Lessons Learned From Profitable System Development

When I build a system that's almost okay, but gets whacked on commissions, I start cutting out periods of the day (most often lunch) and work from there.

Also, if it's some sort of xover system (no matter what the indicator) I'll try smoothing it with something fast (TEMA, or JMA). Sometimes the differences are dramatic.

regards,
laz

 

pivot points. see this thread.... http://www.elitetrader.com/vb/showthread.php?s=&threadid=21377

 

I'm sharing the basic formula I used to create a decent countertrend system in hopes that others may glean something from this process. I'm fairly certain that anyone could use this process with a trend following system they are designing and get good results out of it as a countertrend system.

First off, I started out creating a trend following system. After going through the results and looking at what the system was doing I made a couple of observations:

1) There is a specific open profit size for a given trend following system which is kind of the threshold between countertrend and trend. After the open profit has grown to at least this size, the system is in trend mode and the trend trade will almost always make $$ based on the regular system exits. I observed this point by simply running through the trend system's trades. For each system and market that the system is applied to, this # will probably be different.

2) Some of the best counter trend trades occurred right after the trend trade in #1. Counter trend trades are simply the opposite of the system's signals (when it says buy, you sell and vice versa) In fact, the most reliable system was observed to be taking the next trade after the big trend trade (opposite of the signals the trend system is giving), and continuing to take the trades until a loss occurred. I put in a stop loss to limit the loss as often the loss was the next trend trade (#1), which can turn into a sizeable move.

Something else I'm looking at is that you can (most of the time) reverse on the stop loss and make up the amount of the stop loss on the trend trade. This may be something more geared to human intervention but my observation was that even if mechanically applied as a stop and reverse at the stop loss point it is worth doing.

I think the reason why this works is because of the natural tendency of trend following systems to go through long periods of whipsaws (ironically this was my dilemma a few posts up). Comments?

 

great post but would like to see an equity curve

 

Some who have posted here, trade around their systems. I'd like to hear how you approach this. The assumption I'm making is that by doing so, you are outperforming the system's signals (otherwise just auto-trade the system).

For instance, do you take the entries but manage the exits manually? Or do both manually? How do you deal with it when the system enters and the market moves in the system's favor, and you are on the sidelines? Do you exit trades and look to reenter? When do you do this? Do you skip signals? When do you do this?

Any insights would be greatly appreciated.

 

I trade around my system, but over the long run, it has cost me profits. In the short run it often feels right and reduces the stress level.

I will sometimes allow entry levels to pass when it looks like I can improve on price. More often I exit sooner than the system when it looks like gains may be given back (wrong more often than not).

System is beating me.

g

 

I've actually traded signals before. But one night when I was drinking some beer. I noticed that I'm supposed to trade the market not a damn signal that I created. After some riff-raffs within my head and actual trading results...

It was lack of discipline and ego towards my system that I created it and should trade better etc. etc.. I trade the market by creating a system to make money.

 

So bottom line, you found your system outperformed you, so you came to the conclusion to just follow the system?

 

Great post Smitty! I have a backtesting system that allows me to throw any system at a universe of markets and run various parameter sets over varying time frames. I then chart the results, and not a whole lot of systems really work that well. (Some do, of course, otherwise we would not be doing this (trading) at all.)

Lets say it again: The first step is finding what doesn't work.

 

I'll second that (or third it ). I've spent years working on mechanical systems. Most of the conventional TA wisdom isn't worth squat. But it has sold countless books, magazine articles, seminars and web site subscriptions.

g