This is because many 'traders' go directly from backtesting to real-money trading. They skip the walk-forward testing and real-time paper trading. Like previously said, there are so many ways to backtest the wrong way.
Yes, backtesting can be very useful. I am on year three of a very robust system that has conformed perfectly to initial backtesting I completed in 2014. From my experiences, backtesting can be much more predictive when applied to a specific instrument, especially an index, or index-related derivative, than when applied to a basket of securities. Good luck.
If you backtest 20+ years, 5000+ trades, profitable with an array of similar stocks or etfs and sharpe of 2+ you can be confident that’s it’ll work in live trading.
%% You must study trends also.Better than backtesting. But since we do NOT know much of the future, back testing helps- not on a five minute chart. LOL IF any make a profit on a 5 minute chart= fine with me.
I can say from personal experience that just because you find something that works now through backtesting guarantees you nothing in the future. There are also other risks that don't get mentioned enough when system trading. Things like what to do when a system does not win or lose but simply chops around break-even for an extended period of time. Things like how should you decide when to "offline" a system? Sometimes it is not as simple as "I will stop trading it when it stops making money". You will get chopped up by constantly turning systems "on and off". In many cases these systems are doing nothing more than participating in a limited fashion in the overall underlying trend. This is why your longs in your ES/SPY system will probably outperform your shorts. The same is true in other markets. Being able to trade based on feel/tape reading and other things that are harder to put into a system and measure is much more powerful than system trading. I know many people don't want to hear that. A trading system has a shelf life that is probably just a function of the underlying yearly trend. "Gut feel" for the markets is the last powerful skill now that everyone and their brother is a data scientist/system builder.
Hand backtest. Then your brain will actually learn from the experience, offering up better ideas than you had originally started with. [NoDoji did this all the time...gave her her edges]
Like curve fitting? Question for you: Instead, I should start with incorporating/finding the fundamental factors that affect price?
Golden advice! I speak of personal experience. I learned important things that I would never find while "numbercrunching". "numbercrunching" will pop up with a result, that's all. You should understand why some trades are good and some or bad, how the sytem generates the wrong siugnals (for the bad trades) and how to avoid these signals. There comes the "10,000" hours from. People want to reduce the "10,000" hours by letting the computer do the job. That's not working. You should first understand the markets and use that knowledge as input for the computer. The intellectual part should be done by a human brain.