Execution infrastructure — VPS, broker choice, and why latency matters for your EA
The best strategy in the world loses money if it runs on infrastructure that can't execute it cleanly. This lesson covers the three decisions that determine your EA's execution quality: where it runs (PC vs VPS), who it runs through (broker latency and execution model), and how to measure whether you have a problem.
Last reviewed:
The 90-second version
Infrastructure is not a detail — it's the difference between your EA performing as tested and losing a pip on every single trade it takes.
- Scalper EAs require VPS co-located near your broker's server — latency costs real money per trade
- Swing and position traders can run on a reliable PC, but uptime is still non-negotiable
- Broker execution model matters: ECN/STP gives you tighter spreads; MM brokers may widen spreads at news
- You can measure your real fill quality from MT5's Trade History tab — no third-party tools needed
- IC Markets published data in 2024: co-located VPS averaged <1ms fill; home retail connection averaged 40–80ms
IC Markets benchmark data showing median fill times for co-located vs remote connections — the difference explains why some scalping EAs that look identical in backtests perform very differently live.
IC Markets published execution quality benchmarks in 2024 showing average fill latency under 1ms for co-located VPS users vs 40-80ms for home connections — a difference that costs scalping EAs 1-3 pips per trade in live conditions.
SourceLatency Impact Calculator
Enter your strategy type, broker latency, and expected spread to see whether your execution infrastructure is costing you performance — or whether you can safely stay on your current setup.
Entry cost is only 2.3% of target — your infrastructure is adequate.
The three infrastructure decisions
PC vs VPS — what's actually different
Running an EA on your home PC seems like the obvious starting point. It costs nothing extra, you have full control, and you can watch it work. The problem is that a PC is a general-purpose computer with competing priorities: Windows Update wants to restart it at 3 am, the power company occasionally fails, your internet router needs a reboot, and the cooling fan eventually gives out. Any of these events leaves your EA with an open trade and no pilot.
A VPS (Virtual Private Server) is a rented slice of a data-centre machine that runs 24 hours a day, 7 days a week, regardless of what your home electricity, router, or operating system is doing. You access it via remote desktop, install MT5, set up your EA, and then leave. The server continues running whether your laptop is open or not.
The latency dimension is separate from uptime. A home internet connection introduces 30–150ms of round-trip time between MT5 and your broker's execution server, depending on geography. A VPS inside the same data centre as your broker can cut that to under 2ms. For a scalper with a 3-pip target, a 50ms delay that causes 0.5 pips of slippage on every entry is a 14% tax on every trade's expected value.
When a PC is fine: swing traders on H1 and above, position traders, any strategy where the target is 25+ pips. The latency drag is negligible relative to the profit target. Set your MT5 to disable auto-shutdown, configure an uninterrupted power supply (UPS), and make sure your ISP is reliable. That is a €0 setup that works.
When you must use a VPS: scalpers with 2–8 pip targets; EAs that use very tight stops that require fast execution to get the expected fill; any strategy that must be running at all times including overnight and weekends (for instruments that open at Sunday open). Also consider VPS if you travel and cannot guarantee your home connection.
Broker execution quality — the three metrics that matter
Not all brokers are equal, and the difference is measurable before you deposit. The three metrics that determine execution quality for an EA are: fill time, slippage consistency, and spread behaviour during news events.
Fill time is the round-trip between your EA sending an order and MT5 receiving a confirmation. You can see this in MT5 under Help → About, where the ping to the broker's server is shown. Anything under 10ms is excellent; 10–50ms is acceptable for non-scalping strategies; above 100ms starts to affect short-target EAs noticeably.
Slippage consistency matters more than average slippage. An EA calibrated to a 30-pip stop can tolerate 1–2 pips of occasional slippage. But if slippage spikes unpredictably to 10 pips during news, the risk layer breaks down — the EA thinks it risked 30 pips when it actually risked 40. Consistent, predictable slippage is preferable to low average but high variance.
Spread behaviour during news is the third variable. Many brokers describe themselves as ECN but widen spreads dramatically in the 2 minutes around major releases (NFP, FOMC, CPI). A strategy backtested on normal spreads will underperform live if the broker's spread triples on 30–40 trades per year. Check broker-comparison sites like MyFXBook or FX Blue for independent spread-monitoring data, not the broker's own marketing materials.
