Uploaded by Google on 23.07.2007

Transcript:

SADIK: We're very pleased to have Dr. Michael Orkin here to

give us a talk today.

Dr. Michael Orkin has a very distinguished career.

He's currently Managing Scientist at the Exponent, a

publicly traded center scientific consulting company

here in Menlo Park.

He's a well-known personality.

He's been nationally quoted, CNN, NBC's Dateline and ABC

World News.

He's written several books, including What are the Odds?:

A Chance In Everyday Life.

He's consulted for clients in both government and private

sector, including the FBI, and Las Vegas odds makers.

His software has also been featured in the January 2002

issue of Wired magazine.

So here he is today.

Let's give a Google welcome to Mike.

Could I please ask that because this talk is being

video recorded, any confidential Google questions

should be left till after the cameras have

been switched off.

Thank you.

[APPLAUSE]

DR. MICHAEL ORKIN: Thanks, Sadik.

So today-- well, the title of my talk-- and I'm happy to be

here-- is Chance, Data Mining, and Sports Betting, and we'll

see what those things have in common.

I am currently located at Exponent.

If you haven't heard of it, it's sometimes called Exponent

Failure Analysis.

It's located right on 101 North in Menlo Park.

It's a publicly traded company that has about 800 employees,

in 18 offices and three international offices.

Consultants at Exponent, a majority of whom are

engineers, have expertise in science, math, and

engineering.

And Exponent has been consulted on various problems

and disasters, including some of the major ones you read

about in the paper, such as the grounding of the Exxon

Valdez, the walkway collapse of the Kansas City Hyatt, the

bombing of the Alfred P. Murrah Federal building in

Oklahoma City.

By the way, in this talk, just stop me at any time if you

have any questions, and we'll have more of a

conversational tone.

So I'm in what's called the data and risk analysis group.

I'm trained as a statistician, and do stuff in probability

theory, decision theory, game theory.

And our group develops strategies for decision-making

under uncertainty, and in a variety of contexts.

One of the things our group specializes in is determining

whether a particular activity or product poses an

unreasonable risk, and if so, what to do about it.

Many of our projects involve litigation involving such

things as consumer product issues and recalls.

Automotive lawsuits, and doing risk analyses and failure

analyses to try to develop strategies for companies who

have products.

I can't really to go into too much detail about things that

are in litigation right now.

But that's not what I want to talk about anyway.

I want to talk about one of my specialties, which is chance

and gambling games.

I'll actually be talking about actually a case that is an

Exponent case, along with some software and things that are

not Exponent things, just things I've done.

So first, this a little primer on gambling.

So roulette--

how many of you have been to a casino and have seen a

roulette wheel.

Most of you.

So roulette's a pretty simple game to understand.

There are 38 sections labeled from 1 to 36 along with a zero

and a double zero.

And half of the sections from 1 to 36 are red, half are

black, zero and double zero are green.

You can bet on a color, you can bet on a combination of

numbers, you can make varieties of bets at the

roulette table.

One of the bets is number 17.

In other words, you put some chips on number 17.

The wheel is spun, the ball is dropped in, 17--

the ball lands in the slot marked 17.

You win.

Otherwise you lose.

And since there are 38 sections on the wheel, your

chance of winning is 1 in 38.

If you bet on red, since there are 18 red sections out of 38,

your chance of winning is 18 out of 38.

So you're much more likely to win a red bet than you are a

bet on number 17.

So to get people to make this bet, what does the

casino have to do?

They have to offer higher payoff odds.

So the payoff odds for this bet, a bet on a particular

number, are 35:1.

That means if you bet $1 and number 17 comes up, you'll

make a $35 profit.

You'll get your dollar chip back plus another 35.

Well, there's a mathematical theorem

called the law of averages.

It's just called the law of large numbers by

mathematicians.

It says that if you repeat an experiment independently over

and over again, the fraction of times that something comes

up will be the same as the probability of it

happening in one trial.

So the probability of it happening in one trial on this

bet is 1 in 38, so you don't know what's going to happen in

one bet or two bets or a few bets, but in repeated bets

you'll win an average of 1 in 38, and lose an average of 37

for a net loss of $2.

So that's $2 and $38 bet, which becomes 5.3% profit for

the house, the house being the casino.

So that's called the house edge in roulette.

And it's very straightforward, cut and dry math.

This is what happens if you repeatedly

make a bet in roulette.

Like I said, you can get lucky, you can get unlucky and

lose sooner than at this rate, or you can win and quit.

But in repeated play, you will eventually lose and go broke

at roughly the rate of 5.3 cents per dollar bet.

Sooner if you have a losing streak.

So that's sort of the facts on betting on number 17 in

roulette, and in fact, in making just about any roulette

bet, or a combination of roulette bets, or other

betting strategy applied to roulette.

So you can show mathematically that this is

what happens, period.

For instance, people use some sort of sophisticated

strategies, like double your money if you lose, and keep

going until you eventually win.

