forex-scalping.com

What an execution speed statistic really measures

Brokers love advertising average execution speeds. "Orders filled in 30 milliseconds." Good number. Worth asking: measured from where, to where, and how?

Often the figure is measured inside the broker's own network, from the moment the order reaches the server to the moment it hits the matching engine. It ignores the trip from the client's machine and the trip back. It also excludes rejected orders, which is a convenient omission.

A trader on a home connection in a distant country may see a total round trip ten times longer than the advertised number.

What matters for scalping is the full path: click, network, server, bridge, liquidity provider, and back. The only honest way to know it is to measure it, either with timestamps in a script or by comparing the platform's order time against its fill time over many trades.

Averages also hide tails. A broker with 40 milliseconds on average and occasional two-second stalls is worse for scalping than one with a steady 90.

Has anyone measured the true round-trip time on a live account, and how different was it from the broker's claim?

I did this with a small script that logged the time I sent each order and the time the fill came back, over about two months on a live account. The broker advertised execution under 40 milliseconds. My median round trip from home was around 180, and that was on a decent fibre connection.

The median wasn't the problem though. Roughly one order in fifty took over a second, and almost all of those clustered around news and the London open, which is exactly when I wanted the fill most. That's where my worst slippage came from as well.

Moving to a VPS in the same city as the broker's server brought the median down a lot, but the slow tail didn't disappear completely, which told me some of the delay was on their side. I'd rather know that than trust the brochure number, and it's why I stopped scalping the first minute after releases.