Modeling Book

specialronnie29 · 02-02-11, 12:53 AM

totally useless

what use is time series in sports betting? maybe live markets

guy obviously has no clue about sports betting

sadly the best book on modeling is justin7
and king yao's is a more modern and more useful version of stanford wong's original on general sports betting principle

TU · 02-02-11, 12:54 AM

I've got both of those books. I guess i'll check out some math/stat/modeling books and pick up what I can.

ManBearPig · 02-02-11, 02:07 AM

Conquering Risk is another good book that I've been reading that you may want to check out as well.

Tomahawk · 02-02-11, 04:23 AM

Originally posted by ManBearPig

Conquering Risk is another good book that I've been reading that you may want to check out as well.

Conquering risk is a good start to learn basic modelling by learning it you will be better then the most bettors out their, so you will profit. Obviously there are more good books out there to make you a better and better gambler, but conquering risk is a good start for a beginner modeller.

dmolition · 02-02-11, 09:32 AM

And after Conquering Risk what books for modeling would you suggest TomaHawk?
I am reading a lot of advanced stats but i would like to narrow my reading list a bit.

username474 · 02-02-11, 09:54 AM

http://www.amazon.com/System-Reliability-Theory-Statistical-Applications/dp/047147133X/ref=sr_1_5?s=books&ie=UTF8&qid=1296658133&sr=1-5

http://www.amazon.com/Propensity-Score-Analysis-Applications-Quantitative/dp/1412953561/ref=sr_1_5?s=books&ie=UTF8&qid=1296658283&sr=1-5

There are a lot more books you will want to read, but these will at least give you a good overview.

Wrecktangle · 02-02-11, 10:14 AM

Originally posted by specialronnie29

totally useless

what use is time series in sports betting? maybe live markets

guy obviously has no clue about sports betting

sadly the best book on modeling is justin7
and king yao's is a more modern and more useful version of stanford wong's original on general sports betting principle

^^^^ if it's not a time series what is it?

Sports are absolutely a set of time varying statistics.

Comments like this is why I'm staying away from the tank these days.

specialronnie29 · 02-02-11, 08:36 PM

Originally posted by Wrecktangle

^^^^ if it's not a time series what is it?

Sports are absolutely a set of time varying statistics.

Comments like this is why I'm staying away from the tank these days.

comments like this is why we are glad youre staying away from the tank these days

everything in life is a time series, nothing happens at the exact same time

when statisticians are concerned with time series it is when they are concern about what happens yesterday affecting today

sporting events are almost entirely independent events

Flight · 02-02-11, 10:27 PM

Probably just a misunderstanding of the phrase "time series", as it can change based on context.

In the time series papers I have read, I would agree that it does not apply to sport events, as there is no concept of "state" from event to event. And I have not seen any models that take into account a state variable (I'm not sure what the discrete time step would be for this kind of thing, game-to-game? where k is current game and k-1 is last game? doesn't really work that way). The only place I could see state variables applied are in-game live betting, especially for something like baseball. But then again you would be better off with a state machine based approach like a Markov Model or any FSM technique, as opposed to a time series regression analysis.

Back to OP, just read about regression. Don't feel like you should buy books, just read about linear regressions on the net and you will start learning immediately. Also, search on SBR for econometrics, someone posted a link a long time ago with a PDF for "basic econometrics" that pretty much covers the theory and methods. At that point you just need to get some data and get your hands dirty.

bztips · 02-03-11, 11:20 AM

Originally posted by specialronnie29

comments like this is why we are glad youre staying away from the tank these days

everything in life is a time series, nothing happens at the exact same time

when statisticians are concerned with time series it is when they are concern about what happens yesterday affecting today

sporting events are almost entirely independent events

I think you are way off base here. Yes, 99.9% of regression-based modelers around here probably use just plain old linear (or logit) regressions (or poisson for props) that assume independence. But an entirely different approach is to recognize that, contrary to your statement, sporting events are NOT entirely independent events -- the common thread among events are the teams-players and/or the betting public. Depending on your modeling approach, it may well be enlightening to consider that the Lakers game today has some commonality with the Lakers game yesterday, both because of the Laker players themselves as well as bettors whose betting patterns between the two games may not be independent.

I am NOT an expert in time-series econometrics at all, but just the table of contents from the book in question is indicative that the author has some pretty sophisticated approaches to sports betting, and it's enough to get me interested in reading it. (However, the $95 cost of the book might inhibit me!)

TU · 02-03-11, 12:53 PM

Originally posted by Flight

Probably just a misunderstanding of the phrase "time series", as it can change based on context.

In the time series papers I have read, I would agree that it does not apply to sport events, as there is no concept of "state" from event to event. And I have not seen any models that take into account a state variable (I'm not sure what the discrete time step would be for this kind of thing, game-to-game? where k is current game and k-1 is last game? doesn't really work that way). The only place I could see state variables applied are in-game live betting, especially for something like baseball. But then again you would be better off with a state machine based approach like a Markov Model or any FSM technique, as opposed to a time series regression analysis.

Back to OP, just read about regression. Don't feel like you should buy books, just read about linear regressions on the net and you will start learning immediately. Also, search on SBR for econometrics, someone posted a link a long time ago with a PDF for "basic econometrics" that pretty much covers the theory and methods. At that point you just need to get some data and get your hands dirty.

