A year ago, I saw a backtest curve on a forum. 120% annual return, 8% max drawdown. I thought: this is a money machine.
I bought the EA. Rented a VPS. Turned it on.
One year later, the account is alive. But the detours I took along the way cost me more than they needed to. If I could go back and talk to myself twelve months ago, these are the ten things I'd say.
1. Five days on demo is nothing. Run at least three months.
I demoed for five days. Made money each day. Couldn't wait to go live.
Week two: ranging market. The EA stopped out seven times in a row. I panicked and shut it off manually. What I didn't know: right after those seven stops, a big move hit. If I'd left it running, the EA would have recovered and then some.
Five days of demo covers exactly one market mood. Trends, ranges, high volatility, low volatility, NFP reactions — you need to see all of them. My rule now: three months minimum on demo, covering at least one NFP, one central bank decision, and one unexpected event. Three months of patience vs three weeks of losses. Easy choice.
2. VPS location matters ten times more than specs.
My first VPS was a 2-core, 4GB server from a generic cloud provider. Plenty of power for MT4. But the server was in Asia and my broker was in London. Ping: 280ms. Every order took nearly 300 milliseconds to fill, which is a disaster for any scalping strategy.
Switched to a London VPS. Ping dropped to 2ms. Same EA, same parameters. Monthly return increased by nearly 40%.
When choosing a VPS, step one is finding out where your broker's server is located. Then pick the nearest data center. CPU and RAM are secondary. Our VPS guide covers the details.
3. The same EA on different brokers produces very different results.
I ran a trend EA on Broker A: +12% in three months. Opened the same EA with the same parameters on Broker B: -5% in three months.
The difference: Broker A averaged 0.8-pip spread on EURUSD. Broker B averaged 1.6 pips. Sounds like a small gap. But this EA traded 4–5 times daily. Over a month, that's dozens of extra pips in cost. Add Broker B's worse slippage (averaging 1.2 pips more than A), and the strategy's breakeven point shifted from profitable to losing.
After getting any new EA, run it on demo at two or three brokers simultaneously for a month. Compare the trade logs. The difference will surprise you.
4. Backtest results are always better than live. Plan for a 30–50% discount.
I can't emphasize this enough.
I have an EA with a 10-year backtest: 62% annual return, 11% max drawdown. After one year live: 28% annual return, 19% max drawdown. Return cut by more than half. Drawdown nearly doubled.
Backtests have no slippage, no requotes, no network delays, and fixed spreads. Live trading has all of these, and spreads can spike 10x during news. When I look at a backtest report now, I automatically discount the profit by 30–50% and multiply the drawdown by 1.5–2x. If the discounted numbers are still acceptable, I consider going live. If not, I pass. More on this in our backtest vs live trading gap guide.
5. Your biggest enemy is yourself, not the market.
Last June, my EA had three losing days in a row. Floating drawdown hit 8%. Historical data said this was completely normal. But I couldn't help it. I manually closed two losing positions and opened a reverse hedge.
Result: the two positions I closed would have recovered to profit the next day. My manual reverse trade lost another 3%. An 8% floating loss became a 14% realized loss because I intervened.
I made a rule after that: don't touch the EA unless the preset maximum drawdown fuse triggers. No peeking at the market. No changing parameters. No manual closes. Over the past year, every time I intervened manually, the outcome was worse than doing nothing. Every single time. Read our trading psychology guide if this resonates.
6. VPS and spread costs eat small accounts alive.
Quick math. A London VPS costs roughly $50–80/month. That's $600–960/year.
$1,000 account with 20% annual return = $200 profit. VPS costs alone exceed that. You're underwater before accounting for spreads.
$5,000 account at 20% = $1,000 profit. VPS costs eat 60–96% of it. Add spread costs and you're working for free.
The minimum viable capital for EA trading is higher than most people think. $5,000 is where it starts to make sense. $10,000 is where the economics actually work. If you don't have the capital yet, demo trade and accumulate experience. Don't undercapitalize a live account.
7. One EA is like one stock. Diversify.
My first year: one EA, one pair, one timeframe.
November 2024: EURUSD volatility dropped to unusually low levels. My trend EA opened six trades all month, four were losers. Monthly return: -4.3%.
