Is millisecond latency marketing relevant below discretionary speed?

[LondonScalper]

Millisecond marketing is mostly irrelevant at my discretionary speed.

The decision to click is still measured in seconds for me, not microseconds. A London VPS with a stable 1 to 2 ms ping to the trade server matters more as reliability than as a race. I measure worst ping and disconnects, not bragging-rights averages. If a provider sells magic microseconds but reboots mid-week, they fail my sheet.

Think about the chain: I see a candle, decide, move the mouse, click. That is easily 500 ms to a second on a good day, and much more when I hesitate. Shaving 5 ms off the network leg changes nothing in that chain. What does hurt is a VPS that freezes for three seconds while a stop needs moving, or a home connection that drops during NFP.

Where latency does matter is for EAs reacting to ticks, and for people running stops so tight that a few ms of slip matter. That is a different game from mine.

Below discretionary speed, the costly delays are human hesitation, platform freezes, and news spikes, not the last tenth of a millisecond on a brochure.

[Shadow Trader]

The part of the chain people forget is the remote desktop session itself. If you click through RDP into a London VPS, the screen you're reacting to is already 50 to 200 ms old depending on where you sit, which swamps any difference between a 1 ms and a 3 ms ping.

So I split the jobs. The VPS runs the EA that manages stops and trailing, and I enter manual trades from the terminal on my own machine.

On worst ping, I'd go one step further and log it per minute. My average to the trade server was about 1.4 ms, but the worst minute around a 13:30 US release was over 40 ms, and once the terminal froze for several seconds with the CPU pinned.

That freeze is the failure you describe, and it came from too many charts and indicators on a small plan. Dropping to six charts cured it without changing provider.

The other reason a VPS earns its fee has nothing to do with speed. If your power or home internet is unreliable, the EA keeps protecting open trades when your own machine can't.

Has anyone measured how much their own reaction time changes between a calm hour and a news minute? I suspect mine roughly doubles.

[Shadow Trader]

LondonScalper, I tried answering my own question with a crude test. A sound alert fires at a pre-set level and the terminal journal shows when my order went out, so the gap is my reaction time.

Over three weeks I logged 40 alerts in calm hours and 18 in the first minute after a release. The calm median was 0.7 seconds and the news median 1.6, while the slowest tenth went from 1.4 to 3.2 seconds.

img_13340_r2.png (64.76 KiB) Viewed 6 times

The chart above is an illustration of that split. My guess of roughly double was close for the median, and the slow tail was worse than I expected.

At 3.2 seconds in a release minute, EURUSD can easily be 4 or 5 pips away from where the alert fired. Next to that, shaving two milliseconds off the ping is a rounding error, which backs your point from another direction.

What worked for me was taking the decision out of the release minute. If a level matters around news, I place the pending order beforehand with its stop attached and accept fill slippage instead of my own hesitation.

Do you count hesitation as part of reaction time, or log it separately? The slowest alerts in my sample were all ones where I had doubts about the trade.