I'm a big fan of MACD for longer-term position trading and find that with my own parameters it works very well for me for strong trending stocks. I bought Appel's book (Appel being the guy who 'invented' MACD), and he recommends 9/16/6 for people wanting to use it for shorter cycles, although I haven't tried that. What's interesting is that the simple signal line crossover isn't itself a big signal for him: MACD divergences for example are at least as important. I found this interesting as when I started using it, it was in a very simple, 'when it goes from green to red, sell, and from red to green, buy' kind of way.
Personally I prefer it the other way. I find it whipsaws a lot on the standard setting, so I revised the parameters to suit my trading style: I'm happy to miss either end of a stock's run if I get a good solid trending stock that might still give me 20-30% over three to six months: so I screen for stocks that have historically shown strong trending characteristics (up or down) and set my MACD parameters to 25/65/18.
I've also used 25/65/18 on intraday index trading with some success, ie looking for the bigger moves on 30-minute data, although to be honest I'm moving away from an indicator based system for intraday now. But you might want to have a look at this: eg if you hold stock for 3-20 days then try 25/65/18 on one hour data? Just a thought.
