Saturday, March 22, 2008

How Much To Invest?

One of the most difficult questions in investing is when you find a stock you like, how much do you put in?

Most people determine this based on their conviction; The more confident I am about a stock's performance, the greater the weight it will have in my portfolio.

I offer you a more methodical strategy. Enter the "Kelly Rule".

"How much should I stake - Kelly's strategy

The answer to this question has quite surprisingly been around for ages though it is discussed, analysed and refined often. Named after its author, John L Kelly Jr, if was first published in 1956.

Quite simply, the strategy enables you to find the answer to a question that is common amongst speculators and gamblers. If you have a bank of X, how much should you stake on each occasion to maximise your gain but minimise your loss so that in the long run you can perpetually increase your wealth!

Kelly considered the strategy of betting a fixed fraction of the bank on each occasion. In a favourable game, your fortune ought to grow exponentially, like compound interest. He worked out the way the rate of growth varied according to the fraction you bet. If you bet only a tiny fraction you will not go bankrupt but your wealth grows very little. Make the fraction large and the losses when they occur will wipe you out.

Kelly's answer was simple. The right balance is struck when the fraction you the fraction you bet exactly measures the size of your advantage. If you are being offered even money but the chance of an event occurring is 51% (and therefore the chance of failure is 49%) you should bet the difference between the two, 2%.

Therefore if you started with a bank of £100 and you correctly assessed your chances as discussed above you would need to place a bet of £2. Any higher and your probability of liquidation (risk) increases exponentially to your likely return."

Via Probability Theory


1 comment:

Kyle Wolfe said...

This is very interesting and makes good sense, but how does Kelly come up with the 51% and 49% numbers? I feel these numbers could be subject to forecast risk, however they may be based on probability. If someone could explain I would appreciate it.