Measuring execution speed with nothing but timestamps

[Amanhasnoname]

Brokers love quoting execution speeds like 30 milliseconds. Those numbers are measured inside their data centre under ideal conditions, and they say nothing about the experience from a home connection.

Measuring the real figure is simple. Most platforms log the time an order is sent and the time the fill is confirmed. In MT4 and MT5 the journal and experts tabs show timestamps, though with limited precision. An expert advisor can record the millisecond tick count before and after the order call and write the difference to a file.

Collect a few hundred samples across different hours. Expect a median that looks acceptable, and a tail that looks ugly, with some fills taking over a second.

The tail is what matters. Those slow fills typically cluster around news, rollover, and fast moves, which is exactly when scalpers want clean execution.

It also helps to log the account type and server name next to each batch of results. A broker that quietly moves clients between servers can change the whole latency profile, and without those labels the change shows up as a mystery in the data months later.

Once there's data, compare it between brokers or between VPS locations. Anyone logging fill times, and what does the 95th percentile look like?