There's nothing really wrong with the concept of backtesting, if used sensibly. Lets say you look at a chart, and you form the opinion that price bounces between the upper and lower bollinger bands. So you write a script, backtest and find its not profitable. The problems start when you modify the original idea, so maybe you change the period length, or the deviaton, or maybe you add a stop loss, or scale in, or scale out, or add an extra indicator etc until such point that you get a decent looking equity cureve. At that point, you have a serious data mining bias problem.
If you use backtesting to determine things like max drawdown, or to see how something performs in under particular market conditionss, then there's some mertit, but even thats a bit pointless as markets are non stationary.
To use it as a tool to find ideas that make money is the dumbest thing imaginable. The moment you add in any form of optimisation you are on seriously dodgy ground.
The other issue of course is that these tools are commercial software products, written to satisfy the demands of users (and the users by and large dont really know what they want), in the worst cases, they are tools provided by people who are taking the other sides of your trades !
Apart from that, there are issues with most of the commercial products, bad data, holes in data, indicators not calculating correctly, bizarre quirks in the way the backtester works etc. However, the main point I was making is that its extremely unlikely that a product thats extremely simple to use, and requiring no specialist knowledge is going to return anything meaningful.
