A swing trader who holds positions for days can tolerate 100ms of latency without noticing. A scalper targeting 2-3 pips per trade cannot. For scalping and high-frequency strategies, latency is not a nice-to-have optimization — it is the variable that determines whether your edge exists at all. Every millisecond of delay translates directly into slippage, and slippage on a 2-pip target is the difference between a profitable strategy and a losing one.
This article covers the specific latency requirements for different trading styles, the math behind how delay erodes scalping profits, why your broker’s server location matters more than your VPS specs, and how to configure a VPS specifically for fast execution.
Latency Requirements by Trading Style
Not every strategy needs sub-millisecond execution. The latency tolerance depends on your holding period and profit target.
- Swing trading (hours to days): Under 50ms is perfectly fine. Your entry and exit are separated by hundreds of pips. A few milliseconds of delay is invisible.
- Day trading (minutes to hours): Under 20ms is recommended. You are targeting 10-30 pips, so slippage of 0.5-1 pip from latency starts to matter.
- Scalping (seconds to minutes): Under 5ms is critical. Profit targets of 2-5 pips mean that even 0.3 pips of slippage from latency eats 6-15% of every trade.
- HFT and arbitrage (milliseconds): Under 1ms is essential. These strategies exploit fleeting price discrepancies that exist for only milliseconds. If you are slower than the discrepancy, you lose money instead of making it.
⚠️ Warning: On a 2-3 pip scalping target, even 0.3 pips of slippage per trade eats 10-15% of your profit. Latency-driven slippage is the silent killer of otherwise profitable scalping strategies.
The takeaway: if you scalp or run any strategy with single-digit pip targets, latency is not optional. It is your primary infrastructure requirement.
The Slippage Math: Why Milliseconds Equal Money
Let us put real numbers on this. Consider a scalping EA targeting 2 pips on EUR/USD (approximately $20 per standard lot).
Scenario 1: Home Computer or Budget VPS (50ms latency)
At 50ms round-trip latency, the price can move 0.2-0.5 pips between when your EA sends the order and when the broker executes it, depending on volatility. During news events or fast markets, this gets worse.
- Average slippage: 0.3 pips per trade
- Slippage as percentage of target: 0.3 / 2.0 = 15% of your profit gone
- Over 20 trades per day: 6 pips lost to slippage = $60 per standard lot daily
- Monthly cost of latency: roughly $1,200 per standard lot
📊 Key Stat: For a scalper taking 20 trades/day on standard lots, reducing average slippage from 0.5 pips to 0.05 pips saves roughly $900/month — far more than any VPS subscription costs.
Scenario 2: FXVPS in the Same Datacenter (1-2ms latency)
At 1-2ms round-trip, the price moves 0.01-0.03 pips during execution under normal conditions. The slippage is effectively negligible.
- Average slippage: 0.02 pips per trade
- Slippage as percentage of target: 0.02 / 2.0 = 1% of your profit
- Over 20 trades per day: 0.4 pips lost to slippage = $4 per standard lot daily
- Monthly cost of latency: roughly $80 per standard lot
The difference is over $1,100/month per standard lot traded. A FXVPS Pro plan at $39/month pays for itself on the first trade of the first day. This is not a marketing claim — it is arithmetic. For real-world numbers comparing VPS to home setups, see our VPS vs local PC latency benchmarks.
Broker Server Proximity: The Factor That Matters Most
Here is something that surprises many traders: the raw specs of your VPS (CPU speed, RAM amount) matter far less for execution speed than where the VPS is physically located relative to your broker’s trade server.
Your order travels a physical path: EA calculates signal → MetaTrader sends order → data travels through network → broker’s matching engine receives and executes → confirmation travels back. The network transit portion of this is governed by physics — the speed of light through fiber optic cable is approximately 200 kilometers per millisecond.
A VPS in a datacenter in Dallas connecting to a broker in London covers roughly 7,800 km of cable. That is about 39ms one way, 78ms round trip, just from the speed of light. No amount of CPU optimization or RAM can overcome geography.
Where the Major Brokers Are
The forex industry has consolidated around a handful of Equinix datacenters:
- Equinix LD4 (London): IC Markets, Pepperstone, Tickmill, FP Markets, FXCM, and dozens more. London is the largest forex trading hub, handling roughly 38% of global forex volume.
- Equinix NY4 (New York): OANDA, FXCM, Interactive Brokers, and brokers serving the Americas. Critical for USD pairs and overlap sessions.
- Equinix TY3 (Tokyo): Brokers serving the Asian session and JPY pairs.
- Equinix HK1 (Hong Kong): Hub for brokers serving the broader Asian and Australian markets.
📊 Key Stat: Major forex brokers cluster in a handful of Equinix datacenters (LD4 London, NY4 New York, TY3 Tokyo). A VPS in the same facility achieves sub-millisecond latency because data travels meters, not continents.
FXVPS operates servers inside these same Equinix facilities. When your VPS and your broker’s matching engine are in the same building, the connection is a cross-connect — a direct fiber cable running through the datacenter’s cable trays. The data does not traverse the public internet at all. It travels meters, not kilometers.