You've heard that one, the doubld-up strategy.

So you start out by betting $1, and if you lose you bet

$2, and then if you lose again you bet $4, and eventually

you'll win, thereby covering your losses.

So that's an interesting strategy that people use in

various types of investment situations, namely, you start

losing, bet more.

You can show that that's not a good thing to do.

In fact, it's really not a good thing to

do in a losing game.

One of the reasons it's not a good thing to do is the house

has a betting limit, and if you have an unlucky sequence

of losses, you can't double your bet because they won't

let you bet that much.

Another problem is if you have a betting limit, if you have

an unlucky sequence of losses, you might run out of money

before you can double your bet and come out ahead.

Anyway, so that's sort of the quick tour roulette.

Now, if you talk about roulette betters, you could

say a roulette better who makes a profit after 10 bets

would be lucky.

But a long-run profit would be very weird.

The math says that's really unlikely to happen.

So a possible explanation where somebody makes a

long-run profit in roulette, they're cheating.

So they have some scheme worked out, and they're not

really playing according to the regular roulette odds.

On the other hand, you can also detect if a roulette

better is losing more than he or she should.

In fact, if a roulette better is losing after 10 bets, so

what, they're unlucky.

But if they lose in the long-run and repeated bets

faster than they should, faster than chance predicts,

5.3 cents per bet, that would also be weird.

That would be equally weird, as weird as the winner.

And then possible explanation for that, the

better is being cheated.

So there are two different kinds of weird.

Weird good and weird bad.

But anyway, where do you draw the line between luck and

something else is going on?

So statisticians, that's sort of the basic question of

statistical inference.

Traditionally, 5% is a cut-off point, which is called

statistically significant.

If what you observe has less than a 5% chance of happening

under ordinary circumstances, you conclude that something

else is going on.

So in roulette, it's simple, you can make probability

calculations and answer this statistical question.

Is what you've seen weird, or is it just within

the bounds of chance?

What about betting on football games?

What's the chance that the 49ers will win next week?

Well, 49ers started out looking pretty good in their

first couple games, but what's going to happen next week?

Well, football is not like roulette.

The same experiment is not repeated under identical

independent conditions, and so the law of averages and

various probability models don't necessarily apply.

So you can't make these cut and dried mathematical

statements about the persistent football better

goes broke, which you can say about roulette, or craps, or

Keno, or any of the casino games of pure chance.

Well, football betting is an interesting industry.

It's about a $10 billion industry right now, almost

entirely illegal.

AUDIENCE: How are illegal industries

such as this measured?

How do you-- where does the $10 billion come from?

DR. MICHAEL ORKIN: Surveys.

There are trade organizations that do surveys and find out--

in fact, I'll give you a link in a minute.

It's a good question.

But they're illegal--

Well, here's one answer.

Football betting is illegal in this

country, except in Nevada.

Football betting is legal just about

everywhere else in the world.

So internet gambling is very popular.

And internet gambling is legal, except here.

So that's sort of an interesting legal issue, as

you probably know, because if you bet on a football game

here, you're violating some Federal law, but if you're

playing in an online casino, that's perfectly legal, that's

located in Costa Rica, or Jamaica, or Canada.

So legal is sort of this strange term in this world.

But yeah, how do you come up with a number like?

So there are trade organizations that do surveys

and find out things like that.

So a typical football bet, not all, but a typical football

bet is done with a point spread.

So a point spread is a handicap given to

the inferior team.

And who decides that it's an inferior team?

Well, in some sense the bettors decide.

So here's how it works.

Let me go to the third bullet point.

First week of the season, the 49ers played Arizona, and the

Niners were nine point underdogs.

So that means if you bet on the Niners in that first week,

you get nine free points tacked on to their score.

So it turns out San Francisco lost that first game, 37 to

27, but if you add nine points, the Niners win.

So people who bet on the 49ers in that first

game won their bets.

And the people who bet on Arizona, even though Arizona

won the game, lost.

So that's fine.

Now, the casino has to make a profit, so

they pay 10:11 odds.

So that means if you make an $11 bet with a point spread,

and you win, you make a $10 profit.

So now how does that ensure that the casino is going to

make some money?

Well, what the odds makers want to do is get the same

amount of money to be bet on each team in a game.

And if that happens, because of the 10 to the 11 odds, the

losers pay off the winners and there's a little bit money

left over for the casino.

So if there are two bettors on a game and each bet's $11,

that's $22 bet, and one wins, the other loses.

So the $10 pay is paid off out the $11, that's a $1 left for

the casino.

1 out of 22 is 4.5%.

So if the casino gets an even split of the money on a game,

then they make a 4.5% profit on that game.

It doesn't matter which team wins.

So that's what they want to do.

Now, if one team has an 80% chance of winning and the

other team has a 20% chance of winning, whatever that means,

but everybody likes the 20% team, then the point spread is

going to be shifted.

