Sports not Gaussian?

[vchillin]

I've seen this asserted by user Wrecktangle in various threads and am interested in an elaboration/explanation of this. A relevant link would be sufficient, but I'm looking to have the assertion defended. Thanks in advance.

[hutennis]

I've seen this asserted by user Wrecktangle in various threads and am interested in an elaboration/explanation of this. A relevant link would be sufficient, but I'm looking to have the assertion defended. Thanks in advance.

Well...
Sports not Gaussian. Thats negative statement, is not it? Why would anyone be required to prove negative?
Burden of prove is on someone making a positive statement. Like Sports are Gaussian. And if it is ease to prove -
then great, just do it.

It is one thing, though, to operate under assumption that something is true and quite a bit different to have
an indisputable evidence for it. They've been operating under assumption and making clt models in finance markets for years. We all know how that turned out.

I personaly am quite skeptical myself. It's simply to many ever changing factors and too much of a "human nature"
is involved to be treating sports the same way we deal with defective parts on a production line.

[juuso]

It is. Everything is "gaussian".

[hutennis]

It is. Everything is "gaussian".

Assertion without prove can be dismissed without prove.

[juuso]

Assertion without prove can be dismissed without prove.

Pretty much everything, except chaotic quantum events, can be presented with probability distributions. Sports is no exception. Each score and outcome has it's own probability that fits nicely in a distribution. Problem lies in making good enough model for reasonably good approximations.

[MonkeyF0cker]

Well...
Sports not Gaussian. Thats negative statement, is not it? Why would anyone be required to prove negative?
Burden of prove is on someone making a positive statement. Like Sports are Gaussian. And if it is ease to prove -
then great, just do it.

It is one thing, though, to operate under assumption that something is true and quite a bit different to have
an indisputable evidence for it. They've been operating under assumption and making clt models in finance markets for years. We all know how that turned out.

I personaly am quite skeptical myself. It's simply to many ever changing factors and too much of a "human nature"
is involved to be treating sports the same way we deal with defective parts on a production line.

It's far easier to prove something untrue than true.

[MonkeyF0cker]

Pretty much everything, except chaotic quantum events, can be presented with probability distributions. Sports is no exception. Each score and outcome has it's own probability that fits nicely in a distribution. Problem lies in making good enough model for reasonably good approximations.

Scoring distributions in football?

[betplom]

It's far easier to prove something untrue than true.

Prove it.

[brettd]

Pretty much everything, except chaotic quantum events, can be presented with probability distributions. Sports is no exception. Each score and outcome has it's own probability that fits nicely in a distribution. Problem lies in making good enough model for reasonably good approximations.

Agreed. However I think what Wreck is trying to get at is that the sport and market changes over time. 'Static' models built on interpolating from distributions will eventually (however long that takes) fall apart.

If all you did was to infer from a distribution from an ever increasing population of data, that inference would eventually go to shit in sport. It's not like we're looking at 'absolute' natural phenomena here, (like tree rings or some such), which would lead to an ever greater confidence/precision. Which is what gaussian statistics was designed for, and excels at.

[Wrecktangle]

I've seen this asserted by user Wrecktangle in various threads and am interested in an elaboration/explanation of this. A relevant link would be sufficient, but I'm looking to have the assertion defended. Thanks in advance.

If you want to "prove" a dist is Gaussian (or not) use a Pearson's chi-squared test.

If you want everything to be Gaussian, then assume it. Rule your world they way you see fit.

BTW, I believe JFK was killed by UFOs. Prove he wasn't.

[hutennis]

It's not like we're looking at 'absolute' natural phenomena here, (like tree rings or some such), which would lead to an ever greater confidence/precision. Which is what gaussian statistics was designed for, and excels at.

Exactly. No one will listen though
People need Bell Curve. They just love it. The same way they love their favorite pass time - looking for patterns.
They don't care that most of what they find is only exists in their minds, that correlation does not equal causation, that there is nothing in a law of large numbers that will bail them out of losing streak and so force and so on. They just need their fix. Fix people were on for the past 250 000 years. At least. Bell curve is there to "justify and confirm". Take away Bell Curve and the dream is gone. Sweat dream of being able to predict future based on what happened in the past.

