Read somewhere that like each stock or indicies, any ssytem would have its cycles......meaning times when system would give profits in a row and then go into drawdown periods. Developing simple things like moving average of the system results would help avoid those drawdown period and similarily be extra loyal to system in the profit cycle. Any inputs on this issue, if one is starting with a system with 70% winner trade ratio, then catching the upcycles can always bring us to 90% level.
I have done quite a lot of work over the past year on this. there is no doubt that it pays to have a defined out on a system at some point when it stops performing. A classic example were the trend following FX systems that worked so well in 04 but fell apart so far this year. If you trade a portfolio of systems and products then it is a good idea to have a stop on an individual system at a given draw. This figure can either be arrived at by your personal risk tolerance or by running some sort of historical simulation such as monte carlo on historical results to find a likely worst draw and ascertaining from this when you would consider the system to be no longer working. I think the former is probably the easier to bear in reality but it does need to be defined in advance and stuck to like any other discipline. Generally you would re-introduce the system again once it or if it puts in a new equity high. I hear some people draw trend lines on equity curves too but I have not been able to test this idea.
What has been jump in winner trades once you put an 'out' on system and upon new equity high throw an 'in' for trading. Monte Carlo simulation, where can one get tools to perform this.
Do not understand your first sentence. I use Tradestation for most of my work. Rina systems portfoliostream interfaces with TS and can perform the monte carlo and also apply lots of different MM algorithms. It is expensive however. There are other options out there, probably the most cost effective is "prosizer" from URC trading http://unicorn.us.com/trading/prosizer.html I have not used this myself but it looks like it does the job. It is not actually a complex process. All you are doing is mixing up the order of historical trades and seeing how different the profile becomes. This is essential appreciate likely drawdown and when looking at how you apply sizing algorithm.