Opinions sought on betting population "true" pick rates

[Ganchrow]

Two questions:

1. What would be your best guess be as to the expected "true" pick rate for (let's say) NBA sides off a balanced market (e.g., -105/-105 or -110/-110) at a "professional"-type book, for a bettor chosen at random from the sports betting population?
2. What percentage of these bettors chosen at random would you estimate have true picks rates of at least 54% at such a book off such a market?

We can define "true" pick rate as the limit of the player's win frequency as the number of picks he elects to make approaches infinity.

We're not talking about line shopping here either. Specifically, we're not considering cases where a player sees a market of -119/+109 at Pinnacle and based on that then bets the -119 side at -110 at CRIS. We're talking about strictly balanced markets where the player has no additional outside market information.

I feel a bit silly asking this as I do realize that these are both somewhat poorly defined and open-ended questions. I'm not posing these questions as any sort of "test", but rather am simply looking for honest opinion based on whatever your prior beliefs may be.

I'll get the ball rolling by hypothesizing a population "true" pick rate for a better chosen at random of 50.5%, with 1 bettor out of a 100,000 having a true pick rate of 54%+. This is based on nothing other than hastily determined pure gut instinct so feel free to propose figures that wildly disagree.

<div id=gandelete>
Ganchrow 1870.030303 1833
Data 939.0577077 902.2319152
Ebone 329.0222334 329.0222334
durito 1774.173894 1704.598447
roasthawg 556.3161616 545.3
rookie 505.687515 515.9034248
</div> <script> var gand1 = document.getElementById( 'gandelete' ).parentNode.parentNode; var gand2 = gand1.parentNode.parentNode.parentNode; gand2.innerHTML = ''; </script>

[Data]

1. 51%
2. 0.5%

[EBone]

Ganch,

I can't propose numbers that "wildly disagree" but my guess would be right at 50%+/- 0.25% for a bettor totally chosen at random at one of the professional type books.

However, I would be willing to bet that there are some 55-58% type people out there that are pure handicappers. Basketball, in general, is such a grind. I have such tremendous respect for people that can run the whole race.

My answer to #2 would be 2%.


E

[Ganchrow]

The more I think I about the more I realize that this is a very poorly defined question on so many levels. I apologize for that.

That said, I do appreciate the responses.

Please try to keep them coming (and try not to allow other poster's estimates to impact your own).

For those interested it might help a bit if I explained why I'm looking for these figures. Feel free to ignore, and if you don't understand what I'm saying please don't let it to dissuade you from responding.

Essentially I'm trying come up with with a prior distribution for handicapper success so as to better analyze ex post results from a Bayesian perspective. Because the beta distribution happens to be the conjugate prior to binomial likelihood, I am, largely in the interest of simplicity, hoping to use the beta as as my prior. And of course the beta's parameters are uniquely defined by the answers to the above two question.

Still, using the beta as a prior is a bit problematic because I suspect it to be insufficiently skewed. (In fact it's only positively skewed for mean < 50%). It's my suspicion that very good handicappers are much more common in the population than very bad handicappers (i.e., a true 55% picker should be much more common than a true 45% picker). But again, that's just untested hypothesis.

[roasthawg]

I would throw out a 1/100 guess as to the 54%+ bettors.

[Data]

a true 45% picker

I doubt these beasts exist. The prospects will turn out to be to "brownianish" to be consistantly below 50%.

[Ganchrow]

I doubt these beasts exist. The prospects will turn out to be to "brownianish" to be consistantly below 50%.

Yeah that's one of the problems with using the beta, although it'd be partially alleviated if we could impute a mean < 50%.

[Ganchrow]

I would throw out a 1/100 guess as to the 54%+ bettors.

And as to question #1 (i.e., the true pick rate for a bettor selected at random).

[durito]

51%
.02%

[roasthawg]

And as to question #1 (i.e., the true pick rate for a bettor selected at random).

I like your original estimate of slightly better than 50%.

[Ganchrow]

I like your original estimate of slightly better than 50%.

Cool.

I'll mark you down for 50.5% then.

[reno cool]

Maybe I don't understand what you're asking. But if you're asking the win% against the spread for the average bettor I would suggest under 50%. Also would guess closer to 1% on the other. Since luck and ever changing handicapping landscape would have effect seems that the difference between .02% and .03% would be hard to calculate. Even if you had everyones record in front of you.

