This is a very good question and there’s no easy answer. You’ve already identified the first step, selecting the period of time over which you’d like to hold trades. Its probably worth thinking about what sort of returns you’d like to see, the drawdowns and volatility in equity that you’d be happy with, what trading capital and resources you have etc. Its not just a question of finding a system thats profitable, all these other aspects are equally important.
I started by demo trading other peoples systems, either stuff from forums, or purchased commercially. I ended up designing my own as not surprisingly, there’s very little out there that actually works. I tried hundreds of systems, and nothing worked.
In terms of how you start, I’d suggest that for trend or breakout styles, the first stage is retrospectively identifying trades on historical charts, and then attempting to work out why those moves may have occurred. Thats going to take you a hell of a lot of work, and I’m afraid that you wont find anything on any forum that going to help. I’d start from a fairly simple perspective, for example, was the entry with the major trend, did the entry occur at previously identified support or resistance, did the entry coincide with news, is there a technical set up occurring at the time of entry etc, was the entry a breakout from an established range. If your using TA, multi timeframe analysis is reccommended.
Once you’ve identified entry set up and triggers, you need to think about stops, targets and trade management, position sizing (and probably in that order). Its an itterative process, and you need to be aware of how all of the individual aspects of the system operate synergistically. Its basically back and forward testing. This may be harder for you as your trading longer timeframes, so it takes longer to achieve statistically significant results.
If your looking at hedge style systems, or volatility grids, or arbitrage opportunities then the design approach will be different, but I’m assumming that for your first system you’d want to start with something simple.
If you really want to base your system on technical analysis, to be brutally frank, there’s not that much to choose between different indicators, and if you get the design right, pretty much anything will work. I’d probably suggest that you break down indicators into categories, for example something to identify tend direction, trend strength, momentum, volatility etc, and pick one indicator from each class. If you take the design process seriously, you’ll need to address how you assess the performance of your systems, and once you have this framework in place, its easy enough to try out different indicators and parameter settings.
When you see a system proposed on a forum or in a book, ask yourself why particular indicators are being used, and ask yourself if they are actually doing what theyre intended to do.
It mght be worth you picking up a copy of Van Thaps system design course. It was a couple of thousand bucks when I bought it, but I believe its a lot cheaper now, possibly even available for free from a torrent site. Van’s book trade your way to financial freedom has a simple design methodology that you could follow, as does John Murphys book on TA.
I hate reccommending Van Tharp, but his stuff on system design is a reasonable starting point, and it provides a few ideas on the different trading styles that are commonly used.
A reasonable knowlege of statistics and probability theory certainly helps, as does some discretionary trading experience. I started off designing mechanical type systems and I totally failed. After a couple of years of full time discretionary trading, I understood a lot more about the problem I was trying to solve, and only at that stage was I able to actually start making progress on mechancal strategy design. Its sad to see people that I was working with back in 2002 still releasing EA’s in forums and newsgroups for test, and they are still no further forward, simply because they dont understand the problem that they’re trying to solve. Randomly throwing a bunch of indicators on a chart and hoping to strike it lucky isnt a sensible strategy. You need a design methodology and a process.