Industries are built on misunderstanding Gaussian. Without bell curve money managers and financial consultants would have been called charlatans and not the best of the bread and cream of the crop. People are fooled like retarded kids, get hurt and still practically beg for even more abuse.

Amazing.

[statictheory]

Exactly. No one will listen though
People need Bell Curve. They just love it. The same way they love their favorite pass time - looking for patterns.
They don't care that most of what they find is only exists in their minds, that correlation does not equal causation, that there is nothing in a law of large numbers that will bail them out of losing streak and so force and so on. They just need their fix. Fix people were on for the past 250 000 years. At least. Bell curve is there to "justify and confirm". Take away Bell Curve and the dream is gone. Sweat dream of being able to predict future based on what happened in the past.

Industries are built on misunderstanding Gaussian. Without bell curve money managers and financial consultants would have been called charlatans and not the best of the bread and cream of the crop. People are fooled like retarded kids, get hurt and still practically beg for even more abuse.

Amazing.

Fooled by randomness comes to mind. although the author may be right about all of it, his own hedgefund went in the toilet as he waited and waited and bet and bet on the black swan.

[luegofuego]

Classical probability distribution models are perfectly fine when dealing with enviroments lacking black swans, such as hockey totals or whatever.

If there is really no predictive value in these utterly simple but useless statistical models, why don't you just build one yourself (should be simple enough) and then just fade the shit out of it?

There seems to be this trend of people who have read Taleb and think they've found the holy grail of probability theory in a 250 page book and try to score intellectual points left and right by pointing out how utterly useless "classical" statistics and probability are. Guess what, there's a difference between modelling CDOs and modelling 1Q lines. I'm sure the bell curve does a piss poor job of explaining Greek debt default, but sports betting doesn't really take place in "extremistan" as Taleb himself would have put it.

[Kolotoure]

Classical probability distribution models are perfectly fine when dealing with enviroments lacking black swans, such as hockey totals or whatever.

If there is really no predictive value in these utterly simple but useless statistical models, why don't you just build one yourself (should be simple enough) and then just fade the shit out of it?

There seems to be this trend of people who have read Taleb and think they've found the holy grail of probability theory in a 250 page book and try to score intellectual points left and right by pointing out how utterly useless "classical" statistics and probability are. Guess what, there's a difference between modelling CDOs and modelling 1Q lines. I'm sure the bell curve does a piss poor job of explaining Greek debt default, but sports betting doesn't really take place in "extremistan" as Taleb himself would have put it.

Agreed. Not sure how so many people have completely misunderstood that book

[statictheory]

Agreed. Not sure how so many people have completely misunderstood that book

I just stated the facts. His book is interesting, and he did have to shut down his fund because he took an approach to his own theories that didnt work, simply because waiting for the swan can be a long, long, long, long time.
Yeah , sportsbetting is different, but there are constant questions about sample size etc. In Poker, if you play live, there is a problem with sample size vs human life span. there are still winners, but the sd is very large.
I think with sports, in many areas you have the same problem. Your model might make mathematical sense but too
get into the long run might actually take much longer than you think. So, as an example, you winning for 5 years in a row and then losing constantly could simply be simply flucuation, but it doesnt match up to our perceptions of time etc.
so we think the model is flawed etc.. you math guys can be right and still very possibly lose for your lifetime and still not admit that this probability exists or is much larger than your equations tell you.

[luegofuego]

statictheory,

hutennis was the one picking cheap intellectual points on the back of Taleb, not you.

[allin1]

Fooled by randomness comes to mind. although the author may be right about all of it, his own hedgefund went in the toilet as he waited and waited and bet and bet on the black swan.

This I didn't know.

[hutennis]

Classical probability distribution models are perfectly fine when dealing with enviroments lacking black swans, such as hockey totals or whatever.

If there is really no predictive value in these utterly simple but useless statistical models, why don't you just build one yourself (should be simple enough) and then just fade the shit out of it?

Bad logic, is not it?

Statistical models (those we are talking about anyway) are utterly useless not because they are directionaly wrong and thus can be easily exploited by fading the shit out of them. They are useless b/c no matter what you do you are not expected to be anywhere but at random break even and that is not enough in a commissioned environment.

