EA's DO work
The main reason an ea does not work, is because it was made by someone who doesn't know the basics of risk managment in Forex, OR, they do and they let the bot take extreme risk to inflate the account to entice people to "buy" the ea and make their own money. I am not against selling well made ea's, though I managed to get my hands on a few costly ea's, and they failed terribly, and 90% of them fail terrible on back testing.
There is this idea that backtesting doesn't work, it absolutely does, it is testing the automated strategy on past market conditions, you could argue that it does not represent future market conditions, this is partially true, but what else do you have to analyse? You have to backtest, it is a necesity for developing an EA, then you move on to demo account testing.
I opend one ea in particular to view the code, because it claimed to be "zero risk", and in backtesting it did really well. The EA MAYBE had 20 or 30 lines of code, probably half of that was whitespace. After testing this ea on a $1000 demo for one day, the account lost 99.9%, it busted.
Though I am currently copying the signals of an ea, I do not have access to the code, just the autotrading signals. It has ran for a week, it has over %50 profit. Starting with $1000 account. It has a risk of ruin rating, in losing 10% of account balance of 0.01%, meaning it is very unlikely it will loose even 10% of account balance
This is an exceptional EA, and I really want to see the code or at least an explanation of how it works.
I believe, the best EA's are very low risk, with very high "z-score"(ability of trading system to generate wins and losses in streaks), they must also be able to at the very least be able to perform well with a low starting balance of $1000, if they can build a balance of $100 this is even better.