Thoughts on Position Sizing

B

Babak

I was wandering through www.turtletrader.com where I stumbled on a concept called optimal f or "the Kelly formula" which is purported to be used to solve noise to signal ratios in long distance communication but is also applied to position sizing.

As money management or position sizing is an inexaustible source of learning for me, I was curious to know if anyone has used this in their trading and if it has helped.

More info is found here:
http://www.futuresmag.com/industry/downloads/downloads.html
[F4 and search for 'optimal f']

This master thesis was also interesting:
http://www.turtletrader.com/position_sizing.pdf

This explains the formula in detail:
http://www.hquotes.com/kelly.html
You've got to try their java applet...very cool!
 
Although I cannot answer your question with experience using it for stock trading I am familiar with it enough to know the concept applies to many different things.
I count cards (BlackJack) and anyone who does this effectively is using a bet size system based on or loosely based on kelly sizing. In blackjack speak it's known as "kelly betting".

For all you "gambler = snake oil" folks kelly is a mathamatical concept matching P&L potential to risk size. It wasn't developed/discovered by gamblers. :eek:
 
Take a look at books by Ralph Vince if you want to learn about "Optimal F" and fixed fraction position sizing.
You'll see that actually trading with the 'optimal f' number of contracts is way too risky and you will have drawdowns that are way too big for most. I like the Fixed Ratio idea that Ryan Jones talks about in his book a lot better.
These concepts are mostly used with futures system trading but the idea of increasing your contract size as your account equity increases and lowering your contract size as your account equity decreases(anti-martingale bet sizing), will benefit any type of trader. Of course, this will only work if you have a "positive expectation" over the life of your trading and you can withstand the drawdowns that will occur.

The beauty of this idea is that if you find a simple technique/system/method that is not impressive or exciting but is consistant with small profits; this is all you need! You can stop searching for big trades or 100% accuracy. Just perform your small edge over and over and focus on your position sizing.
 
I stumbled on something really interesting:

I went to http://www.hquotes.com/tradehard/simulator.html and then played with the variables in the equity curve generator. There are two variables that really matter. The win/loss ratio and the win probability.

I found that the win probability matters much more in generating a steeper (more profitable) equity curve than the win/loss ratio. I came to this conclusion by increasing each, one at a time, at specified increments. Like 20%, 40%, 60%, etc.

I thought this was really fascinating. It seems, from this atleast, that it is more important to be right and make some money repeatedly, than to make a lot of money (when right less often). Hope that made sense. Also, it seemed to not be linear. The equity curve was effected positively in an excellerated manner when the win probability was increased incrementally.

If you see the applet you'll know what I mean. BTW I used 100 independant random equity curves (variable called 'Lines Qty')

Am I missing something or is this generally true?
 
You must log in or register to reply here.
Back
Top