As I said before, I dont know whether sport results normally distributed or not and I dont think it matters much.
What I do know that it is the fact, that number of people winning in sports does not exceed what would be expected to be produced simply by chance in a overall huge population of participants. That speaks volume about random (unpredictable) nature of the beast.
At very least, anyone who claims otherwise and points to his particular results as to "prove" that sports are predictable and therefore beatable would have to deal with and ultimately win against this argument to be considered seriously.

It is not using Gaussian what is wrong, but misusing Gaussian on a backdrop of misunderstanding and underestimating of randomness.
Thats what, to me at least, is the main point of Taleb's books.

If these statements are what you call "scoring intellectual points", so be it.

[donjuan]

What I do know that it is the fact, that number of people winning in sports does not exceed what would be expected to be produced simply by chance in a overall huge population of participants. That speaks volume about random (unpredictable) nature of the beast.

LOLOLOLOL no.

[luegofuego]

Bad logic, is not it?

Statistical models (those we are talking about anyway) are utterly useless not because they are directionaly wrong and thus can be easily exploited by fading the shit out of them. They are useless b/c no matter what you do you are not expected to be anywhere but at random break even and that is not enough in a commissioned environment.

As I said before, I dont know whether sport results normally distributed or not and I dont think it matters much.
What I do know that it is the fact, that number of people winning in sports does not exceed what would be expected to be produced simply by chance in a overall huge population of participants. That speaks volume about random (unpredictable) nature of the beast.
At very least, anyone who claims otherwise and points to his particular results as to "prove" that sports are predictable and therefore beatable would have to deal with and ultimately win against this argument to be considered seriously.

It is not using Gaussian what is wrong, but misusing Gaussian on a backdrop of misunderstanding and underestimating of randomness.
Thats what, to me at least, is the main point of Taleb's books.

If these statements are what you call "scoring intellectual points", so be it.

You're talking out of your ass about a subject you admit to know nothing about.

[hutennis]

LOLOLOLOL no.

You're talking out of your ass about a subject you admit to know nothing about.

Two of the most brilliant, eloquent and thoroughly thought out arguments I have ever seen.
Congrats!

[luegofuego]

Why don't you go read a beginners book on quantum mechanics, then head into a physics forum and wax philosophical about how useless Newtonian physics are at describing projectile motion or whatever, and then get butthurt when no one bothers to write out an eloquent response there aswell?

[statictheory]

This I didn't know.

What's scary is that not long after he closed shop we had the Wall Street Crisis. Timing is everything

[hutennis]

Why don't you go read a beginners book on quantum mechanics, then head into a physics forum and wax philosophical about how useless Newtonian physics are at describing projectile motion or whatever, and then get butthurt when no one bothers to write out an eloquent response there aswell?

Another excellent analogy. Even better than "why don't you fade the shit out of simple models" one.

Let me repeat my argument so you can point out in which way it is so similar to an unsolved mysteries of a quantum world and, and, most importantly, what's wrong with it.

After billions, if not more, of bets made by millions upon millions of bettors its very likely (quoting Edward Ras in this case) than less than 1% are long term winners.

There are two basic explanations for this remarkable success.

1. Winners are exceptionally talented group of people who were able to uncover deeply hidden inefficiencies in sports markets and skillfully exploit them for profit.
2. From standpoint of probability theory, this minuscule rate of success does not require any other explanation but simply a chance.

Basic, universal rule tells me to go with the second one. It also asks anyone who insists that he belongs to the first category to prove it.

I'm sure somebody, somewhere can do it and I'd love to see his/her arguments.
I'm not sure though if "LOLs, you are stupid ****, you have no idea what you are talking about..." and such can count.

[Pot luck]

After billions, if not more, of bets made by millions upon millions of bettors its very likely (quoting Edward Ras in this case) than less than 1% are long term winners.

My guess is that the study is BS.

I, for one, am banned/limited from pretty much all recreational books - why would they do this to me if it were a case of 'getting lucky'?

I also think you misunderstood Taleb. That is just my vote, nothing to prove or disprove.

