Free strategy check — send your trading rules and we tell you within 24 hours if they can be automated. No obligation. Send your rules →
Testing2 Oct 2026 · Eye Algo

A strong backtest is a good start, not a promise. These are the usual reasons an EA performs differently on a live account, and a simple way to test so that fewer surprises reach your real money.

Common causes

  • Spread and slippage: the tester may use a fixed or average spread; live spreads widen at news and rollover.
  • Data quality: testing on generated ticks instead of real tick data can hide problems on short timeframes.
  • Execution: requotes, partial fills and latency from your PC or VPS to the broker.
  • Overfitting: settings tuned to fit past data often fail on new data.
  • Broker differences: server time, symbol specifications and trading hours.

Spread and slippage in more detail

A strategy that aims for small targets is hit hardest by costs. If the average spread in the tester is lower than the spread you pay at 23:00 server time, or during a news release, results will be worse live. Always test with your broker's real spread and commission, and look at what happens in the hours when spreads are widest.

Overfitting

If you optimise many settings on the same period, you can find a combination that fits the past almost perfectly and has no link to the future. Signs of this include very precise values, results that collapse when one setting changes slightly, and a long list of conditions that each remove one losing trade. Fewer, simpler rules usually hold up better.

How to test properly

  1. Backtest on real tick data with realistic spread and commission.
  2. Check results on a period the settings were not tuned on.
  3. Forward test on a demo account for at least a few weeks.
  4. Go live with small size and compare live trades with the demo.

What to compare

  • Entry and exit prices of the same signal in the tester and on demo.
  • Number of trades over the same period.
  • Spread at the moment of entry.
  • Any trades that opened or closed late, and the reason.

If the live and demo trades differ without a clear reason, stop and find the cause before adding size. A log file from the EA helps, so ask for logging when you write your brief.

Every Eye Algo delivery includes a demo build so you can test on your own account before you pay the balance.

Where we can help

We build the EA to follow your rules and show a working demo before the balance is due. Free revisions (3 rounds) cover anything that does not match the agreed rules, and bug fixes are free for 15 days on simple projects and 30 days on complex ones. See the process page for the steps. What we cannot do is promise how a strategy will perform. An EA automates rules; it does not make a strategy profitable, and trading carries risk.

The gap between tester and live also shows up in other areas. Converted TradingView strategies are a common case; read Pine Script to MT5 for the details.

Settings that should not be changed too often

After every losing week, it is tempting to adjust the settings. If you do this often, you are tuning the EA to recent history, which is the same trap as overfitting. Decide in advance how long you will observe before changing anything, and change one setting at a time so you can see the effect. Keep a simple log: date, what you changed, and why.

Using a VPS

A VPS close to your broker's server can reduce delay and keeps the EA running when your PC is off. It does not fix a weak strategy or remove slippage, but it removes problems such as a dropped home connection. Test the EA on the VPS on demo before using it live, because the environment can differ from your own computer.

Keep expectations realistic. A tester run on past data cannot include every condition that will happen next month. Treat the numbers as a check that the rules work as written, not as a forecast, and size your live trades so that a bad stretch does not put your account at risk.

Frequently asked questions

How long should I forward test on demo?

At least a few weeks, and long enough to include several of the market conditions the strategy is meant for. A shorter test may not show the problems that appear around news or rollover.

Is a good backtest useless, then?

No. It shows whether the rules were coded correctly and how they behaved on past data. It is just not enough on its own to expect the same live results.

Why does my EA open trades at different prices live?

Prices move between the signal and the fill, and your broker may fill at a different price than shown. Latency and spread also add differences. Compare the same trade on demo and live to see the size of the gap.

Have a strategy you want automated?Get a free strategy check
← All articles