Why Do EAs Fail?

Most expert advisors that look brilliant on a sales page lose money in a real account. The reasons are boringly consistent, and almost none of them are bad luck. An EA fails when the conditions it was built for stop holding — because the market changed, because the backtest was never realistic, or because the strategy had no way to survive a bad run. Knowing the common failure modes is the fastest way to spot a system that will not last, before you fund it.

The gap between backtest and live

A backtest is a strategy's best possible past. It runs on clean historical data, often with a fixed spread, instant fills and no missed ticks. A live account has none of that. The most common reason an EA fails is simple: its real results never come close to the backtest it was sold on, and that gap was built in from the start. Most of the failures below are just different ways of saying the same thing — the conditions the EA needed were never quite real.

The main reasons EAs fail

Curve-fitting — the settings were tuned so tightly to past data that they describe noise rather than a real edge. It falls apart the moment the market stops repeating itself.
A changing market regime — the strategy was built for a trending or a quiet market, and volatility, ranges or correlations shifted. Nothing broke; the conditions simply left.
Unrealistic backtests — fixed spreads, no slippage and instant execution flatter every result. Real spreads widen around news, fills arrive late, and the edge disappears.
Martingale and grid blow-ups — systems that add to losing trades show smooth equity for months, then hand it all back in one trend that never reverses.
VPS latency or downtime — a slow or offline server means missed entries, late exits and orders that never fire. The logic can be sound and still lose money on execution.
Broker differences — different spreads, swap, execution and symbol names mean the same EA behaves differently from the broker it was tested on.
No risk controls — no stop loss, no drawdown limit and no cap on size turns an ordinary losing streak into a wipeout.
Vendors that stop updating — an EA is not finished software. When the seller disappears, nobody adapts it to the market it now has to trade.

How to cut the risk

Forward test on demo first, then on a small live account. A strategy that survives real spreads and real fills has earned a little trust.
Insist on a live track record rather than a backtest alone. Months of real trades tell you far more than any optimised curve.
Set a hard drawdown limit, and decide in advance what you will do when it is hit.
Prefer fewer parameters and a strategy you can explain. Complexity usually hides fragility.
Monitor every account continuously, so a stalled or misbehaving EA reaches you in minutes rather than at month end.

Common red flags

A flawless backtest — a near-straight equity line with tiny drawdown.
A high profit factor built on very few trades.
Martingale or grid mechanics sold as "no stop loss needed".
No live results anywhere, only screenshots or a demo.
A vendor who guarantees returns or will not discuss the worst month.

Why it matters

An EA that fails rarely announces it. It keeps running, the balance drifts, and by the time you notice a month of gains is gone. Most of these failure modes are avoidable with two dull habits: test before you trust, and watch the account after you fund it. A strategy that survived a proper forward test and is monitored every day can still lose — but it will not lose quietly for weeks while you assume it is fine.

The Karnek note

Most EAs do not fail with a bang. They stop trading, drift into a drawdown or quietly miss entries, and you find out weeks later. Karnek watches the live terminal and alerts you when an EA stalls or a drawdown creeps toward your limit, so you can act while it still matters. It reads read-only and can never place a trade or touch the account.

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