Sometime I think out a trading idea and then I back test it on one stock, which shows nice; after that I would test it on many stocks, by which at last I find the strategy runs well on several stocks but also failed on many other stocks. Shall I believe this strategy(only use it on several stocks)? or the strategy is just random and which should not be believed at all(that means we only believe a strategy which runs well almost all stocks)? Or another question: how can you confirm your strategy is ok to real trading?
Your Q is quite similar to this: https://www.elitetrader.com/et/threads/is-this-over-fitting.373413/unread Stocks have various behavior/personalities/characteristics ... They move in various manners/fashions/speeds... No such thing as one-size-fits-all-stocks things.
In order for the startegy to reach it's peak efficiency, all of the names involved, must meet certain criteria. Look for the similarities among the few that works and find more names like that.
Stocks have different profiles.You cant expect a thickly traded megacap name to act like some thinly traded microcap name. Adjust your strategy accordingly.
Thank you for your link and I review all its reply and I think if we can understand the reason why strategy not fit then we can be more confident about it.
But in many cases the differences are significant, and important. It may help you to pre-scan and create lists of stocks with different characteristics, prior to back testing and forward testing. For example, EPS (earnings), dividends, TR(Range/Volatility), industry, etc. etc. As just one example, at one time I traded a "pullback" system that looked for weakness in equities. I scanned and created a list of stocks that had strong volume, and strong and growing EPS (profits). The reasoning, which worked, was that buying into technical weakness on otherwise strong companies was preferable to just buying into a random list. This takes time, but you may find that there are one or two specific characteristics that respond well to your strategy, or to a new strategy.
Could you give me an example of how it worked on a specific stock and not the other? I think each stock has a specific personality based on volitility,overall strength, institutional support. A great example would have been INTC and AMD.
You can't. All you can do is make an educated guess. You're flying blind unless you have some idea of what edge or factor the strategy is actually exploiting. Most simple retail strategies are momentum-based and thus work well on hot stocks in bull markets, and get destroyed otherwise. Or, maybe you're trading an RTM strategy on REITs or utilities; that can work, but beware of the risk factors (interest rate risk in particular) which can affect that strategy and render a given setup invalid. Remember that market prices and buying/selling activity are a product of real world factors specific to any given historical moment - the stock market isn't just a random-number generator.