Brokers and VPS providers advertise latency figures, but they often measure different things.
Common numbers:
Ping: network round-trip time from your computer or VPS to the broker's server. Often 1–5 ms from a nearby VPS, 20–80 ms from home.
Execution time: time from the server receiving the order to confirming the fill. Depends on the broker's systems and liquidity providers.
Total order latency: ping plus execution time plus platform processing.
For most manual scalpers, differences of 20–50 ms in ping matter much less than execution time and slippage. For automated strategies reacting to fast price changes, network latency becomes more important.
MT5 shows ping in the status bar; execution time can be estimated from order timestamps in the journal.
Have you measured both ping and execution time with your broker?
I measured both on my setup last quarter, and the gap surprised me. Ping from the VPS sat around 2 ms, but the journal timestamps showed order to fill averaging 38 ms, with the slowest 5 percent above 120 ms.
Those slow fills clustered around London open and the first minute after data. So the number on the status bar was nearly useless for the moments I care about most.
From home my ping was about 45 ms, and I expected that to be the bigger difference. On market orders during calm hours, fill quality between home and VPS was almost identical, which matches your point about manual scalpers.
One measurement trap is clock drift between the terminal and the server. If the VPS clock is off by even 50 ms, any execution time worked out from mixed timestamps is noise, so I only compare server times with server times.
Has anyone compared execution time on limit orders versus market orders with the same broker? My sample of limits is still too small to say anything useful.
I now have enough limit orders to answer my own question, about 140 over the last two months with the same broker. The execution time on them turned out to be the wrong thing to measure.
Once a limit sits on the server, my ping stops mattering because the order is already there when price arrives. Fills came through at the price every time, so slippage on limits was zero by design.
The cost showed up as missed fills instead. On 31 of the 140, price touched my level exactly and the order was not filled, mostly in the busy first minutes of London.
Stop entries were the opposite case. They trigger on the server and then fill as market orders, and in the minute after data they slipped about 0.4 pips on average against roughly 0.1 in quiet hours.
So my log now tracks three separate numbers by order type, which are fill time for market orders, missed touch rate for limits, and trigger slippage for stops. Lumping them together as latency hid all three.
Does your broker show a reason in the journal when a touched limit stays unfilled? Mine just leaves it pending without comment.
Latency measures how long a message takes to travel. Jitter measures how much that time varies from one message to the next. A connection averaging 20 milliseconds with wild swings between 5 and 150 milliseconds is often worse for scalping than a steady 40 milliseconds.
Inconsistent timing makes execution unpredictable. A strategy calibrated for fast fills may work most of the time and fail occasionally, and those failures tend to coincide with volatile moments when the network is busiest.
Measuring jitter requires repeated pings over time, not a single test. Running a continuous ping for an hour and recording the standard deviation gives a usable figure. Tools such as MTR or similar network diagnostics reveal where along the route delays occur. If jitter is high, switching VPS providers, changing the network route, or choosing a different data center may help. Avoiding wireless connections on the machine running the platform is also a basic but often neglected step.