So to get more money bet on the team, that really should

be the one that everybody's betting on.

So the betting public is foolish, the point spread may

give a statistical advantage to the smart bettor who knows

that the money's coming in on the wrong team.

So that's different from roulette.

On the other hand, if the betting public is--

it's different from roulette in the sense that it's sort of

like the stock market.

The odds are determined by the public, and if the public

doesn't know what they're doing, the bettor who does

know what he or she's doing can take

advantage of that situation.

On the other hand, if the betting public is smart,

equalizing the betting action will also equalize each team's

chance of winning, and the game will be

equivalent to a coin toss.

So you don't want to get 10:11 odds for

betting on a coin toss.

So in order to win in the long run, you have to find

situations in which the betting public isn't smart.

In other words, in which your chance of winning is

greater than 50%.

So it turns out that if you just look at some basic

situations, like how often does the home team win, well

if you look at the years 1993 to the present, the home team

actually wins the game almost 60% of the time.

But when you factor in the point spread, it's almost a

50/50 split.

So if you just think I'm just going to bet on the home team

because I know they win 60% of the time, you're going to lose

because the point spread completely equalizes that.

So what do people do?

Well, in one form or another, a lot of sports bettors do

data mining.

Of course, some of them don't actually use software, they

just pour through data by hand.

But nowadays there are a lot of sophisticated sports

bettors who actually use software.

So in general, data mining software is used to find

predictive patterns in large data sets.

So people use data mining for all kinds of things--

investment strategies, web surfing patterns just to try

to find patterns.

I wrote some software to search football data that sort

of points out some of the issues of data mining.

I was actually motivated by badgering from some attorney

friends who were avid sports bettors and didn't like doing

all the stuff by hand.

They thought it would be great to automate the process of

trying to find good betting situations.

So I did that a couple years ago.

It basically uses sequel to query a football database.

It has a user-friendly interface

necessary for lawyers.

There's a procedure in the software that automatically

generates query, so it does a random queries.

And it finds good results and saves them in a separate file,

and just forgets about bad results where good means

better than coin tossing.

So it uncovers trends or possibly--

it uncovers trends that the user wouldn't think of.

There's no best strategy when you look for

patterns like this.

But instead you try to collect a bunch of good strategies,

such as it turns out that by data mining from 2000 to 2006,

the Baltimore Ravens were 17 and 3 versus the point spread.

That's an 85% success record when they lost their previous

game and their opponents play their

previous game on the road.

So the question, of course, is for any thoughtful data miner,

is this really meaningful or is it just a random

coincidence that you found by data mining?

So first of all, you have to define what a

good strategy is.

Win percentage doesn't work because 1 and zero is 100%,

and that's, of course, meaningless.

17 and 3 is only 85%.

So as I mentioned a minute ago, one way to define good,

or weird, or out of the ordinary is to compare the

strategy to an appropriate probability distribution.

In this case, coin tossing.

So you see if the strategy does a lot

better than coin tossing.

It turns out that that 7 and 3 record of the Ravens, if

that's really coin tossing.

In other words, if it's just a random fluke.

The chance of getting something like that

is about 1 in 1,000.

So just use the binomial distribution.

For those of you who know basic probability

distributions, the chance of getting 17 and 3 are better,

out of 20 is about 1 in 1,000.

So a statistician would reject the null hypothesis, which

means you would say that's not just chance.

So in other words, that would be considered, in some

context, a good strategy.

So the question is should you bet on

Baltimore in this situation?

Well, interestingly enough, technology changes the meaning

of unlikely and context in situations like this.

There's something called the law of very large numbers,

which says given enough opportunity, weird things

happen just due to chance.

So when you're mining through data doing lots of queries,

saving the good ones--

by good, I mean good compared to coin tossing--

discarding the bad ones, the meaning of 1 in 1,000 sort of

changes a little bit.

So in fact, using a modest PC, this particular data mining

software does about 1,000 queries a minute when it's

doing the automated query thing.

It's generating queries at random, analyzing to see if

they're better than coin tossing, saving them in a file

if they are.

Anyway, that's lots of opportunity for weird things

to happen just due to chance.

So even if you generate this data by tossing coins, you'll

find 17 in 3 result about once a minute.

And you can turn this on and let it run all night, so you

can find all kinds of great stuff.

So statisticians consider something statistically

significant if the chance of occurring is less than 1 in 20

under a suitable hypothesis.

In the data mining environment, even 1 in 1,000

doesn't necessarily mean rule out chance.

So more generally, these kinds of techniques in which you're

doing repeated queries and analyzing them sort of comes

under the general heading of multiple comparisons.

You're testing lots of statistical

hypotheses at once.

What happens is you might find some useful things.

You might find some strategies in which you'll win money, but

a lot of the stuff you're going to find

is just random stuff.

So that's one of the issues with--

it's one of the big issues when you're doing things like

data mining of this type that you have to deal with.

