I agree that most people are better off doing something else. But that's because most people don't understand that backtesting is a lot more than coding an idea and running it on historical data. One has to deal with nonstationarity, survivorship bias, modeling error, risk management, money management, etc. But it's wrong to categorically say that "back testing simply cannot work" as you did... maybe that's true for the definition and expectations of backtesting you had in mind though. Does this give you pause? He's still doing very well with technically based systems. "I know of no way to validate conjectures concerning technical trading without back testing" --William Eckhardt
"I know of no way to validate conjectures concerning technical trading without actual trading" --jimbojim Thatz too stupid of you trader6666. Backtesting cannot validate anything. It can only provide HOPE. Eckhardt I don't know who he is but he sound like a paper trader. Conjectures can never be validated anyway. Too stupid to say that. Conjectures can only be falsified. My conjecture is that trader666 is a paper trader. If he displays a real trading record my conjecture is falsified. Good Lord...
manual, discretionary trading can also provide only HOPE that it will generate profits in the future. Eckhardt manages over 300 million dollars and is profitable over many years. let's hope we will always have thousands of dumb ignorants like you - fool and his money need to be separated resulting in money flow into smarter hands ie. eckhardt.
My conjecture is that you're a fool, especially since you wrote your drivel without even knowing who William Eckhardt is or bothering to google him... good luck idiot, you'll need it.
Actually it's $700 million. I gotta laugh when people on ET think they know better than someone like Eckhardt when the fact is that nobody who's ever posted on ET is even close to being in the same league. And in jimbojim's case he thinks he "knows it all" but never even heard of him
Those of you hating on optimization: I do know what you mean by excessive curve fitting but almost nobody talks about under-optimization. Optimizing parameters is not inherently evil and should be done on a continuous basis to prevent system decay. Also, a system blowing up isn't a bad thing as long as its done responsibly. You have to be proactive about pulling the plug when the edge dissipates before you trade your profits away. Sometimes its earlier than you would like, but you have to manage the risk as it comes.
You are correct regarding the experiments. Just do not know what has happened to the fellow who started them all, was doing a good job. Hopefully someone else can fill up the gaps.
You are correct, JaiSreeram, they were indeed good. Is that guy around? Anyway, let us hope that someone else "fill up the gaps".