Backtest vs Live Performance

Almost every EA looks better in the tester than in the account. That is not always fraud or a broken strategy; it is the normal difference between a simulation that knows the past perfectly and a live market that does not care what your backtest showed. The useful questions are how big the gap is, what causes it, and whether you can shrink it. Expect live to be somewhat worse than the backtest and you will be planning around reality instead of a hopeful chart.

Why live rarely matches

Several things quietly separate the tester from the account:

Real spread — backtests are often run on a tight fixed spread. Live, the spread breathes and widens exactly when your EA wants to trade.
Slippage — the tester fills you at the price you asked for. Live, fast markets fill you a little worse, and it adds up.
Requotes and rejections — a backtest never has an order refused. A live broker can, especially around news.
Latency — the milliseconds between your terminal, the VPS and the broker's server. Harmless for slow EAs, fatal for scalpers.
News and gaps — real markets jump, gap over stops and briefly turn untradeable. Interpolated backtests glide through all of it.
Broker differences — the EA was tuned on one broker's prices and feed; yours are not identical.
Over-optimisation — if the backtest was curve-fitted, the live account is where the fantasy meets the bill.

The honest expectation

Assume live will be worse than the backtest; the only question is by how much. A well-built system on good data might give back 10–20% of its backtested edge to real-world friction and still be worth running. A curve-fitted one gives back all of it and then some. If a vendor's live results happen to match or beat a spotless backtest, be more suspicious, not less — that is the unusual outcome, and a backtest is easy to fake.

An example

A scalping EA backtests at £2,000 a month with a profit factor of 2.2 on a fixed 0.8-pip spread. Live, the spread averages 1.4 pips, each fill slips a fraction, and the occasional requote kills a trade it counted as a winner. The same EA now makes around £600 a month at a profit factor of 1.3. It still works — but anyone who sized their account on the £2,000 figure is now badly overexposed and wondering what went wrong.

A normal gap and an alarming one

Some slippage between backtest and live is healthy and expected. What should worry you is the shape of the difference. A live curve that sits a little below the backtest but tracks its ups and downs is behaving; the friction is just costing you a slice of the edge. A live curve that ignores the backtest entirely — losing when the backtest won, drawing down far deeper — means the backtest was measuring something that was never there. The first is a cost. The second is a warning.

How to narrow the gap

Backtest honestly — real ticks, variable spread, commission and swap. A realistic backtest sets a realistic expectation.
Forward test — run it on demo or a small live account in real time first, on your own broker and feed.
Start small — trade minimum size live before scaling. The market will tell you things the tester could not.
Monitor closely — track live profit factor, drawdown and win rate against what the backtest promised, and watch for the gap widening.

You will never close the gap completely. The goal is to know its size before it costs you, not after.

The Karnek note

This is the whole reason live monitoring exists: the backtest is a promise, and the live account is what actually happens. Karnek reads your EA's real trades read-only from the terminal, so you can put the live profit factor and drawdown next to what the backtest claimed and see the gap for yourself. It can never place a trade or touch the account — it only tells you the truth of it.

See it in Karnek: monitor grid and martingale EAs, with drawdown alerts before they run away.