This is how FXVPS achieves measured latencies like 0.38ms to certain brokers. That number is not theoretical — it is the physical result of being in the same building with a direct fiber path. Check the specific latency to your broker on the broker latency page, where FXVPS publishes real measurements for 195+ brokers including IC Markets (2.50ms from LD4), Pepperstone (1.99ms), and FTMO (2.32ms).
VPS Configuration for Scalping
Once you have the right location, here is how to configure the VPS itself for maximum execution speed.
Dedicated Cores Are Non-Negotiable
Shared vCPUs mean that when another customer on the same host server runs a CPU-intensive operation, your available processing power drops. For a scalping EA, this manifests as random execution delays — your order takes 15ms to process instead of 1ms because the hypervisor scheduled another VM’s workload on your core at that moment.
FXVPS provides dedicated CPU cores on all plans. Your cores are reserved exclusively for your VPS. Other customers cannot consume them. This eliminates the “noisy neighbor” problem that plagues budget VPS providers.
For scalping, the Pro plan at $39/month is the minimum recommended tier. Scalping EAs need fast single-thread performance for order calculation, and the Pro plan’s dedicated cores deliver consistent processing speed without contention.
NVMe SSD for Tick Data
Scalping EAs often process every incoming tick rather than waiting for candle closes. On a busy pair like EUR/USD during London session, that is 5-20 ticks per second. Each tick involves a disk read (chart history) and a potential disk write (logs, state files).
NVMe SSDs deliver 2,000-3,500 MB/s throughput compared to 80-160 MB/s for traditional HDDs. The difference is not about total throughput (MetaTrader does not move gigabytes) — it is about access latency. NVMe responds in 0.02-0.05ms per I/O operation versus 5-15ms for HDD. Over thousands of ticks per hour, this adds up. FXVPS uses NVMe SSD storage across all plans.
Terminal Optimization for Speed
Strip your MetaTrader installation down to the essentials.
Market Watch: Right-click → Hide All, then add back only the symbols your EA trades. Every visible symbol in Market Watch is a live tick stream consuming CPU and bandwidth. If your EA trades EUR/USD and GBP/USD, you should see exactly two symbols.
✅ Best Practice: In Market Watch, right-click and “Hide All,” then add back only the symbols your EA trades. Every visible symbol is a live data stream consuming CPU and bandwidth for zero benefit.
Charts: Open only the charts your EA requires. If your scalping EA runs on EUR/USD M1, you need one chart. Close everything else. Each chart is a data processing pipeline — fewer charts means more CPU available for order execution.
History bars: Go to Tools → Options → Charts and set Max bars in history to 10,000-20,000. Scalping EAs rarely look back more than a few hundred bars. Loading 500,000 bars of history consumes RAM and slows terminal startup for no benefit.
News and alerts: Disable news (Tools → Options → Server → uncheck Enable news), disable email alerts, disable sounds. Every disabled feature is one less background thread competing for your CPU.
For a detailed walkthrough of all optimization steps, see the MT4/MT5 performance optimization guide.
Tick Charts vs Timeframe Charts
Some scalping EAs use tick charts rather than time-based charts (M1, M5, etc.). A tick chart draws a new candle after a fixed number of ticks rather than after a fixed time period. This captures price action more granularly during volatile periods.
MT5 supports tick charts natively. On MT4, tick charts require a custom indicator that converts the tick stream. Both approaches increase CPU and memory usage compared to standard timeframe charts.
If your EA uses tick charts, account for the additional resource usage. A single MT4 terminal running a tick-chart scalper on 3 pairs uses roughly the same resources as 2-3 terminals running standard timeframe EAs. Size your VPS plan accordingly.
Scalping Checklist: Before You Go Live
Before deploying a scalping EA on your VPS:
- Verify your broker’s latency from your chosen VPS datacenter at the broker latency page. If it is above 5ms, consider switching to a closer datacenter location.
- Confirm dedicated resources. Ask your VPS provider whether CPU cores are dedicated or shared. On FXVPS, they are always dedicated.
- Strip the terminal. Follow the optimization steps above. A lean terminal executes faster.
- Monitor the first week. Check Task Manager for CPU and RAM usage during market hours. If CPU exceeds 60% sustained, your EA is compute-heavy and may benefit from the Scaling plan.
- Check the Journal tab. Frequent “no connection” or reconnection messages in MetaTrader’s Journal indicate network instability, not just latency. This is a different problem that requires investigation.
- Test during high volatility. Run your EA through a major news event (NFP, FOMC, ECB) on a demo account first. Tick rates spike 5-10x during these events, and your VPS needs headroom to handle it. Our news trading execution speed guide covers this in detail.
The Bottom Line for Scalpers
Scalping profitability comes down to a simple formula: your edge minus your execution costs. Latency is an execution cost. Slippage is an execution cost. Running on shared, oversold infrastructure in a datacenter 50ms from your broker adds execution costs that can exceed your strategy’s edge entirely.
FXVPS puts your terminal in the same building as your broker, on dedicated hardware, with NVMe storage, starting at $29/month for the Core plan or $39/month for the Pro plan recommended for scalping. The latency numbers are published and verifiable. The infrastructure is built for exactly this use case.
Check your broker’s measured latency at the broker latency page and get started at FXVPS pricing.