Speaking of regression, does anyone know how to accomplish regression on a Mac. The new Excel version for a Mac came out and did not include the Analysis Took Pak. I've google searched and can't find anyway to get it done. I guess i'll just go look for a new laptop but thought I would see if I was missing something.

specialronnie29 · 02-03-11, 04:52 PM

Originally posted by bztips

I think you are way off base here. Yes, 99.9% of regression-based modelers around here probably use just plain old linear (or logit) regressions (or poisson for props) that assume independence. But an entirely different approach is to recognize that, contrary to your statement, sporting events are NOT entirely independent events -- the common thread among events are the teams-players and/or the betting public. Depending on your modeling approach, it may well be enlightening to consider that the Lakers game today has some commonality with the Lakers game yesterday, both because of the Laker players themselves as well as bettors whose betting patterns between the two games may not be independent.

I am NOT an expert in time-series econometrics at all, but just the table of contents from the book in question is indicative that the author has some pretty sophisticated approaches to sports betting, and it's enough to get me interested in reading it. (However, the $95 cost of the book might inhibit me!)

If you want to beat the stock market you better know time series analysis. why.

because the stock price today is primarily a function of what it was yesterday

if you graph a sport teams margin of victory over time and try to use a time trend to predict what it is tomorrow, i wish you luck!!

specialronnie29 · 02-03-11, 04:53 PM

Originally posted by TU

Speaking of regression, does anyone know how to accomplish regression on a Mac. The new Excel version for a Mac came out and did not include the Analysis Took Pak. I've google searched and can't find anyway to get it done. I guess i'll just go look for a new laptop but thought I would see if I was missing something.

learn another stats package like sass, stata, etc

bztips · 02-03-11, 04:59 PM

Originally posted by specialronnie29

If you want to beat the stock market you better know time series analysis. why.

because the stock price today is primarily a function of what it was yesterday

if you graph a sport teams margin of victory over time and try to use a time trend to predict what it is tomorrow, i wish you luck!!

Uh, a lot of betting and betting models have nothing to do with margin of victory... moneylines for example.

CHUBNUT · 02-03-11, 05:25 PM

usual stuff, fella asks a simple question and within minutes nerds morph the thread to quantum physics

as a beginner, a read on gambling psychology and understanding how the brain reacts to gambling can go a long way to saving serious losses which leads to disciplined betting which leads to acceptable losses which gives you a chance to make the unlikely step to small profits and a springboard to the very few who actually make a decent profit, like my friend durito.

specialronnie29 · 02-03-11, 05:45 PM

Originally posted by bztips

Uh, a lot of betting and betting models have nothing to do with margin of victory... moneylines for example.

Using last game's moneyline to predict this games true moneyline isntt going to get you anywhere.

the implicit defn of time series here is that the order of a sequence of events (margin of victory, total pts scored, etc) matters

if 2 days ago google's stock was $600 and yesterday it was $500 then a very rough prediction would be that today it would be $400

if instead 2 days ago it was $500 and yesterday it was 600 then the prediction for today would be 700

in time series forecasting the order matters. the implicit assumption when you say youre using time series analysis to predict something is that youre using yesterday's value as at least one of your predictors of today's value

go look at historical data and regress, for example, the laker's margin of victory at time t on the laker's margin of victory at time t-1. the coefficient will not be significant.

just because a team's margin of victory over time is a time series (any sequence of data measured over time is a time series) doesnt mean time series forecasting/statistical methods will be useful

imagine recording the results of a 1000 coinflips of an unfair coin and trying to use time series to predict the result of the next flip. no instead the mean of the past 1000 coinflips is going to be a much better predictor and that is the essence of how real bettors model sports

bztips · 02-03-11, 08:00 PM

SR, your simplistic explanation of how time series econometrics might or might not be useful for sports betting might be more convincing if you had actually read the book in question. I haven't, and I assume you haven't either. Unlike you, I am not willing to ASSUME that I know the author's approach is invalid just because it's not how more traditional (conventional) sports modeling is done.

You can have the last word, I am done here.

bztips · 02-03-11, 08:05 PM

Originally posted by CHUBNUT

usual stuff, fella asks a simple question and within minutes nerds morph the thread to quantum physics

as a beginner, a read on gambling psychology and understanding how the brain reacts to gambling can go a long way to saving serious losses which leads to disciplined betting which leads to acceptable losses which gives you a chance to make the unlikely step to small profits and a springboard to the very few who actually make a decent profit, like my friend durito.

As many others have noted before, you have no clue what you're talking about. The OP didn't just "ask a simple question"; he asked whether a particular book, which appears to employ fairly sophisticated statistical techniques, might be useful for learning how to model. If modeling is of no interest to you, or it's beyond you, that's fine; just go over to the PT forum and rummage around there.

Flight · 02-04-11, 06:00 PM

The book OP posted actually does look interesting, but probably not for a first time modeler. Start small and simple (props)

TU · 02-04-11, 06:52 PM

Originally posted by CHUBNUT

usual stuff, fella asks a simple question and within minutes nerds morph the thread to quantum physics

as a beginner, a read on gambling psychology and understanding how the brain reacts to gambling can go a long way to saving serious losses which leads to disciplined betting which leads to acceptable losses which gives you a chance to make the unlikely step to small profits and a springboard to the very few who actually make a decent profit, like my friend durito.

Those are good points but i've been gambling for about ten years. The last 5 or 6 have been pretty good to me. I just want to graduate to a different level. I've got a Finance degree and taken some complex stat courses so i'm comfortable with the material, but I want to learn how to relate it to sports gambling. In order to really get to the next level I need to learn how to model. Thanks for all the input by everyone in this thread.

Dukebluejms · 02-04-11, 08:56 PM

The biggest problem is finding the right variables. No matter how many statistics on teams you have there isn't enough. Too many intangibles in sports to model them completely.

dbDan · 02-04-11, 11:14 PM

Originally posted by Dukebluejms

The biggest problem is finding the right variables. No matter how many statistics on teams you have there isn't enough. Too many intangibles in sports to model them completely.

You really can't be serious. If you believe this, why post in the Think Tank?