And of course, there are some standard procedures for

dealing with that.

The main remedy is you apply your strategy that you've

developed by looking at some data to a fresh set of data

and see how it works there.

So what statisticians would call it, or decision

theorists, you'd have a training sample and then a new

data set that you'd apply your thing to.

So if it's just due to chance from mining the training set,

it won't work on the new data.

Or you can do a simulation, do a lot of random queries and

compare your results to a probability model.

So for example, if you're doing a lot of random queries

on this football data, and you draw a graph of the results,

you should see a record of 17 and 3 occurring with a

frequency of about 1 in 1,000 if it's just due to chance.

If you get some weird pattern in your histogram or whatever

graph you're looking at, then maybe there's something other

than coin tossing going on.

Any questions so far?

AUDIENCE: Well, it seems like you could also just bet on all

of these clues.

And the ones which are still-- which are useless are going to

be 50/50 for you.

It would be a small amount, as we've seen today.

And the ones which are actually useful

will win the money.

DR. MICHAEL ORKIN: That's exactly what lots of people

do, except with the part about winning money.

It turns out that the 10:11 are a little bit deceptive,

and you have to win more than 52.4% of the time in order to

make a profit because of that.

But that is a standard strategy.

People they, either by hand or using software, find lots of

these types of situations that look like they give you an

edge, and they'll just bet on them.

And they'll even write them, like this one's really good.

So I'll bet more on the one that's really good than the

one that's just sort of good, and I think some people are

modestly successful doing things like that.

Any other questions?

So let's see, an earlier version of the software was

featured in Wired magazine in 2002 in the

street credit section.

But let's actually look at the software for a sec, if I can

get it up Here So this is called the optimizer.

Let's say you want to see how the home team does from--

so it has this sort of user-friendly interface--

2001 to 2006 in the month of September.

Can you see that?

I guess you can sort of see it.

You get the statistics, and you can see that the home team

from 2001 to the present, not including last week's games.

So SU means straight up, they're 144 and 110.

They won close to 57% of their games.

But versus the point spread they're almost 50/50 at home--

this is just the home team.

And this Z-value is sort of that's the measure of how far

away from coin tossing it is.

A Z-value, a large Z-value means very

unlikely under coin tossing.

There's another kind of bet in football called the over/under

where you bet on whether the total score will be over or

under a particular number.

So this software keeps track of that, too.

So if I wanted to look at a particular team,

let's look at Atlanta.

How did they do in September?

So Atlanta's 5 and 3 on the road, 4 and 4 at home versus a

spread in September and so on.

Suppose I wanted to see which team had the best September

record, looking at all records of this type.

I just click optimize and I see that Jacksonville is

actually 12 and 4 versus the spread in September, which

gives them a--

so that's another one of these good records.

So you can also look at what happened in the last

game, won or loss.

So let's look at games where the home team--

just playing around here--

won their last game and played at home in their last game,

how did they do?

So pretty much 50/50 against the spread.

So this is basically just a statement of

the query up here.

Then you can also look at a list of the

games for that query.

So this is just a query output.

Then this part here accesses the file that stores good

situations.

So now the automated part.

So I'm going to clear this.

Let's say you want to look at just--

I'm going to restrict the year range--

2000 to 2006.

Let's see if we can find any good

strategies for the Niners.

So I'm just going to click random search.

So it's looking for good strategies for the 49ers, and

that search count tells how many queries it's looked at

and analyzed.

It's only saving the ones with Z-values greater than 2.5, and

so far it hasn't found anything.

Oops, there it found one.

I'll let it run for another couple seconds.

Two.

So it's found two strategies.

Let's just look at the best one.

This is a strategy.

The 49ers from 2000 to the present, when their opponent's

last game was on the road, and the opponent's last game was

under-- that means not a lot of points were scored--

then the 49ers on the road are 2 and 13 versus the spread

with two ties.

That's a Z-value of 2.84.

What that suggests is when this situation comes up, you

bet against the Niners because they're 2 and 13.

Then here's a list of all those 49er games.

Then, as I said, you can look at the save

situations in there.

Another thing you can do with the software is you can get

updates from the web.

So if you click on that, it'll actually automatically

download the scores and point spreads for this week.

You can manage the situations--

that means the good strategies you've already found.

You can share it with other people, you can import last

year's, and so on.

You can also print out any of this stuff.

So I don't have many saved situations, but if I wanted to

look at all the games in September, I can go

through each one.

I don't think anything's on here, but let's see.

I'll just look real quickly.

So if you have a lot of-- some people use this software, have

thousands of situations.

It'll find the ones that apply to a particular game.

So anyway, the part about this software that I think is sort

of interesting is the random search thing.

And that is just an automated query generator.

So any questions?

Yeah.

AUDIENCE: Is it interesting to do things like to know that

you got rid of those instead of just randomly generating--

DR. MICHAEL ORKIN: Yeah, so that's an interesting

question, and we were messing with that for a while.

