Timestamp resolution: when a second is too coarse

[Fairman]

Timestamp resolution: when a second is too coarse

Many platforms record time to the second. For a trade lasting minutes, that is fine. For a trade lasting a few seconds, a second is a huge unit.

Within one second, dozens of ticks can arrive. If the data only says that they happened at 10:02:15, their order is unknown, and a test must guess. The same problem hits fill logs: two events within the same second cannot be sequenced.

Millisecond timestamps are the standard in serious research. They make it possible to measure latency, slippage relative to the quote at the moment of the order, and the true sequence of events around news.

Some platforms offer higher resolution data, and many data vendors do. Custom logging in an expert advisor can capture its own timestamps with millisecond accuracy.

Anyone serious about studying short-term execution benefits from recording the time of every click and every response in milliseconds.

Even a simple log with millisecond stamps of order sent, order confirmed, and quote at each moment unlocks analysis that second-level data cannot. Slippage by latency bucket, for instance, shows how much each extra 50 milliseconds costs.

What timestamp precision does the trade journal use, and is it enough for the trades being taken?