What is the Kelly Criterion for Sports Bettors?

While you will find numerous articles on coming up with techniques to determine whether a bet is worthwhile, you will not encounter as many articles that tackle the topic of money management.  Assuming you have discovered a bet is worth taking, how do you put a monetary value on it?  How do you decide how much to risk?

In several recent articles about common methods for betting on sports, I have been tackling this question by exploring a number of different systems.  Some of the systems I have talked about include progressive systems like Martingale and Fibonacci, fixed amount systems, and proportional systems where you strive to always risk the same percentage.

Another method for staking bets which is very popular is the Kelly Criterion.  This one is quite a bit different from the others I have delved into so far.  What is the Kelly Criterion and how does it work?

The Kelly Criterion is a mathematical equation:

(bp – q)/b

Where:

b = the decimal odds -1

p = the probability of success

q = the probability of failure

The easiest way to explore this idea is to look at a coin toss.  Assume that you are betting on one side at decimal odds of 2.00.  You have a 50/50 chance of winning, so your equation would be:

(0.50x1 – 0.50)/1 = 0.00.

In other words, the Kelly Criterion would suggest you not bet.

Now imagine that the coin is weighted against you as such:

(0.45x1 – 0.55)/1 =  -0.10.

When the Kelly Criterion gives you a negative percentage, that is again an indication that betting is a bad idea (a worse idea than if the odds were even).

Now imagine that the coin is weighed in your favour, not against it:

(0.55x1 – 0.45)/1 =  0.10

In this case, the Kelly Criterion recommends that you do take a bet, and moreover that you wager 10% of your account. 

Based on this, you should be able to understand the principle behind the Kelly Criterion and why it (sometimes) works. 

Whereas systems like Martingale and Fibonacci recommend that you increase your risk as you are losing bets, the Kelly Criterion recommends that you increase your risk only when you have a significant chance at winning.  Not only that, but it recommends that you do so proportionally. 

This makes a lot more sense intuitively than either Martingale or Fibonacci.  When the risk is high or the odds are even, the system further suggests that you avoid betting until the odds are back in your favour and you have a real shot at winning.

Unlike Martingale and Fibonacci, which recommend increasing bet sizes for reasons that have nothing to do with the actual real-life situation, this system recommends increases in conjunction with actual real-life reasons. 

So why do I still not give the Kelly Criterion a huge recommendation myself?

I am still not a massive fan of this system because the amounts that it recommends that you wager can quickly range pretty high.  Consider the example that I just used above with the coin toss.  Since the coin is weighed significantly in your favour, the 10% suggestion is not entirely unreasonable, and it could help you to grow your account quickly.  But losing 10% of your account on a single bet would be very painful.  And while it is not likely that you will lose 10 bets in a row and blow your account, it could still happen.

More relevantly, the real world does not work like that weighed coin toss.  With the weighed coin, you know the coin is weighed in your favour, and there isn’t going to be some random fluke that changes that.  Flukes are estimated to determine the outcomes of maybe half of all sports matches.  Do you really want to ever wager as much as 10% on something that could be overturned by a fluke?

I also am not a huge fan of the idea of wagering substantially more or less on a bet because it implies that you are taking a fair number of subpar wagers.  You should only be betting when you are really confident.  Start breaking with your discipline and you are going to have problems.

With all of those considerations, I do think the Kelly Criterion is a valuable tool.  You can actually use it to help you determine whether a given bet may be likely to succeed, and if so, how likely.  This helps you to figure out whether you should have confidence in a wager, and whether you should take it.  At that point, you can wager a fixed percentage which is comfortable for you and which is reasonably conservative (I always recommend 1-3%, maybe 5% at the very outside).  Learn more about that system in my instalment on Common Methods for Staking Bets, Part 3 under the heading “Proportional Betting.”