How to evaluate a broker before you deposit: look for independently published execution statistics, not just Trustpilot reviews. Sites like myfxbook.com/brokers/slippage and fxblue.com publish aggregated slippage and fill-time data from real accounts. Prioritise brokers that publish this data proactively — they have nothing to hide.
Measuring your actual execution in MT5
The most important skill in this lesson is not knowing which VPS to buy — it is knowing how to verify that your current setup is working at an acceptable level. MT5 gives you everything you need without third-party tools.
Start with the connection ping: in MT5, look at the bottom right of the screen. There is a small icon showing connection status and, if you hover over it or check Help → About, the current server ping. This is the single-hop latency from your MT5 instance to the broker's server. Note it when markets are quiet and note it again during a busy news release — the change tells you how your connection behaves under load.
For actual execution quality, open the Trade History tab (View → Terminal → Trade History). Each executed trade shows an open time and a close time. But more usefully, if you enable FX Blue logging or use the Journal tab, you can see the timestamp of when the order was placed versus when it was filled. The gap between those two timestamps is your real fill time — not the theoretical ping, but the actual time the broker took to process the order.
The comparison that matters most is demo vs live on the same broker. Run your EA for 2 weeks on a demo account with a live chart feed (not a demo data feed — a live MT5 feed set to demo mode). Then run it for 2 weeks on a small live account. Compare average fill times and average slippage across those 20–40 trades. If demo shows consistently better execution, that tells you the broker adjusts live execution differently from demo — which is a red flag.
FX Blue's trade copier and trade analysis tools can automate this comparison. You paste your MT4/MT5 history into their analyser and get a slippage breakdown, fill-time histogram, and spread analysis. The free tier covers enough trades for a meaningful sample.
Key terms
A real benchmark from 2024
In 2024, IC Markets published execution quality benchmarks that quantified something traders had argued about qualitatively for years: the latency gap between co-located VPS users and home retail connections is not a marginal rounding error — it is a 40x to 80x multiple.
IC Markets published execution quality benchmarks in 2024 showing average fill latency under 1ms for co-located VPS users vs 40-80ms for home connections — a difference that costs scalping EAs 1-3 pips per trade in live conditions.
The benchmark showed co-located VPS accounts averaging under 1ms fill latency against IC Markets' NY4 server. Home retail connections — even fast fibre in major cities — averaged 40–80ms. For a scalping EA targeting 3 pips with a 1.5-pip spread, that 40ms delay translates to approximately 0.4 pips of additional entry cost from slippage on every single trade. Across 200 trades per month, that is 80 pips of pure infrastructure drag — more than many scalping EAs make in a month. The implication is not that everyone needs a VPS: a swing trader targeting 50 pips finds that 0.4 pips is noise. The implication is that infrastructure cost is strategy-specific, and choosing your infrastructure without knowing your strategy's pip-target profile is trading blind.
SourcePractice
Audit your current execution setup
This three-step practice takes 20 minutes and gives you a concrete baseline for your infrastructure quality. You do not need to change anything yet — the goal is measurement.
- 1
Measure your MT5 ping: open MT5, look at the bottom-right connection icon, and hover over it to see the current ping to your broker's server. Note the number. Do this at three different times: during low-volatility Asian session (2–4 am GMT), during London open (8–9 am GMT), and during NY open (2–3 pm GMT). Write down all three numbers — the range tells you how your connection behaves under different load conditions.
- 2
Open Trade History and count your last 10 executed trades. For each trade, look at the entry price versus the price shown on the chart at that exact second using your chart's crosshair. The difference is your slippage. Calculate the average. If your average entry slippage is more than 30% of your average spread, your execution is costing you more than the broker's quoted spread.
- 3
Compare the same EA on demo vs live: if you already run an EA on a live account, set up an identical copy on a demo account (same broker, same parameters) for 2 weeks. At the end, export both histories and compare entry prices on the same signals. Systematic differences suggest the broker treats demo execution differently from live execution — an important discovery before you scale up.