[hutennis]

My guess is that the study is BS.

Could be, but 1% is pretty much inline with numbers taken from other speculative fields (poker, financial and real estate markets etc.)

I, for one, am banned/limited from pretty much all recreational books - why would they do this to me if it were a case of 'getting lucky'?

I'd say because they don't like/need short term damage regardless of the reason.
Luck, skill, manipulation, combination of all of the above in unknown proportion. Who cares! Splitting hares for a long time for no reason or not big enough reason is not their business model, I guess.

I also think you misunderstood Taleb. That is just my vote, nothing to prove or disprove.

Fair enough.

[Pot luck]

Incidentally,

After a series of concurrent bets, I think you move from 'mediocristan' to 'extremistan' by betting a proportion of your bankroll on each wager .

For example, start with $3, after 3 fair coin flips and a wager of $1 on each of the flips your final bankroll is normally distributed. But if you bet your entire roll on each flip then 7/8 times you will have $0 and 1/8 you will have $24.

[donjuan]

If it's purely based on chance that there are that many successful bettors, then of that 1% only 1% should be successful over the next of that number of wagers in the future. This is clearly not the case but good luck with chapter 2 of your intro to stats book.

[hutennis]

Incidentally,

After a series of concurrent bets, I think you move from 'mediocristan' to 'extremistan' by betting a proportion of your bankroll on each wager .

For example, start with $3, after 3 fair coin flips and a wager of $1 on each of the flips your final bankroll is normally distributed. But if you bet your entire roll on each flip then 7/8 times you will have $0 and 1/8 you will have $24.

Right.

Thats the reason, if my memory serves me right, he puts gambling in "mediocristan" under 2 conditions - known edge and constant wager size

[hutennis]

If it's purely based on chance that there are that many successful bettors, then of that 1% only 1% should be successful over the next of that number of wagers in the future. This is clearly not the case but good luck with chapter 2 of your intro to stats book.

Thank you.

BTW, "clearly not the case" - is that your opinion or is that the fact? Or you just want it to be the fact?

[luegofuego]

You come in and randomly blurt this 1% out as fact with no backing whatsoever, other than some vague reference to some dubious authority, whose insight into the exact percentage of winners/losers in sports betting is even more dubious. You're just saying things completely at random. Over how many bets does this 1% number hold? How many bets did the winners make? How many did the losers make? How big are the winnings? How big are the losses? What vig did they pay? You have no idea about any of this, you're just being ******* random.

And the the most ironic thing of all is that you're trying to use TRADITIONAL STATISTICS to prove that this 1% number is somehow in line with the results of pure chance. How would you even know? Did you build a monte carlo and test it? My guess is NO, because the closest thing you've come to probability theory is some stupid pop-culture book by a brilliant self-promoter who spends more pages name-dropping Poincaré and talking about drinking espresso at the Bonn airport café than he does talking about actual probability.

[luegofuego]

And by the way, even with the exponential stake sizes of sports betting, I wouldn't exactly compare the collapse of entire economies comparable with a MLB losing streak. In sports betting, the maximum loss is predetermined by your stake. In financial derivatives, stuff like the credit crunch happens. Sports betting decidedly takes place in mediocristan and I have no clue how anyone could argue otherwise.

[donjuan]

Thank you.

BTW, "clearly not the case" - is that your opinion or is that the fact? Or you just want it to be the fact?

It's certainly more factual than the random stuff you've made up so far.

[Pot luck]

And by the way, even with the exponential stake sizes of sports betting, I wouldn't exactly compare the collapse of entire economies comparable with a MLB losing streak. In sports betting, the maximum loss is predetermined by your stake. In financial derivatives, stuff like the credit crunch happens. Sports betting decidedly takes place in mediocristan and I have no clue how anyone could argue otherwise.

From its pre 2008 high to the its lowest point during 2009 the Dow Jones shed approximately 50%. My sports betting bankroll has dropped over 40% in the space of 3 weeks before and I have had several swings of around 30%. Even greater movement on the up side.

[durito]

Thats not extreme. Extreme might be losing your whole role in a few days like thremp. But i would really think it would be losing 5x ones roll in a day.