Some MQL5 EA backtests look perfect because the EA already knows what the market did. I've been testing EAs sold on
the MQL5 Market, the popular ones near the top of the list. Most behave like normal trading systems. A few don't, and
they fall into two groups.

Group 1: what's a backtest with the answers built in?

These EAs look into the future. The EA carries past market data inside it, so when you run it in the strategy tester it
doesn't have to guess where price goes next. It already knows, and it just looks up the answer.

That's why the backtest looks unreal: win rates of 97–99 %, profit factors of 30, 50, sometimes more, an equity curve
that's nearly a straight line.

The trick shows up once you change the test a little. Shift the prices slightly or move the dates by a few days, and the
EA can't find its answers anymore. The same backtest falls to around break-even, and below that once spread and
commission come out.

Live trading happens in the future, and the EA has no data for the future. So what you're left with is the strategy
underneath, and in every case I tested that strategy had no edge. Several also use a big stop loss with a tiny take
profit: a long run of small wins, then one loss that takes a large part of the account.

These authors also release new versions often. Each new version adds the weeks since the old one, so the backtest
looks perfect right up to the release date. It never covers the days after, and those are the days you trade.

Group 2: what does "sales page vs the real EA" mean?

Here the backtest is honest. The problem is that the sales page or the author's signal shows something the EA you buy
doesn't do. A signal that runs settings you don't get. A grid with no stop loss sold as "capital preservation". An "AI
SuperTrend" EA where SuperTrend is only drawn on the chart.

You find out the same way, by running the demo yourself, and the result for the buyer is much the same: a slow loss or a
blown account.

So what's my advice?

Don't buy EAs like these. Theres no setting that fixes them.

What comes next?

One EA per post. Every title says which group it's in: "a backtest with the answers built in" or "sales page vs the real
EA". Each post shows what the page promises, the test that exposed it, and how you can run that test yourself.