Friday, October 13, 2006

Picking Winners

With all due respect to Jedi, and I congratulate him on his week, lets take a good look at picking NFL winners.

It is obviously not enough to win 50% of your games (unless you are the sports book). Lets say for arguments sake we are playing at a normal sports book with a 10% vig (or juice) on your loses. What would an expected win percentage of 50% earn? Let’s use the following formula assuming a $100 bet:

(winning the bet * probability of winning) – (losing the bet * prob of losing)

(100*.5) – (110*.5) = -5 (dollars in this case… on average you should expect to lose 5 dollars)

So how much do you need to win just to break even? Let’s set this win percentage at “X”

(100*X) – (110*(1-X)) = 0

Do the math (or take my word for it), and X = 52.4%. So a winning percentage of 52.4% will equate to breaking even while covering your juice.

Now let’s think about those that claim to always win. Let’s say that someone is able to accurately predict 60% of the time. That’s 6 out of 10 winners at a consistent basis. This sounds like a fraction of what some of these guys claim, but let’s be humble with our assumptions just to prove a point. Our expected value for one week of football action is:

(100*.6) – (110*.4) = 16

Multiply this by 17 weeks in the season, plus 4 weeks of post season equals 21 times to take advantage of the 60% win percentage. This equals: 16 * 21 = 336. That is a 336% expected return on your asset over a six month time! That is insane! There are many of you reading this who understand finance. What would you do if someone claims a 336% expected return on their asset? If you had the ability to more then triple an investment wouldn’t you try? The answer is yes, leading to capital mobility on the winning side, pushing the lines toward the expected outcome of the game.

In short the market self corrects, and should render these predictions useless in a short period of time. To come up with a sustainable winning probability this high requires some serious mathematical modeling. And I don’t know for sure, but it doesn’t read to me like these guys are running any regression analysis to come to their conclusions. Again, no offense.

No comments:

Post a Comment