Ask ten traders what “latency” means and most will quote their ping. Ping is real, but it is only one hop out of roughly eight that a price and an order have to travel. A 2 ms ping can still come with a 150 ms fill, and a trader who moves to a faster VPS can be surprised that nothing much changed. The reason is simple: latency in forex trading is a chain, and the slowest link decides what you actually get.
This guide walks the whole chain, hop by hop: where the price is born, how it reaches your screen, what happens to your order on the way back, and which of those delays you can shorten yourself. All timing figures below are illustrative ranges to show orders of magnitude, not measurements of any specific broker.
Two kinds of latency: price latency and execution latency
Before looking at the hops, separate two things that are often lumped together:
- Price latency (also called quote or feed latency) is how old the price on your screen is. If the market moved 40 ms ago and your terminal still shows the previous quote, you are looking at the past.
- Execution latency is the round trip from the moment you (or your software) send an order until the fill confirmation comes back. It includes the network both ways plus everything the broker and its liquidity providers do in between.
They have different causes and different fixes. A trader can have a fresh price feed and slow execution (a broker that holds orders), or fast execution on a stale feed (a broker whose price server throttles ticks). Most “my broker is slow” complaints become much easier to solve once you know which of the two you are dealing with.
The full chain, hop by hop
Here is the path a quote and an order take on a typical retail FX setup with a MetaTrader or cTrader broker. The top row is the price coming to you; the bottom row is your order going back.
1. The liquidity provider sets the price
Banks, non-bank market makers and ECNs stream quotes from their own engines. Their internal latency is tiny, but it is the origin of the “true” price that everything downstream is chasing. You cannot change this hop; you can only be closer to or further from it.
2. The aggregator and the broker’s price server
The broker collects streams from several providers, picks or blends the best bid and ask, applies its markup and pushes the result to its trading platform. This is where a lot of hidden price latency lives. Some setups filter or throttle ticks (for example, sending at most a few updates per second per symbol) to save bandwidth. The result is a feed that looks smooth but lags the market by tens of milliseconds in fast moves.
3. The trade server
The MetaTrader or cTrader server receives the broker’s prices and distributes them to every connected client. A busy server with thousands of clients and many plugins adds its own small but variable delay.
4. The network to you
This is the hop people measure with ping. It is governed mostly by physical distance: light in fibre covers roughly 200 km per millisecond, and real routes are never straight. A VPS in the same data centre as the broker’s server can sit below 1 ms; the same city is a few milliseconds; London to New York is on the order of 30-40 ms one way; a home connection on another continent can easily be 100-200 ms or more.
5. Your terminal and your strategy
The quote now has to be processed by your terminal and whatever logic you run. An overloaded terminal with dozens of charts, heavy indicators or a slow script can add more delay than the network. This is the most underrated hop because it never shows up in a ping test.
6-10. The order goes back
Once you decide to trade, the order crosses the network again (hop 6), waits in the trade server’s queue and passes risk and margin checks (7), goes through the broker’s bridge to a liquidity provider or the broker’s own book, possibly with a deliberate dealing delay (8), may be held by the provider’s last look window and then accepted or rejected (9), and finally the fill report travels all the way back (10). If you want the detail of how bridges route orders, our comparison of FIX API connectivity versus an MT4 bridge covers that hop in depth.
A sample latency budget
Putting rough numbers on each hop makes the picture concrete. Treat the ranges as orders of magnitude; real values vary by broker, server load and time of day.
| Hop | Illustrative range | Main driver | Who controls it |
|---|---|---|---|
| LP price formation | sub-ms to a few ms | Provider engine, venue | Nobody (market) |
| Aggregator + price server | 1-20 ms, more if ticks are throttled | Aggregation logic, tick filtering | Broker |
| Network, server to you | <1 ms (same data centre) to 200+ ms (another continent) | Physical distance, routing | You (where you run) |
| Terminal + strategy | 0.1 ms to 50+ ms | CPU, chart load, code quality | You |
| Network, you to server | Same as the way in | Physical distance, routing | You |
| Server queue + risk checks | 1-30 ms | Server load, plugins | Broker |
| Bridge + dealing delay | 0 to several hundred ms | Execution model, dealer settings | Broker (your choice of broker) |
| LP last look | a few ms to ~200 ms | Provider’s hold window | Liquidity provider |
Two things stand out. First, the network is the largest item only when you are far away. Second, once you are close, almost everything that is left belongs to the broker and its liquidity. The next chart shows that shift.
