How to pick an EA without losing your shirt

I get asked some version of “is this one any good” several times a week, usually with a link to a sales page and a screenshot of an equity curve going up at forty-five degrees.

I can’t test everything. But most of what you need is on the sales page already, if you read it in the right order. Here’s mine.

Start with the account, not the returns

The first thing I look for isn’t a number. It’s whether there’s a verified live account at all, and how long it has been running.

Not a backtest. Not a demo. A real account with real money, on a third-party verification service, with a history you can scroll back through.

If there isn’t one, you’re done. Everything else on the page is a claim about software that has never been tested against a real spread, real slippage and a real broker deciding whether to fill you. That is not a small gap.

Then check how old it is

Fourteen months is the number I care about, and it’s not arbitrary — it’s roughly how long you have to run before you’ve seen a few different market regimes.

Anything under six months tells you almost nothing. Plenty of strategies look brilliant for a quarter. Grid systems in particular can print money for months before the one move that takes it all back, which is precisely what makes them so easy to sell.

Look at the deposits, not just the balance

This is the one most people skip, and it’s the one that catches the most dishonesty.

Pull up the account history and look for deposits. Then check the dates against the drawdowns.

A percentage return is a ratio. You can fix a ratio without making any money, by quietly increasing the bottom half of it. An account that gets topped up during every bad patch will show a healthy percentage forever, while the person running it loses money continuously.

If deposits keep landing mid-drawdown, the headline number is decoration.

Read the drawdown as the real price

Most people read the profit figure and glance at the drawdown. Do it the other way round.

Your maximum drawdown is the amount you have to be able to sit through without closing everything at the worst possible moment. Whatever number is on the page, assume you’ll experience worse, because you’re buying it after its good run, not before.

And ask what produced it. A 30% drawdown from a strategy that takes a stop is a different animal from a 30% drawdown from a martingale that hasn’t blown up yet. The second one isn’t a drawdown, it’s a countdown.

Find out what it actually does

You don’t need the source. You need to know whether it:

• adds to losing positions (martingale)
• runs a basket of open trades with no stop (grid)
• holds through news
• relies on tight spreads at a specific hour

None of these are automatically disqualifying — plenty of people run grids deliberately and know exactly what they’ve signed up for. But a vendor who won’t tell you which category their thing falls into has told you something anyway.

Check whether the developer has a past

Search the name. Search the product name plus “scam”, and then ignore the top three results, which are usually affiliate pages using the word to catch that exact search.

What you’re actually looking for is a previous product — a rebrand, an EA that used to be sold under a different name, a developer who has done this before with something that no longer exists. That pattern is common, and it’s the single most useful thing you can find, because it tells you what happens after the sale.

What I’d ignore entirely

Backtest screenshots on the sales page. Anyone can produce a beautiful one. Optimise hard enough on past data and you’ll get a curve that goes up at forty-five degrees and falls apart the moment it meets a week it hasn’t seen.

Percentage-per-month claims. Meaningless without the drawdown and the deposit history sitting next to them.

Countdown timers and licence-count warnings. “Only 3 copies left at this price” has been true on some of these pages for four years.

Reviews on sites that sell it. Obviously.

The actual test

If it survives all that, and you still want it, run it on a demo for a month before it sees a penny of your money — and then on the smallest live size your broker allows for another month.

Yes, that’s two months of not making money. It’s also the cheapest possible version of finding out, and the alternative is finding out at full size.

Most of them don’t survive the demo month. That’s not pessimism, it’s just what happens when software written for a backtest meets a real spread.