Page 1 of 1
Server clock drift and order timestamps: why your journal lies
Posted: Mon Sep 14, 2026 11:15 pm
by LondonScalper
Server clock drift and order timestamps: why your journal lies.
I had tickets tagged outside a news blackout that were inside it once I aligned platform time, OS time, and statement stamps. Drift of one to two minutes is enough to corrupt event studies and "reject at open" stats.
Habits
1. NTP sync before session.
2. Note platform vs local offset in the daily header.
3. Prefer platform timestamps in the journal when available.
4. For disputes, screenshots beat reconstructed memory.
If your reviews feel inconsistent around data releases, check clocks before checking strategy.
Have you caught a journal conclusion that was really a timestamp problem?
Broker statements, platform logs, and your notes should be reconcilable to the minute on event days. If they are not, fix tooling before you fix strategy. Many "mystery fills" are clock mysteries wearing a strategy costume.
I am interested in how others on this desk handle the same problem without turning it into folklore. Concrete rules and log fields beat slogans. If you have a version of "Server clock drift and order timestamps: why your journal lies" that survived contact with live sessions, what does the rule look like on a sticky note -- and what did you try that failed?
Re: Server clock drift and order timestamps: why your journal lies
Posted: Sat Oct 03, 2026 2:24 am
by PropScalpDesk
Yes, though mine wasn't drift. A month of "CPI blackout" stats was wrong because the calendar export I was matching against had been saved in New York time while the platform history was in server time. Nothing looked obviously broken, every timestamp was plausible, and the conclusion was still nonsense until I lined up one known release by hand.
So my version of your point 2 is a field in the daily header for each clock I use: platform server time, local PC time and calendar time zone, all written as an offset in seconds or hours against a reference checked at session start. It takes a minute and catches both the drift and the dumb mistakes.
If you run anything on a VPS, check its clock separately. Mine had its time sync disabled after a provider image update and slowly ran about forty seconds slow before I noticed in the logs. For news filters measured in minutes that's survivable; for "seconds after the release" studies it isn't.
Re: Server clock drift and order timestamps: why your journal lies
Posted: Fri Oct 16, 2026 10:09 pm
by Shadow Trader
Mine was a daylight saving week, which fits PropScalpDesk's calendar story. Europe and the US change clocks on different dates, and this year the gap runs from 25 October to 1 November.
On New York close server time the server is normally two hours ahead of London. For that week it is three, so events matched with a fixed offset land an hour out, and my US data blackout covered an empty hour.
The same thing happens for about three weeks in March, so that is four weeks a year where a fixed offset is quietly wrong. I only caught it because a CPI stamp sat on a flat M1 bar where the spike should have been.
My sticky note version of LondonScalper's point 2 is to find one known release in the M1 bars every Monday and write down the server time of the spike bar. If it is not where the offset says, nothing from that week goes into event stats until it is fixed.
On the VPS I run a sync check at session start and log the offset beside the others. Forty seconds is survivable for minute filters, but for spread stamps in the first ten seconds after a release it decides which side of the print a tick lands on.
Re: Server clock drift and order timestamps: why your journal lies
Posted: Sat Oct 17, 2026 11:54 pm
by Shadow Trader
One more source of drift for anyone who logs from an EA. In MT4 and MT5, TimeCurrent returns the server time of the last tick received, not the current server time, so in a quiet minute it simply stops moving.
On EURUSD at the London open that hardly matters. On EURNZD at 21:30 London I have seen it sit 20 to 40 seconds behind, and my EA was stamping its log with that frozen value.
The fix was to log two fields per event, TimeCurrent and TimeLocal, with the PC clock synced as PropScalpDesk describes. The gap between them gives a rough measure of how stale the feed was at that moment, which is useful in itself.
Manual journals have a different version of the problem. When I compared 30 hand typed entry times against platform history, the median error was about two minutes and six were off by more than five, nearly always rounded towards a round number.
So I no longer type times at all. The journal row gets the ticket number, and a script pulls open and close times from the account history at the end of the day.
LondonScalper, do your statements show partial closes with their own times? Mine merge them into the final close, which hid a few early exits.
Re: Server clock drift and order timestamps: why your journal lies
Posted: Mon Oct 26, 2026 10:10 am
by Shadow Trader
Accurate clocks matter for trading systems. If a machine's time drifts from true time, timestamps in logs become unreliable, time-based rules misfire, and comparisons with exchange or broker data become difficult.
Windows machines synchronize with time servers only periodically by default, and drift of several seconds can accumulate. For strategies that act on events at specific moments, such as news releases, even a one-second error can matter.
Using a dedicated synchronization tool, configured to update frequently against reliable servers, keeps the clock accurate to a few milliseconds. Checking the offset between the local clock and the broker's server time gives a practical measure of alignment. Logging both local and server timestamps with each order reveals discrepancies. When analyzing latency, accurate clocks are essential, since measurements depend on comparing times from different systems. A routine check of time synchronization, part of regular system maintenance, prevents subtle problems that can be difficult to diagnose after the fact.