Tuesday, February 9, 2010

J200 and Normal distribution - PART II

I’ve received several questions about previous post (J200 and Normal distribution) from my trading friends on Skype and on one of the forums where I am active (www.page88.co.za), so I have decided to respond here on blog.

If we assume that we have ideal trend following end-of-the-day (EOD) trading system (100% win rate), then that system is going to mirror market and will produce results that are not normally distributed, as our market is not normally distributed when it comes to difference of daily price changes at market close.

However, as our trading system is not perfect I’ve decided to test distribution of trades for that particular system and for additional eleven trend following trading systems that I trade and/or paper-trade (5min-15min-30min-EOD). For all of them it was more than clear that trade distribution is not normally distributed, with systems successfully relying on market extremes and successfully catching market events during which market reaches extremes several standard deviations from the mean. Of course, this kind of distribution catches some big fat-tail losers too, but as long as these systems continue to catch more positive fat-tail events, they are going to continue to perform. (Please note that these fat-tail events can result from daily returns and/or follow through or can be result of massive one-directional short-term burst in volatility.)

One of the items on my to-do list is to look at normally distributed trend following systems with 60%+ win rate, just in order to have few different machines when compared to these twelve systems I follow, as for all of them win rate is somewhere between 30 & 45%, with average win/average loss ratio being between 2 & 3.

When it comes to mean-reverting strategies and their distribution – I am not actively exploring anything in regards to mean-reversion and I am not willing to comment on this, as I haven’t done homework yet.

Finally, all of this gives me nice platform to come to the point that I wanted to make since I started with original post:

If one’s system has proven (with enough data sample; over many market conditions; with enough out of sample data, etc.) to be able to provide trade entries that over time yield positive results through fat-tails, then trader has to be extremely disciplined in letting profits run, as over time system will lead to positive results. All of this, of course, until system breaks down and that is something that I will be writing more on in the future.

Proper money & risk management are to be employed, as almost always trading great system(s) with too little capital will lead to ruin of one’s trading account.

Trade with trend!


Attached graph shows trade distribution for one of the variations of our 30min system for period: 19.12.2007 – 03.02.2010.