So a genetic algorithm would be taking a good strategy,

looking at parameters of a good

strategy, like won or loss.

Let's say that that Baltimore strategy they won, when they

lost the last week and their opponent was whatever.

So you have a bunch of variables in some vector.

A genetic algorithm would do something like perturb the

different values in the components of the vector and

see if you can get something better.

So would sort of try to genetically mutate the vector,

which defines your strategy, and get a

better one out of it.

So yeah, we sort of messed around with that.

It's not part of the software right now.

Yeah.

AUDIENCE: Can you potentially use the software to identify

features which are aspects of the--

bookmakers themselves are using to come up with the

point spread?

And more importantly, [INAUDIBLE]?

DR. MICHAEL ORKIN: Well actually, that's a good

question. the answer is the way the bookmakers come up

with the point spread is they look at the amount of money

that's being bet on each team.

And if it's not equally balanced, they'll make the

point spread more favorable for the team that's

not being bet on.

They're not really looking at much of anything, except the

betting money, and that's all automated.

Now, they have to start with something.

So at the beginning of the week, like right now, in fact

on my way in, in the lobby, I stole the sports section of

today's Mercury News, and even though football betting is

illegal, sitting right here in the lobby is the Mercury News

latest line, Pro Football.

So I can see that this coming week Philadelphia's a six

point favorite over the 49ers.

Buffalo is a 5 and a half point favorite over the

Jets, and so on.

Then they also have the lines for college football.

If it's bad to bet on pro football, it's supposed to be

a real sin to bet on college football because you're sort

of corrupting the integrity of the game.

But here it is in the San Jose Mercury News.

I don't know.

AUDIENCE: That's the Nevada issue.

DR. MICHAEL ORKIN: What?

AUDIENCE: That's the Nevada issue.

DR. MICHAEL ORKIN: That's the Nevada issue, right.

This is the online issue.

AUDIENCE: So doesn't work if the point changes--

DR. MICHAEL ORKIN: Correct.

So that's a good question.

So he had asked, so your question was how do the odd

makers decide what are the important factors?

Well, they start out with something, because here it is

Monday and they've already printed a line, which is

roughly resembling the various lines in Las Vegas.

So there hasn't been that much betting action yet.

So they had to start with something.

So what they start with is they look at software, and

they look at the records of teams, and they have formulas

that give, what we called, power ratings, which are just

ratings of how good teams are.

Then they just sort of sit around and actually just sort

of talk about how they think the public is going to bet.

Then they'll come out with what's called an opening line,

and they'll modify the opening line throughout the week.

And it doesn't change by that much, actually.

They're pretty good at coming out with the opening line.

So the things that they are-- and they know a lot about

football, and whatever sport they're handicapping.

In fact, the companies in Las Vegas, handicapped sports has

experts in each sport who really know a

lot about the sport.

They also have software.

But the software is software that just gives trends and

patterns, not exactly data mining software.

But they use that too.

But they rely heavily on how they think the

public is going to bet.

So right now the 49ers--

sorry, I didn't answer your question.

Right now, if you bet on Philadelphia, you have to give

six points since Philadelphia's playing 49ers

next week, and they're a six point favorite.

So let's say that during the week lots of people bet on

Philadelphia until they move the point spread up to seven.

When you made your bet it was six, so you get six.

That was your question, right?

Yeah.

So you get the point spread that was available when you

made your bet.

Yeah.

AUDIENCE: [INAUDIBLE]?

DR. MICHAEL ORKIN: Well, if you're doing this data mining

method or something similar to it, that's correct.

AUDIENCE: [INAUDIBLE].

DR. MICHAEL ORKIN: That assumption is only moderately

reasonable for a few important reasons.

One is that teams change, coaches

change, stadiums change.

So it's not like the roulette environment.

We get these new situations.

Like this year, the Raiders have a new coach, a new

offensive coordinator--

offensive is a good word for him--

and they have a new quarterback.

So it's like--

AUDIENCE: [INAUDIBLE].

DR. MICHAEL ORKIN: So what people who bet on sports do is

they get these patterns, they get these things like

Baltimore Ravens are 17 and 3, let's bet on this week.

And then they look at other things that might be more

connected with the real world, namely--

and they do actually what a statistician would do.

They would say OK, I have this weird situation now.

What's a plausible explanation?

And if there's no plausible explanation, a smart person

would say this is just random nonsense.

AUDIENCE: So why not weight things differently.

So the further they are [INAUDIBLE]?

DR. MICHAEL ORKIN: People have used other techniques.

What people tend to do, use software like this or is to

restrict the year range.

I mean I know some people who say if you go back more than

two years you're an idiot, because everything that's

relevant has happened within the last two years.

That's almost like mark-off change or something.

But then I know this guy who's a sports bettor who says you

have to go back at least 15 years, because only then will

you get enough data to establish a pattern.

