Sabermetrics etc. and Gambling

The Game Room
1

These days there has been an enormous amount of statistical analysis on sports, much of it about the value of individual players, but some of it geared to predicting likelihood of victory. You see it all over the place, e.g. 538.com, but also elsewhere including here. It’s to the point where people are running simulations and models on each play of big games, to tell you what the updated likelihood of victory is based on the latest play.

The thing is that if you can really determine accurate odds on victory, you could theoretically make a lot of money gambling on the sport. So the question is how feasible this is. I see several (overlapping) possibilities:

[ol]
[li]It’s actually feasible.[/li][li]It’s not feasible because the increased odds of success that you have from these statistical predictions is not enough to overcome the likelihood of being ruined by the inevitable bad run.[/li][li]It’s not feasible because the statistical methods tended to validate the gambling odds already in use.[/li][li]It’s not feasible because the increased of statistical methods has tended to influence the gambling odds themselves, whether in influencing the bettors or bookies or both.[/li][/ol]
Any thoughts? Data?

2

It’s feasible, but it requires a great deal of work. The problem you face is that ten thousand other people will do the same thing. It doesn’t take a real genius to figure out sabermetrics (or whatever the term is for a given sport) as a result of which

  1. You have to constantly discover new facts to keep ahead, and
  2. You must have a very large bankroll, in order to make many bets, because your edge will be very narrow.

The living example of this is Bob Voulgaris, who made a great deal of money betting on NBA basketball. The thing is, Voulgaris studied NBA basketball to an extent that would put almost any sportswriter completely to shame; it wasn’t just basketball he studied, but coaching tendencies, how payers behaved in different contractual situations, the effect of specific travel schedules, and a thousand other things. He spent way more time on studying professional basketball than most people do at any job - but IIRC, his winning percentage was only something like 55 percent. If you’ve got the bankroll to bet on every game’s result, the over/under, and many other side bets, that’s guaranteed money. If you can make 10,000 bets at an expected win rate of 55 percent, there is almost no chance you will lose. But few have the bankroll.

The classic Voulgaris example, of course, is that for a long time he was making a stupid fortune on the side bet of “halftime points,” e.g. how many points both teams will have at halftime; until the early 2000s, the oddsmakers would figure this number out by simply taking the estimated points for each team, based on their points scored and allowed in regular time, and cutting that in half. That is, of course, incredibly stupid, because it is really obvious that NBA teams score more points in the fourth quarter than they do in any other quarter; the oddsmakers were systematically overestimating how many points would be scored in the first half. Voulgaris noticed this and just bet “over” on almost every single game, and won millions. Of course, whatever edge he has now, he’s not telling you.

But eventually everyone caught on and he lost that edge, so he had to study more and find other edges. I believe he endded up actually hiring a mathematician to construct incredibly advanced models. Of course, the model but be subject to constant maintenance as teams trade, release, sign, a guy’s got a bad ankle, the coach’s job is in jeopardy, so on and so forth.

Now everyone has a computer and everyone’s done this math. What you knew five years ago isn’t enough, isn’t even close to enough. You need to know more than the professionals do - it needs to be your two full time jobs AND you have to be just a really clever, insightful person.

3

Fascinating stuff, thanks.

Article about that guy here.

4

Typical payoff is 10-11 so you need to win 52.4% just to break even. Your friend’s advantage was therefore 5% — very substantial, but not like the huge 10% advantage from .55 -v- .45 on an even-money bet. (Using the Kelly Criterion he could bet 5% of his bankroll on such a bet if he were confident that his 55% assessment was correct. Disclaimer: Winnings must be reported to the IRS; this will affect net returns.)

I have a friend who became a famous multi-millionaire with various ventures, one of which was basketball betting. He once offered me a full-time salary to base myself in Las Vegas and just place bets for him! Of all the many zigs where I should have zagged in my checkered career, I wonder what accepting his offer might have led to! :rolleyes:

5

The real trick is to know where (and when and how) people bet irrationally. See, the Vegas oddsmakers don’t actually care about the probabilities per se: They only care about how people will bet. If the bookies have set the odds (or over-under or spread or whatever) at the right point, then they don’t care what the actual result is, because their profit will be identical in every case. And they’re very good at finding that right point.