- 4
Check your PC's uptime record: on Windows, open Event Viewer → Windows Logs → System and look for unexpected shutdowns in the last month. Count how many times MT5 would have been restarted mid-session. If you see more than one unexpected restart per month, you have a reliability risk that a VPS would eliminate.
- 5
Use the LatencyImpactCalc widget above to plug in your actual latency and spread figures. Identify which verdict category your current setup produces for the strategy type you are running. If the result is 'Marginal' or 'Infrastructure upgrade needed', prioritise fixing the infrastructure before making any other changes to your EA.
Mastery check
Four questions. Pass at 75% (3/4). Focus on the practical distinctions — the point is not memorising definitions but being able to make the right infrastructure decision for a given EA type.
Mastery check — Lesson 8
Test your understanding with 4 questions. Pass with 75/4 correct.
Reflect
Reflection
Type your honest answers — saved on this device only. Use them next week to spot patterns in your trading thinking.
Pro deep dive
The retail infrastructure picture described in this lesson is the practical floor. Professional and near-professional traders face a different set of trade-offs. Here is the fuller picture for those who want to understand the execution stack from the bottom up.
FIX API vs retail MT5 execution
MT5 communicates with the broker's execution server using a proprietary MetaQuotes protocol. Professional traders who need sub-millisecond execution use FIX (Financial Information eXchange) API connections directly — bypassing MT5 entirely. FIX connections offer lower latency (often <0.5ms within a co-located environment), richer order types, and programmatic account management. The trade-off is complexity: FIX requires a dedicated development environment (usually Java, C++, or Python with a FIX library like QuickFIX), a prime broker relationship, and significantly higher minimum capital thresholds. For retail traders running MT5 EAs, FIX is not relevant — but understanding it explains why professional execution quality benchmarks (often published by banks and hedge funds) cannot be used as a target for MT5 EA traders.
Co-location economics for retail traders
The break-even analysis for co-location is strategy-dependent and worth doing precisely. A scalping EA that trades 10 times per day and saves 0.3 pips per trade (from reduced latency) recovers 3 pips/day = 60 pips/month. On a 0.1 lot position with a standard 10 USD/pip EURUSD value, that is $60/month in recovered performance — against a $20–25/month VPS cost. The ROI is clear. For a swing EA that trades 3 times per month, the same calculation: 3 trades × 0.3 pips recovered = 0.9 pips/month = $0.90 recovered. The VPS costs $20. The ROI is negative. The co-location decision is not 'VPS is better' — it is 'is the latency drag larger than the VPS cost for this specific strategy's trade frequency and pip target?'
Spread behaviour during news events
Retail forex spreads are not fixed — they are variable and reflect the broker's cost of sourcing liquidity. At major news releases (NFP, FOMC statements, ECB rate decisions), interbank liquidity temporarily contracts as market makers widen their own quotes to account for direction uncertainty. Retail brokers pass this widening on to clients, typically widening spreads by 5–20x for the 30–90 seconds around the release. An EA that enters during this window executes at a spread that its backtest never modelled (backtests typically use average or fixed spreads). The practical defence is a news filter: the EA checks the next high-impact event from an economic calendar API and blocks entries within a configurable window. For EAs without this filter, the spread spike effectively adds a random tax to every news-window entry. Lesson 11 covers EA monitoring, including how to audit whether your live EA is accidentally entering during news windows despite having a filter configured.
Execution vs strategy as the binding constraint
There is a common sequencing mistake in EA development: spending weeks optimising the strategy while running it on infrastructure that guarantees poor execution. The order matters. Before you try to improve a strategy, verify that your infrastructure is not the binding constraint. A simple test: run the EA on a co-located VPS for 30 days and compare performance to the same period on your current setup. If performance improves noticeably, infrastructure was the binding constraint and you can skip strategy optimisation for now. If performance is identical, infrastructure is not the issue — your strategy parameters are. This distinction saves months of misdiagnosed optimisation work.
Sources
Show answer
Where it runs (PC vs VPS, and if VPS, co-location proximity to the broker's server), who it runs through (broker execution model, fill time, slippage consistency, spread behaviour at news), and whether you are measuring it (Trade History analysis, demo vs live comparison, independent slippage monitoring).
Educational material only — not investment advice. Trading carries risk of capital loss. Always practice on demo and use a stop-loss. ← Back to Automated Trading