And that there are certain underlying physical

characteristics and dynamics of football that you can

uncover by looking at lots of years.

So it's a mixed opinion on that.

Yeah.

AUDIENCE: [INAUDIBLE PHRASE]

what position you play.

At least some combinations of scores might be because of the

way that the [INAUDIBLE PHRASE].

DR. MICHAEL ORKIN: That's actually an important issue

for a couple of scores things.

Like a point spread, like 17% of games land on a three-point

difference.

So if you're a bookmaker and you change the point spread

from 3 to 3.5, you're doing a huge thing.

Whereas if you change the point spread from 1 to 1.5,

that doesn't mean very much.

So in order to move the point spread off of 3, then there

has to be a huge imbalance in the betting.

So there are certain score differences, like 7 is the

other number.

But 3 is the big number.

And then, in fact, you could even buy a half a point at

some casinos.

So instead of getting 10:11 odds, you'll get--

or 11:10, however you want to say it, you'll get 6:5 odds,

which is worse, and they'll move the point spread a half a

point just for your bet.

So yes, the attention is mainly from the book makers

who are much more hesitant to move off of a popular actual

difference in score, a frequent one like 3, and then

also bettors.

For instance, if the point spread's 3.5, then a lot of

the so-called wise guys who are the professional gamblers

will take the underdog, because they know that a lot

of games will end on 3, so they'll take the 3.5 points

knowing that they'll win a three-point game, a

three-point loss.

So that brings up another little dynamic of sports

betting, and that is there are basically two camps of bettors

who the bookmakers call the squares and the wise guys.

It sounds kind of funny, and it is.

The wise guys are the professional bettors and the

squares are like, you go to Las Vegas for the weekend and

you bet on the home team.

But they actually use that.

They will not move the point-- if some wealthy square comes

in and makes a huge bet on a team, they're not going to

move the point spread in the same manner that they would if

a well-known gambler comes in and makes the same bet.

So there are some well-known gamblers--

this is a little football betting lore, or sports

betting lore--

well-known gamblers who actually hire people to go

make bets for them, because they don't want to have an

effect on the odds.

And the gamblers who go bet on them are called beards who bet

that for them.

AUDIENCE: It sounds like a whole [INAUDIBLE PHRASE].

DR. MICHAEL ORKIN: So your question I think is do to the

wise guys do better than the squares?

AUDIENCE: Not only that, but are there people who are so

good that they make a lot of money from it, as much as--

DR. MICHAEL ORKIN: Well let's put it this way.

There are people who are so good that they make a lot of

money from the sports betting industry.

Now, I know a couple of these people, and I know that they

make a lot of money by selling information.

Whether they actually make any money by betting on games is

an open question.

But they will have tout services, in other words,

they'll sell.

There's this publication called The GoldSheet that

comes out of Los Angeles, which you

can buy at any newsstand.

It costs I think $7 now or something.

It comes out every week.

They did it in football and basketball.

So the guy who originally found it was

an avid sports bettor.

He passed away now, but I used to know the guy, and I don't

know if he ever made anything actually betting.

But he had 35,000 subscribers to this weekly little sheet,

plus lots of people who buy off of newsstands.

So he made a lot of money in the sports betting industry.

So typical gamblers--

but there are some people who seem to make money at it.

I don't know how much.

Anything else?

AUDIENCE: [INAUDIBLE].

DR. MICHAEL ORKIN: Some of them have performed well.

I see there are a few.

I don't really market it anymore because I'm too busy

doing other things, but it's available on our website.

There are some people those who--

a group of users who swear that the strategies work, and

that they made lots of money.

I personally haven't really been doing much betting.

I'm more of the developer, not the gambler.

AUDIENCE: I was wondering--

I don't know too much about sports betting in the US, but

where I'm from, the [INAUDIBLE]

have quite reasonably different odds spreads against

what they're offering.

Is that the case here or would that be

something else entirely?

DR. MICHAEL ORKIN: Typically in football, it'll stay pretty

much the same, because as soon as one changes a little bit,

people will go and bet on what they think is the right team

and it'll force them to comply.

So it's just the betting public is so versatile,

especially because of computer technology and stuff like

that, they won't let odds be much different from one casino

to another.

Because you can do things like play hedge, you could hedge if

there are different odds.

And so you can make bets on each team in different places

if you have different points spreads.

So hedging is not easy in pro football betting because

everything, especially with online gambling, there's so

much of it going on that everything is within a half a

point usually.

So you can say it's the point spread plus or minus half a

point just about everywhere by the end of the week.

During the week, if an injury occurs or something happens,

then, of course, the point spread might move a lot.

Any other--

I'm going to go to something else now.

It still has to do with sports betting.

So let's see.

So this is the website.

If you're interested you can download a free 10-day copy.

If you send me an email I'll send you an

authorization key to use.

You can use it for 10 days anyway.

It's snoopdata.com of the software.

I'm going to talk about how about another little--

we have a few minutes.