That same month, GBPJPY volatility was normal. If I'd had a second EA running GBPJPY, it would have offset the EURUSD losses.
Now I run three EAs: trend following on EURUSD, mean reversion on a cross pair, and range breakout on gold. Different logic, different pairs, low correlation. After combining them, worst monthly loss dropped from -4.3% to -1.8%, and annual return went from 19% to 26%. The smoother curve lets me use slightly larger position sizes.
8. Boring EAs are the ones that actually make money.
I tested two EAs side by side.
EA-A: backtest 110% annual, 32% max drawdown. Wild equity curve. Exciting to watch.
EA-B: backtest 31% annual, 9% max drawdown. Nearly a straight line. Boring.
I chose EA-A. Who doesn't want 110%?
Three months live: EA-A hit 40% drawdown during a volatile stretch. I panicked and shut it off. Net result: -7%.
EA-B? I put it on later. Six months of live trading: +14% net profit. No month worse than -3%. Boring? Yes. Account growing? Also yes.
High returns and high drawdowns are twins. You don't get one without the other. For most people, the psychological stability of low drawdown is worth far more than the potential upside of high returns. A strategy you can actually hold through a drawdown is the only strategy that makes you money. We wrote about realistic return expectations in detail.
9. Without a news filter, one NFP night costs you a month of profit.
September 2024, first Friday. US employment data. My EA had no news filter. Two minutes before the release, EURUSD spread widened from 0.9 to 12 pips. My EA triggered an entry signal at exactly that moment. Filled at the inflated spread.
Data dropped. Price spiked 80 pips in three seconds, then reversed. My stop loss was hit. Lost 3.2% of the account in one trade.
After installing a news filter that pauses trading 30 minutes before and after major events, my maximum single-trade loss dropped from 3.2% to 1.1%. The filter costs you a few missed trades per month. It saves you from the one trade that would have ruined your month.
10. "Set and forget" is a myth.
EA sellers love saying: "Just turn it on and make money in your sleep."
I believed it. Then I discovered my VPS had run out of memory, MT4 crashed, and the EA had been offline for three days. I missed two profitable trades during those three days.
Another time: the broker updated their server IP. MT4 couldn't connect. Took me two days to notice.
Another time: a currency pair changed its minimum lot size from 0.01 to 0.1. The EA was still using old parameters and the position management logic broke silently.
My routine now: 5 minutes daily to confirm MT4 is running. 30 minutes weekly reviewing trade logs against expectations. Monthly parameter review. Quarterly evaluation of whether each EA still fits the current market regime.
You don't need to watch the market. You do need to monitor the system. Those are different things.
After one year
EAs aren't the Holy Grail. They're tools. Used correctly, they save you time and remove emotional interference from execution. Used incorrectly, they help you lose money faster.
These ten lessons cost me real money. If you're about to start running EAs, I hope they save you some. Check out the FXTool marketplace for EAs with verified backtest data, or start with our what is an EA guide if you're still at the beginning.
FAQ
What's the minimum capital for running an EA?
$5,000 minimum to cover VPS costs and trading expenses without eroding all your profit. Below that, use demo to build skill and wait until you have enough capital. Never borrow money or use living expenses.
The backtest looks great but live is losing. Was I scammed?
Not necessarily. A 30–50% performance gap between backtest and live is normal. If live results are the complete opposite (backtest profits, live consistently loses), check for overfitting or broker execution issues.
Do I need to understand trading to run an EA?
Yes. You don't need to code, but you need to understand the strategy logic. Is it trend following or mean reversion? When does it perform well? When should you turn it off? Handing money to a system you can't explain is gambling, not trading.
How long does an EA last before needing replacement?
No fixed answer. Some run for years, others fail in six months. Monitor continuously. If actual performance deviates significantly from historical averages for three consecutive months, evaluate whether the strategy has stopped working or the market has simply been unfavorable. Don't change constantly, but don't expect one EA to work forever.
About the author: The FXTool team builds and tests MetaTrader trading tools daily. We run every EA we sell on live accounts and publish the results. This guide reflects what we've learned from building 50+ EAs and working with thousands of retail traders.
Forex trading involves significant risk and may result in total loss of capital. This article is for educational purposes only and is not investment advice. Understand the risks and consider your financial situation before trading.