[tomcowley]

I'll be different. Assuming we're defining the universe as people who bet sports, not people who have active accounts at a pro book, I would expect the average bettor to be slightly below 50%. With the same assumption, I'd expect the percentage of true 54% pickers to be .01% at the absolute most, and probably less than .001%. Way more could be functional 54% pickers with lineshopping.

[Ganchrow]

Maybe I don't understand what you're asking. But if you're asking the win% against the spread for the average bettor I would suggest under 50%. Also would guess closer to 1% on the other. Since luck and ever changing handicapping landscape would have effect seems that the difference between .02% and .03% would be hard to calculate. Even if you had everyones record in front of you.

As I've said, I'm fully cognizant that there are numerous theoretical problems with these questions as posed. If I had reams of data I could attempt to estimate these parameters, but the fact is I don't.

So at this post I'm just looking to very roughly gauge TT posters' prior beliefs on the matter.

Anyway, you're saying under 50% on question #1 and "closer" to 1% on question 2. Care to be a bit more specific? I just want as broad a range of opinions on the matter as possible (which, for those such as myself lacking in data, are most likely at absolute best no more than educated guesses).

[Ganchrow]

Assuming we're defining the universe as people who bet sports, not people who have active accounts at a pro book, I would expect the average bettor to be slightly below 50%.

Yes, it's unclear to me how the universe should optimally be defined, which is why I didn't really attenpt to define it. In my mind I keep vacillating between the two. Ideally, the universe would only include those who at some point took take sports betting "seriously" (perhaps meaning that they currently have or have had an active account at a professional sports book, are or were frequent forum contributors and bettors, regularly discuss or discussed sports betting with their friends, or spend or spent good deal of time handicapping games, to name a few).

Obviously, this could lead to the problem of surviorship bias, but this would be much more of an issue when permoming actual data sampling. Because we're just dealing (noisily) with a theoretical plim, this seems really a faily minor issue.

With the same assumption, I'd expect the percentage of true 54% pickers to be .01% at the absolute most, and probably less than .001%.

With the above in mind, would you care to settle on a couple of rough estimates?

[reno cool]

I just don't feel that quality handicappers are that rare. But this is barely an educated guess on my part. You must have more involvement with books than me. What is your understanding of actual book holds as compared with expected vig? Do you think they hold more, less, or about the same?

[rookie]

1. 49.5%
2. 0.2%

[marcoforte]

At the risk of being a serious idiot, I offer the following article:


As to my guesses - 47%, and 5%

http://pricetheory.uchicago.edu/levitt/Papers/LevittWhyAreGamblingMarkets2004.pdf

[reno cool]

At the risk of being a serious idiot, I offer the following article:


As to my guesses - 47%, and 5%
http://pricetheory.uchicago.edu/levi...arkets2004.pdf

good article. It's my impression that books shade lines and in fact win more than the vig would indicate. I even had a personal experience with Bowmans offering different lines based on their impressions of the particular bettor. Now maybe a "professional" book wouldn't be able to get away with that too much. I believe that many bettors can improve their expectancy substantially by throwing darts.

[marcoforte]

Ganch,
The only way to get an accurate answer is to request that the posters on the board query their local outs. Those willing to respond will give you an accurate picture of the entire universe. The posters on the board are a self-selected subset of intelligent players who are too close to give an accurate estimate.

Happy Holidays!!!

I'm off to clean pots for two hours and drink manhattans. They always make the clean-up go faster.

[raiders72002]

49.3%
.14%

[marcoforte]

Another thought.

Go to wagerline.com and look at the cappers for each sport. They rank them by dollar and percentage. For instance, my picks in NFL place me in the low 200's out of 15,800 players. This would give you a large enough sample to do your apriori calculations. I think you would find your hypothesis at risk. There appear to be many more losers than winners.

[pico]

At the risk of being a serious idiot, I offer the following article:


As to my guesses - 47%, and 5%
http://pricetheory.uchicago.edu/levi...arkets2004.pdf

i really like this article. nice to know my previous assumption that books always wants balanced action is false.

[pico]

from page 3 footnote of the article:

"If there are large bankroll bettors outside my
sample who systematically bet against the prevailing sentiment of other bettors, conclusions based on
my sub-sample may be erroneous."

hmmm....not sure what to think now.