I'm going to talk a few minutes about an actual case

that I got involved in as an expert.

So I'm going to talk a little about internet gambling.

And I'm going to answer your question in a minute.

Internet gambling is about a $12 billion industry.

Actually, that's more like a $20 billion industry now.

Illegal in the US except for horse racing.

So I'm not just talking about sports betting.

Internet gambling is illegal in this country, technically,

except for horse racing.

Now, what does that mean?

So I mean I'm sure a lot of you have seen websites like

PartyPoker where you can play poker online.

If you play on PartyPoker in the State of Washington, you

are guilty of a felony right now.

You're not going to get arrested, but there's a new

law in the State of Washington that makes internet poker

playing a felony.

So it's a very weird situation.

Now, notice I say except for horse racing.

So that's interesting.

And in fact, there's a law that's sort of moving its way

through Congress right now, in which the people supporting

the bill and making these impassioned speeches on the

floor of Congress about how internet gambling is

destroying our moral culture, and ruining the lives of

compulsive gamblers, and it has to be outlawed, except

horse racing.

Horse racing, of course, betting on horses is legal in

California and various states.

So they sort of have a lot of lobbying pressure.

So horse racing for some reason is above the fray here

and isn't immoral.

You can bet on horse racing on the internet and it's

perfectly OK.

So right now, except for horse racing, there are over 2,000

internet casinos, and they're all offshore, except for the

horse racing websites.

There are about 23 million internet gamblers in 2005,

eight million from the US.

In April of 2005 the World Trade Organization ruled that

the US is violating certain rules because they don't treat

foreign and domestic internet gambling businesses equally in

the sense that everything's illegal except horse racing.

So far nothing has happened.

However, recently, the Federal government has arrested and

thrown in jail two people who are CEOs of internet gambling

websites offshore.

One guy from England who is the CEO of the company that--

a well-established website completely legal in England.

And in fact, it is traded on--

it's a publicly traded company on the London Stock Exchange.

He was going to Jamaica and he had a stopover in Dallas, and

the Feds busted him in Dallas and threw him in jail where he

still sits a couple months later.

It's not quite clear--

he's charged with violating the US wire laws.

But still, no one's gone after individual

bettors at this point.

So there are some articles that talk about these things.

So here's an article from the Louisville, Kentucky Courier

Journal, which talks about how horse racing is good.

Here's an article about that poker law from the Seattle

Post Intelligencer, which talks about how poker is bad.

Then here's one of these trade organizations, the American

Gaming Association, so if you Google the American Gaming

Association, you might find this article.

I can email a copy of this if you want, if you're interested

in where these stats come from that I'm putting on some of

these slides.

They're one of the organizations that gathers

data on that industry.

So here's a story.

There was a businessman a few years ago,

about five years ago.

And he raised $50 million in investments for a company that

was supposed to be on the verge of obtaining a patent

for a lucrative new tooth whitening product.

Raised $40 million.

There was such a product, but this guy had no

connection to it.

He was a swindler.

In fact, he had seen this product advertised on the

shopping network.

So he started raising money for it, claiming he had a

patent and that he was going to sell it

to Proctor & Gamble.

Well, he spent $10 million of this $40 million to buy homes,

condos, nice cars.

And he gambled away all the rest of it, $30 million

betting on baseball at an offshore casino.

So he went completely broke.

The investors started asking for their money.

He promised them 100:1 return on their investment, and he

didn't have any.

So he got busted, he declared bankruptcy,

he was sent to prison.

So the victims of the swindle retained a law firm that

specializes in recovering funds from bankruptcy cases.

So they got the homes and the car seized and got the money

back from that--

they got a few million bucks back from that.

Enough to pay the attorney fees, basically.

Then they filed a lawsuit against the casino.

They said a bunch of things.

One, they said the casino wasn't

doing its due diligence.

They should have known this guy was a swindler.

This was all done online and by telephone.

They said with this kind of money and stuff, they should

have checked up.

Well, in fact, the guy was betting directly by wires from

the Bank of Montreal that were sent into

the offshore account.

So the Bank of Montreal was equally culpable.

So they sort of gave up on the due diligence argument.

So then they alleged in this lawsuit that the online casino

has to return the money because that the casino was

either cheating the swindler, or the casino was involved

with the swindler in a money laundering scheme.

In other words, the swindler wasn't really losing this

money, he was just feeding it to this offshore account, and

he was going to later split it with the

guy who ran the casino.

Well, the guy who ran the casino had pretty good records

and he could show that he was keeping that money for himself

and wasn't going to share it with anybody, especially some

guy who just got out of jail and was

working at a gas station.

So they decided to focus on this.

If the casino had given the swindler proper betting odds,

he couldn't possibly have lost $30 million

over a two year period.

In other words, the result couldn't

have been due to chance.

Sort of what I was talking about before

with the good strategy.

In other words, the swindler's losses were just too extreme

to be explained by chance.

