Measuring latency yourself: a weekend project

[Shadow Trader]

Brokers advertise execution speed in milliseconds. Measuring it independently is surprisingly doable.

The basic method: send a small order, record the timestamp when the request leaves the machine, and the timestamp when the confirmation arrives. The difference is the round-trip execution time, including network, platform and broker processing. Repeat a few hundred times at different hours and plot the distribution.

Use the same account type and instrument as in real trading, with minimum size so costs stay tiny. A demo account gives a rough idea but may not match live infrastructure.

Look at more than the average. The 95th and 99th percentiles show how bad the slow cases get, and those are the ones that cause missed fills. Compare results from different VPS locations or home versus server.

Record the platform's own timestamps too, since clocks across systems can drift.

A few hours of work turns marketing claims into numbers that can be trusted.

What did measured execution times look like, and how far were they from the advertised ones?