[cubsker]

Just read The Black Swan by Taleb who loves to rip the Bell Curve. Is sports betting Gaussian and how would you go about quantifying such a thing especially when it comes to betting sides? Clearly, you could look at past distributions of points scored to look at totals and see if it follows the bell curve. Any thoughts on this or past threads you can recommend?

[Dark Horse]

I sense a Ganch appearance.

As a separate comment, would it be helpful to have a SBR thread with Wikipedia links to probability theory? Including links to threads where each topic was discussed?

http://en.wikipedia.org/wiki/Gaussian_distribution

[RogueJuror]

I dont want to pollute the think tank

or ruin this thread


but dark horse

what is this avatar about?

sinceryly, a drunk fan

[Dark Horse]

That's Nataraja (Na-Ta-Ra-Ja), the Cosmic Dancer.

lol

[Wrecktangle]

Just read The Black Swan by Taleb who loves to rip the Bell Curve. Is sports betting Gaussian and how would you go about quantifying such a thing especially when it comes to betting sides? Clearly, you could look at past distributions of points scored to look at totals and see if it follows the bell curve. Any thoughts on this or past threads you can recommend?

Which info set are you talking about? NFL scores are absolutely not as are no other sports scores I can think of (all variants of poisson at best). Since most stat tests work off Gaussian dist. many folks like to think they are Normal. Some folks think totals are Gaussian "like" and can be approximated by it but they "drift", so no, I'd say hardly any sports distributions related to scoring are Gaussian.

[tweek]

As for baseball, I think team runs scored follows a Weibull distribution. But, I don't think this is necessarily what you're after...?

[cubsker]

Which info set are you talking about? NFL scores are absolutely not as are no other sports scores I can think of (all variants of poisson at best). Since most stat tests work off Gaussian dist. many folks like to think they are Normal. Some folks think totals are Gaussian "like" and can be approximated by it but they "drift", so no, I'd say hardly any sports distributions related to scoring are Gaussian.

Thanks for your thoughts. I wish I had continued my math education after calculus and knew a lot more than I currently do. The reason I'm asking is b/c I am wondering how models are developed. Don't these models usually use regression which relies on normal distributions? If I'm way off, please say so and if you can, point me in the right direction.

[cubsker]

As for baseball, I think team runs scored follows a Weibull distribution. But, I don't think this is necessarily what you're after...?

I'm not really sure what I'm after. I'm not familiar with other types of distributions, so I guess I'll start with that and see where it goes.

[maxdalury]

i mean i dont' really know what your talking about. hockey and soccer are approximately poisson, you can probably use a gaussin distribution as a starting point for basketball but you are gonna have to do lot more work than just using a standard prob distribution to make money.

[Wrecktangle]

Thanks for your thoughts. I wish I had continued my math education after calculus and knew a lot more than I currently do. The reason I'm asking is b/c I am wondering how models are developed. Don't these models usually use regression which relies on normal distributions? If I'm way off, please say so and if you can, point me in the right direction.

Most regression techniques are not necessarily based on Gaussian dist. (least squares, typically) but many folks look at residuals to see if they are Normally distributed to see if the fit is good. If the residuals are not bell-shaped, your model "may not" be decently fitted. From that stand point some folks think standard regression techniques are based on Normal dist.

If you are using Stepwise Regression to pull in/out factors, an F dist. is typically used to test for inclusion/exclusion.

There are all sorts of URLs on the net talking about this stuff, you need to do a search.

[cubsker]

will do, wrecktangle. thanks for your thoughts.

[pico]

i see black swans all the time. usc lost to ore st. i remember usc lost to ucla on sat then colts lost to titans on sunday the day later, both are -400 to -500 chalks. j blake lost in straight set with -500 opening line and getting chalkier as the line approach closing. the list goes on.

i thought about statisitcal distribution and sports events for awhile, my impression is that the odds is basically market perception of the sports events and we don't have enough data to quantify a single game. maybe you can use statistical methods to predict season totals, but the book who set the lines already has people crunching the numbers, so your model (if it is working), should pretty much be in line with most season odds posted. now if you found couple of team lines are way off then it can either be 1) your model is crap, 2) you forgot or don;t know some key info, or 3) market is off. with major sports leagues, chances are your model is off.

i don't think you can predict a single game. predict single game is like guess whether dow is going up or down the next 5 minutes. unless you have shit of money and know you're going to move the market, you can't really say for certain.

i don;t think regression works that well either even for totals. you're gonna have data like team A average 4.5 runs, team B average 4.5 runs...the total is around 9...what the hell are you gonna do?

[Dave Head]

"The Black Swan" is trash. It's not math it's ideology.

The Wall Street emperors have no clothes on and Taleb is offering them his ideology as a fig leaf to wear over their private parts. Now they can say "We didn't sell you out for a quick buck. We were misled by incompetent statisticians who didn't understand how useless their mathematical models were". Bull.

[durito]

a -500 loser is far from a black swan

[pico]

a -500 loser is far from a black swan

how about losing -1300 chalk back to back in a single day

[durito]

nope

[billhat]

its always gueein