Well the swindler was betting on baseball games.

He was betting on 10 games a day, 15K a game.

And when the baseball season wasn't on, he

was betting on hockey.

So if you translate that, $15,000 a bet, 10 bets a day

over two years, you get about $110 million in betting

action, which is called churn.

So that's how much betting action we have from this guy

before he finally got busted.

He didn't go broke, he just got busted.

I guess he was pretty much broke at the time.

So the question is how much on average will a gambler lose

when he bets $110 million in relatively small

increments over time?

And of course, the answer is it depends.

It depends on the bets.

In particular, it depends on the house edge, which is

sometimes called the hold percentage in sports betting.

So with roulette, the house edge is 5.3%.

So if the gambler makes roulette bets with a house

edge of 5.3% at that rate, 10 bets a day, $15,000 a bet,

you'd lose an average of about $6 million.

In fact, you can also compute a

probability here for roulette.

The chance of losing $30 million or more making those

kinds of bets is less than 1 in 1,000.

So it's really unlikely that if this guy was playing

roulette he would lose that kind of money.

As we've seen, sports betting is different from roulette,

but still you can talk about the hold percentage.

Well, in baseball betting, just real quickly, there's

something different than a point spread, it's called the

money line, and the money line is just another way of stating

payoff odds.

So the favorite has a negative money line, which means you

have to bet more to win a certain amount.

The underdog has a positive money line.

So it's all in units of 100.

So anyway, you can take these money lines, and what they do

is they shade one of the money lines so that there's a slight

imbalance so it isn't a fair bet.

That if you bet on both teams you break even.

So when they do that, what they'll do is they'll take 20

points off the underdog line, starting with some number that

they think is good for dividing the

betting action again.

The bookmaker makes a certain profit.

And in fact, for a typical baseball bet, the hold

percentage will be in the neighborhood of 4%.

Similar to point spread bets.

So baseball, just betting on a team in baseball, yields about

the same profit margin for casinos per dollar bet.

and it depends very much on how much money is bet on each

team in a game.

But anyway, so typically that's what bookmakers make

though on baseball betting, somewhere around 4%.

Well, if you do that, if the swindler had been making

random bets with a 4%, 3.8% house edge for that whole run

of $110 million, he would have lost about $4 million.

If he actually was using any skill in baseball betting, he

might have even done better.

He might have lost less than $4 million.

But you can have skill and still lose money, of course.

You can have skill and just do worse than randomness, and

because of the odds, you can still go broke.

So you can be skillfully bankrupt.

But in any event, this guy lost $30 million.

So the question is was he being cheated?

Well it turns out there's another kind of baseball bet

called a parlay bet.

Some of you may know what a parlay bet is.

Parlay bet just means you bet on a bunch of different events

and they all have to happen for you to win.

So for example, here's a 14--

and you can do that in any sport--

here's a 14 parlay.

You bet that San Francisco beats Florida, Pittsburgh

beats Milwaukee, Oakland beats Kansas City, and the Yankees

beat Boston.

If you make that bet, all those teams have to win in

order for you to win your bet, otherwise you lose.

So the payoff odds are somewhat high.

They're huge payoff odds.

In fact, you get the payoff odds by multiplying the

probabilities together for winning for each of those

teams, roughly speaking.

Taking a little bit off for the bookmaker profit.

So it turns out that these types of bets are highly

profitable if you win, but on the other hand, there's a huge

house edge or hold percentage because the casinos don't want

to expose themselves to the risk of taking huge parlay

bets because of the big payoff odds.

So they'll make the odds such that they're not so good for

the bettor.

So a typical 14 parlay bet, the house edge is

between 25% and 30%.

The same is true if you make--

if you look at games with high payoff odds, the house edge is

always worse, like the State Lottery, for instance.

The house edge equivalent is 50%.

The State keeps 50% of all the money bet.

Keno is another game.

If you've seen the game of Keno in a casino, that's a

type of lottery.

Keno bets typically have a 20% to 30% house edge because you

can win a whole lot of money for a $1 bet.

So on games like that have very high payoffs with a very

low chance of winning, the house edge is typically high.

So these 14 parlay bets, this guy was betting 4, 5 and 16

parlay bets generally.

So he was exposing the casino to-- he was betting $15,000 a

bet, these 14, 15, 16 parlay bets.

So any win, he would get hundreds of

thousands of dollars.

So the online casino was hesitant to take the money on

such large bets.

They don't usually, so they spread it out around other

casinos, so the money trail gets sort of dispersed.

And anyway, let's just look at the bottom line.

Let's say the house edge is about 27.5%.

And that's what this guy was doing was making

these parlay bets.

And the $110 million bet, you multiply by

27.5%, you get $30 million.

And that's exactly what the guy lost.

So I'm going to stop with that because we're running out of

time, but I will answer questions.

Anybody want any other information, just come on up

and get it.

Thank you very much.