Now, if every bettor were making their bets based on the actual probabilities and expected values, then the bookies’ figures (if set properly) would end up being equivalent to those probabilities. But that’s not the way everyone bets. Suppose, for instance, the alumni of one particular school were known to be die-hard fans, who would always bet on their team, even when it’s a negative expected value. The bookies would then set odds that overestimate that team’s chances, because that’s what’s needed to get balanced bets. So if you can identify such situations, it makes sense to bet against the popular team.

6

As an example of what Chronos speaks of, I’ve heard that betting against the Dallas Cowboys was a money-winning idea. “America’s team” often won but failed to beat the spread. Unfortunately, Googling led me only to the ten most recent years of results, but the idea would have still won money during 2007-2015. Dallas vs Spread was 30-42-0 at home during these years, and 36-35-1 away.

The idea lost during the 2016 season: Dallas was 5-2-1, 5-3-0.

ETA: One bookie blunder my friend sometimes took advantage of was parlaying the Win and the over/under. For example, my friend got the bookie to pay off 3-1 if my friend picked both team and over/under correctly; he lost otherwise. This is a valid parlay, from bookie’s viewpoint, only if winner and over/under are uncorrelated — they’re not.

7

There’s one more (also overlapping) possibility:

 5.  It's not feasible because you win too much and bookies refuse to take bets from you anymore and/or modify their rules so you can't bet like you want

I don’t know how much this is a factor in sports betting, but it’s THE crucial factor in blackjack card counting.

As noted, there are examples given above of bad (for the bookies) bets that bookies used to offer, but presumably don’t anymore – not because the bookies did the statistical research themselves to set odds better, but because the bookies noticed Mr X kept winning these bets.

8

[/QUOTE]

Has your browser been hijacked also? That makes two of us! My hijacker is pwwysydh.

9

septimus, I’m not seeing anything like that in Fotheringay-Phipps’ post. Maybe that’s from your own browser, too?

10

Bookies care more about there being equal action on both sides of the line, and this is the main reason why sports betting can be profitable - knowing when the odds are better than they should be because there are a lot of less savvy betters throwing their money around.

I might imagine that bookies might take a slightly larger vig from those who are known to spot good bets more often than not, but I can’t imagine them not accepting the bets at all. It’s not like the person can continuously bet against you like they would in Blackjack; sports events only happen with a certain frequency, and there’s only so many events that have action available on them.

11

I don’t see it in Fotheringay-Phipps’ post; I see it in Quercus’ quote thereof. I also see it, just now, on my son’s computer, which runs a different Windows version from mine and, presumably, a different set of viruses, if any.

12

FWIW, I also see it in Quercus’ quote. I assume it’s not anything in my computer. May have been something in Quercus’ computer.

13

Oops, yeah, I see it in Quercus’s post, too.

16

The theory I’ve heard is that you should always bet on the team with the smaller fan base.

17

My emphasis to the above quote. Among professional sports bettors 55% is the Holy Grail. Although you see people advertising their 60+% winning systems it is statistically impossible to achieve results like that in the long term.

18

Except that as soon as the first person with a deep bankroll spots this pattern, they’ll bet in the opposite fashion to the point where they drag the odds back to neutral again. Arbitrage isn’t just about spotting opportunities, it’s about being the only person to spot them and acting in a way that then eliminates that opportunity for subsequent participants.

19

You’re assuming that most people with big bankrolls are rational (or rather, that most bankroll is controlled by rational people, however that’d distributed). But maybe there are also some of those big-bankroll folks on the blind fanatic side. Or maybe there are enough blind fanatics that they’re able to collectively overwhelm the savvy big bankroller.

Plus, of course, nobody has a big enough bankroll to saturate the action on every positive-expectation bet that comes along. Something like the Superbowl, sure, that attracts everyone’s attention. But Mr. Moneybags isn’t going to research each and every college basketball game to find all of the underpriced bets.