Every forex VPS provider advertises a latency number. Ours included. It is a useful number, it is easy to measure, and it is genuinely meaningful — but on its own it is a bad predictor of whether your EA will trade well.
Here is the reason, stated plainly. Two connections can both average 4ms to your broker. One delivers 4ms on essentially every packet. The other delivers 1ms most of the time and 90ms during the London open, with occasional packets that never arrive at all. Both providers can honestly print “4ms average” on the pricing page. Only one of them will fill your orders the way your backtest assumed.
The variables that separate those two connections are jitter and packet loss. Almost nobody advertises them. They are the numbers you should be measuring.
The Three Numbers That Describe a Connection
- Latency (round-trip time): how long a packet takes to reach the broker and come back. Measured in milliseconds. Lower is better.
- Jitter: the variation in that latency between consecutive packets. If your round trips are 4, 4, 5, 4, 4ms, jitter is near zero. If they are 4, 31, 6, 78, 4ms, jitter is severe. Measured in milliseconds of deviation. Lower is better.
- Packet loss: the percentage of packets that never arrive and must be retransmitted. Measured as a percentage. Zero is the only acceptable answer for trading.
Latency sets your floor. Jitter and packet loss determine how often you fall through it.
📊 Key Stat: A single lost TCP packet typically costs far more than the raw latency figure suggests. The sender must detect the loss and retransmit, and that recovery cycle can add tens or hundreds of milliseconds to a single order — on a connection whose “average latency” still reads in single digits.
Why Jitter Hurts More Than Latency
Consistent latency is something a strategy can be built around. If every order takes 5ms to reach the broker, that 5ms is a fixed cost baked into your expectancy. Your backtest is wrong by a constant, and a constant is manageable.
Jitter is not manageable, because it is not predictable. And jitter has a nasty property: it is usually worst exactly when it matters most. Network congestion peaks during the same high-volume windows as market volatility — the London open, the New York open, the minutes around a Non-Farm Payrolls release. Your connection is at its least reliable during the sessions your strategy makes most of its money in.
Concretely, here is what that does to different trading styles:
Scalpers are hit hardest. A strategy targeting five to ten pips cannot absorb an occasional 80ms delay on entry, because in a fast market 80ms is enough for the price to move past your level. The requote or the slipped fill turns a positive-expectancy system into a break-even one. Our scalping and HFT on VPS guide covers the sensitivity in detail.
News traders are hit at the worst possible moment. The whole premise is executing in the first seconds of a release, and that is precisely when every retail connection in the world is congested. See news trading on VPS.
Grid and martingale systems are hit structurally. These strategies place many orders and depend on each level filling where expected. Inconsistent execution across a grid distorts the whole position ladder — read grid and martingale VPS requirements.
Swing traders are genuinely less affected, and it is worth saying so. If you enter three positions a week and hold them for days, a 50ms jitter spike is noise. The honest advice is that your reason for using a VPS is uptime, not jitter.
⚠️ Warning: Be suspicious of any provider that publishes only a best-case latency figure with no mention of consistency, and especially of “sub-1ms” claims measured to something other than your actual broker’s server. A number measured to a nearby router, or averaged over a quiet Sunday, tells you nothing about Tuesday at 08:00 London.
Where Jitter and Loss Come From
Understanding the causes tells you which ones you can fix.
The residential last mile. This is the dominant source for home traders and it is not fixable by tuning. Consumer broadband is contended — you share upstream capacity with your neighbours — and it is optimised for throughput and cost, not for consistent low-latency delivery of small packets. Evening congestion, DOCSIS scheduling on cable, and interleaving on DSL all add variable delay.
Wi-Fi. Even excellent Wi-Fi adds jitter, because the radio medium is shared and retransmissions are routine. A trader on Wi-Fi is adding variance before their packet has left the building. If you trade from a local machine at all, use Ethernet.
Hop count and routing. Every router between you and the broker is a chance to queue. A path with eighteen hops crossing three networks has far more opportunity for variance than a path with two hops inside one datacenter.
Cross-continent distance. Physics sets a floor: roughly 5ms of round trip per 1,000km of fibre, before any equipment delay. A London-to-Tokyo path cannot be made consistent by any amount of optimisation. This is why matching your VPS to your broker’s datacenter is the single highest-leverage decision — see choosing the best VPS location.
Oversubscribed VPS hosts. This one is a provider problem, not a network problem, but it presents identically. If your virtual machine is sharing a physical CPU with noisy neighbours, your terminal can be descheduled for milliseconds at a time. The packet arrives on schedule and your EA processes it late. Our dedicated cores versus shared vCPU explainer covers the mechanism.
How to Actually Measure It
You can do this yourself in fifteen minutes, and you should — for your current setup and for any VPS you are evaluating.
Step one: find your broker’s real server address. The MetaTrader connection status shows latency to your broker, but for a proper test you want the hostname or IP. Our guide to extracting your broker’s server address walks through it.
