest and the order fills at that tick. In real trading, there's a delay between the signal and the fill, typically tens to hundreds of milliseconds.
During that delay, price can move. For a scalper aiming at small targets, a typical delay can erase a good portion of the edge. The faster the strategy's signals decay, the more it matters.
Simulating it is straightforward with tick data. Instead of filling at the signal tick, fill at the first tick after a set delay, such as 200 milliseconds. Run the test with several delay values and plot how performance changes.
A robust strategy loses a little with each additional delay. A fragile one falls off a cliff.
The result also guides infrastructure decisions. If performance drops sharply beyond 100 milliseconds, a VPS near the broker becomes essential. If it barely changes, the money is better spent elsewhere.
How sensitive is your strategy to a 200 millisecond delay?
During that delay, price can move. For a scalper aiming at small targets, a typical delay can erase a good portion of the edge. The faster the strategy's signals decay, the more it matters.
Simulating it is straightforward with tick data. Instead of filling at the signal tick, fill at the first tick after a set delay, such as 200 milliseconds. Run the test with several delay values and plot how performance changes.
A robust strategy loses a little with each additional delay. A fragile one falls off a cliff.
The result also guides infrastructure decisions. If performance drops sharply beyond 100 milliseconds, a VPS near the broker becomes essential. If it barely changes, the money is better spent elsewhere.
How sensitive is your strategy to a 200 millisecond delay?