Jitter: why the average lies
A single latency number hides the most dangerous part: variation. Jitter is how much the delay changes from one message to the next. A connection that averages 15 ms but occasionally takes 90 ms will fill most orders fine and then deliver a bad fill exactly when the market is moving fastest, because that is when servers, bridges and providers are under the most load.
When you test a VPS or a broker, record many round trips and look at the 95th and 99th percentiles, not just the mean. A setup with a slightly higher average but a tight tail is usually better than a fast-on-average one with spikes.
Platform issue or market? How to diagnose a slow fill
When an order fills late or at a worse price, work through these checks before blaming anyone:
- Check your ping at the time. If the terminal’s connection ping was normal, the delay happened after your order reached the server, not on your side of the chain.
- Read the journal timestamps. The gap between “order sent” and “order filled” in the platform journal is your real execution latency. Compare it with a quiet period.
- Look at your own machine. High CPU, dozens of open charts or a heavy script can delay the decision and the send. A clean, dedicated terminal is the cheapest fix there is.
- Note when it happens. Delays only around news releases or the daily rollover usually point to thin liquidity and last-look rejections: a market effect, not a broken setup.
- Check whether it is selective. If only your fast, profitable trades are slow while losing or slow trades fill instantly, that asymmetry is typical of a deliberate dealing delay. Our guide on how brokers flag and slow down fast accounts explains the pattern.
- Compare against a second account. Running the same order at the same moment on another broker or a demo account separates a broker problem from a market event.
Why the chain matters so much for arbitrage
For most traders latency is a cost. For latency arbitrage it is both the opportunity and the cost, and they sit on opposite rows of Figure 1.
The opportunity is price latency at the broker: if hops 2-3 at a broker lag a fast reference feed, its quote is briefly stale. That gap is what latency arbitrage trades. The cost is your own execution latency: the gap is only worth something if your order arrives and is filled before the quote catches up. Every millisecond in hops 6-9 eats into it, and dealing delays or last-look rejections can erase it entirely. That is also why slippage analysis is central to the strategy; we break the numbers down in our guide to beating slippage in latency arbitrage.
The practical consequence: an arbitrage setup must be fast on the hops you control so that the result depends only on the broker-side hops, which you then judge by choosing venues carefully.
How to reduce latency, in the right order
- Move next to the server. Run your terminal on a VPS in the same data centre or city as the broker’s trade server. For forex that usually means London (LD4), New York (NY4) or Tokyo (TY3); our VPS setup guide for LD4, NY4 and TY3 shows how to pick the location and measure it.
- Clean up the terminal. Close unused charts, remove heavy indicators, keep one strategy per terminal where possible and leave CPU headroom.
- Measure the tail, not the average. Log round trips over a full trading week, including news days, and compare percentiles.
- Choose the execution route. Where a broker offers it, direct connectivity can shorten the server-and-bridge hops compared with a standard platform account.
- Judge the broker on hops 7-9. Once your side is fast, persistent slow or selective fills are a property of the broker’s execution model. That is a reason to change venue, not hardware.
Frequently asked questions
What is latency in forex trading?
It is the delay between a market event and your reaction to it being executed. It has two parts: price latency (how old the quote on your screen is) and execution latency (the round trip from sending an order to receiving the fill).
What is a good ping for forex trading?
For manual trading, anything under about 50 ms is comfortable. For automated or speed-sensitive strategies, traders aim for single-digit milliseconds, and for arbitrage below 1-2 ms, which in practice means a VPS in the broker’s data centre.
Does a VPS remove latency completely?
No. A VPS shortens the network hops and gives your terminal a stable machine, but the broker’s server, bridge, dealing settings and the liquidity provider’s last look still add their own delay.
What is last look?
Last look is a short window in which a liquidity provider can check an incoming order against the current market before accepting or rejecting it. It protects the provider from stale-price orders and adds latency and rejection risk for the trader.
Why are my orders slow only during news?
During news releases liquidity thins, prices jump and servers, bridges and providers are at peak load. Queues lengthen, last-look rejections increase and requotes become common, so the slow tail of the distribution shows up exactly then.
Which platform has the lowest latency: MT4, MT5, cTrader or FIX?
The platform matters less than the broker’s infrastructure behind it. Direct FIX connectivity generally removes some layers compared with a retail platform account, but a well-run MT4 or MT5 server near your VPS can beat a poorly run alternative. The bridge and execution model are usually the deciding factors.
About The Author