Step two: run a sustained ping, not a quick one. Open Command Prompt and run a long count rather than the default four packets:
ping -n 200 your-broker-server.com
Two hundred packets takes a few minutes and gives you something statistically meaningful. Read the summary at the end: minimum, maximum, average, and the loss percentage.
Step three: read the gap between minimum and maximum. This is your practical jitter indicator. A min of 3ms and a max of 6ms is a healthy connection. A min of 3ms and a max of 240ms is a connection that will hurt you, no matter how good the average looks.
Step four: repeat during your actual trading hours. A test at 11pm on a Sunday is close to meaningless. Run it during the London open, during the New York overlap, and during a scheduled news release. That is when you find out what you actually have.
Step five: check loss specifically. Any sustained packet loss above zero is a red flag for trading. Occasional single-packet loss over a long test can be normal internet behaviour; 1% or more is a problem you need to solve.
💡 Tip: Run the same 200-packet test from your home machine and from a trial VPS to the same broker server, at the same time of day. That side-by-side is the only comparison that answers your actual question, and it is far more informative than any provider’s published figure — including ours.
What Good Looks Like
For a VPS colocated in the same datacenter region as your broker’s matching engine, these are reasonable expectations:
- Latency: low single-digit milliseconds — typically 1-5ms within the same facility
- Jitter: min-to-max spread of a few milliseconds, holding during volatile sessions
- Packet loss: zero across a sustained test
For a home connection to a broker on another continent, expect 100ms or more with a min-max spread that can run into the hundreds. Between those poles sits everything else. Our VPS versus local PC latency benchmarks has measured comparisons.
The structural reason a colocated VPS wins is not that the wires are faster. It is that the path is shorter and less contended. Fewer hops means fewer queues, and a datacenter uplink is engineered for consistent delivery in a way that consumer broadband is not. You are removing variance sources, not adding speed.
🚀 Try FXVPS free for $1.99 — get 7 days on the Core VPS and run the 200-packet test yourself against your own broker. That is the only benchmark that counts. Cancel anytime.
Choosing a Plan With Jitter in Mind
Network consistency is largely a function of location and provider infrastructure rather than plan size, so the first decision is the datacenter, not the tier. Pick the FXVPS location that matches your broker’s matching engine — London (LD4) for the majority of retail forex brokers, New York (NY4) for US-routed brokers and futures, Tokyo or Hong Kong for Asian execution.
Core Plan ($29/mo) — 2GB RAM, 1 vCPU. Correct for a single terminal with a handful of EAs. You get the same network path as any other tier.
Pro Plan ($39/mo) — 4GB RAM, 2 vCPUs. The right choice if you run multiple terminals or heavy indicator stacks, because processing delay on an overloaded machine looks exactly like network jitter from your EA’s point of view.
Scaling Plan ($79/mo) — 8GB RAM, 4 vCPUs. For multi-account operations where you want certainty that nothing on the box is competing for CPU during a volatile open.
✅ Best Practice: Before you conclude your network is the problem, check the machine. Open Task Manager during your worst session and watch CPU. If you are sitting at 95% while five EAs recalculate on every tick, your execution delay is local processing, not the wire — and no amount of latency shopping will fix it. Our performance metrics guide covers what to watch.
Frequently Asked Questions
Is 20ms latency with low jitter better than 5ms with high jitter?
For most automated strategies, yes. A predictable 20ms is a constant your system absorbs. A 5ms average that spikes to 150ms during the London open produces slippage precisely when your position sizing assumed clean fills. Consistency beats a headline average.
How do I measure jitter on Windows?
The simplest proxy is a long ping run — ping -n 200 <server> — and reading the gap between the reported minimum and maximum. A tight spread means low jitter. For continuous monitoring, pathping gives per-hop loss and latency statistics, which also tells you where in the path the problem sits.
Does packet loss cause requotes?
It contributes. A lost packet forces a retransmission, which delays your order reaching the broker. By the time it arrives the quote may have moved, and depending on your broker’s execution model that produces a requote or a slipped fill. Our slippage and requotes guide covers the full chain.
Will a VPS eliminate jitter completely?
No, and anyone claiming otherwise is overselling. A colocated VPS removes the largest and most variable segment of the path — the residential last mile — and replaces it with a short, engineered datacenter route. Residual variance remains, but it typically drops from tens of milliseconds to low single digits.
Should swing traders care about jitter?
Honestly, not much. If you hold positions for days, execution variance measured in milliseconds is irrelevant to your results. Your reason for running a VPS is uninterrupted uptime and reliable EA operation, not fill quality. Buy the plan that gives you reliability and stop optimising for a number that does not affect you.
Related Reading
- VPS vs Local PC: Real Latency Benchmarks — measured comparisons of the two paths
- Slippage, Requotes and VPS Latency — how network delay becomes a cost on your statement
- Dedicated Cores vs Shared vCPU — why processing delay masquerades as network delay
- Tips for Choosing the Best VPS Location — the decision that dominates all the others