## Does Vegas undervalue bad teams and overvalue good teams?

The prevailing view is that Vegas is an example of an efficient market. If there were obvious trends that oddsmakers ignored, it would be easy for people to make money gambling on football, and we know that’s not the case. But I thought it would be interesting to investigate some claims I’ve heard over the years, so I’m introducing the Efficient Vegas tag to Football Perspective.

One theory I’ve heard is that when good teams play bad teams, the smart money is to bet on the bad teams. That’s not because Vegas doesn’t know what it was doing, but that oddsmakers know that fans like to bet on good teams when they play bad ones. But is this true? Here is how I decided to test that question.

From 1990 to 2013, there were 792 games that met the following four criteria:

• the game was played in weeks 7 through 16
• entering the game, one team had a winning percentage of 0.600 or better
• entering the game, the opponent had a winning percentage of .400 or lower
• entering the game, the two teams had a difference in winning percentage of at least 0.250

There were 792 such games. In those games, the team with the better record went 378-399-15 against the spread (and 580-211-1 straight up, for a 0.733 winning percentage). If you had bet \$110 on the underdog in each game — assuming you need to bet \$110 to win \$100 — you would have won more often than you lost but you would have still walked away down money due to the vigorish. In other words, this does not appear to be a theory one could use to exploit the public’s tendency to back favorites, and is therefore evidence of an efficient market for Vegas. In fact, one could argue that these results represent a more lucrative market than a true 50/50 split, if we believe that more people tend to back the favorite (since Vegas would be winning more than half the time). For what it’s worth, in 2013, the team with the worse record went 15-17 against the spread in these games, making this “zig against the public” strategy an even worse one than normal.

Some trivia notes. In 36 of the games (4.5% of the sample), the team with the better record was actually the underdog, with two teams being as large as seven-point dogs. Those games were in 2012, when Andrew Luck and the 7-4 Colts upset the 4-7 Lions, and in 1991, when the 5-1 Lions lost to the 2-4 49ers, 35-3. In those 36 games, the underdog team with the better record went just 14-22 against the spread.

If you want to make a suggestion for a future topic of investigation on the efficiency of the Vegas line, post it in the comments.

• Chris Leveroni

The most interesting question to my mind is whether the market or the experts does a better job pricing the games. If you presume that the original line is Vegas’s best attempt to handicap the game then you could test lines that deviated a certain amount from the original number and see which was more accurate. For example, take a line that started at 6 and went off at 7.5 would one have been more profitable backing the expert line or the market line? You would obviously have to filter out line movements that were produced by major injury or some other abberation.

• Chase Stuart

Thanks for the comment, Chris. Unfortunately, I don’t have any data on line moves, so I can’t answer that question. If you happen to know of a data source, please let me know.

• Ty

sbrodds might have the data. They use an opener and have closing lines for multiple books. Covers might have one, though I’m not sure about that (I know they have the closing line, but I’m not sure if they have the opening line).

• Kriss Bergethon

I’d be curious what the results of the same study would be if you control for the ‘public’ teams. The Vegas sharps tend to watch the games played by the Steelers, Giants, Bears, And Packers because they know those fan bases can tip the lines with heavy betting on their own team. You might also see similar swaying on ‘big’ games – Monday night games, crucial division games, playoff games. These tend to attract a lot more public action that might be swayed by perception more than reality.

• Chase Stuart

I don’t buy much into the idea of “public” teams. Over the last ten years, the Giants (84-73-3) and Packers (84-69-7) have the 5th and 6th best wining percentages against the spread. The Steelers are 81-76-3, and the Bears are 72-81-7. If fans are betting on those teams, they’re winning.

• MP

Potentially interesting: Since 2004, .250 or worse teams (6 games played minimum) playing against a .550 or better team went 139-99 (58.4%) ATS. It suggests that teams with truly terrible records are undervalued by the market, although the record is only 17-15 ATS over the last two seasons.

• Chase Stuart

Good comment. I didn’t check the extremes, so good thinking. The question now is has the market caught up to this or are the last 2 years just the result of a small sample size.

• George

Nice stuff. Re: the ATS split, and in the answer to the question, I think Vegas probably doesn’t but the general “square” (if there is such a term anymore with the amount of information out there) public probably does and that is reflected in the line Vegas sets. There is an assumption that certain teams will keep covering, and the line will get increasingly moved against those teams until the public catches on.

In a lot of those games mentioned, it would be interesting to know how many of the lines were at 7.5, or 10.5 (just to make it more than a touchdown, or a touchdown and field goal and not just a soft/easy cover so to speak). Conversely as it shows efficiency both ways there are probably games where the line was at 5 or 6.5 and it probably should have been 7.5 (and the underdog had a 4th quarter backdoor cover).

I think Kriss’s point above hits it squarely on the head – if you had known public teams as a control (e.g. I am assuming you are getting more “action” on a Giants vs the Cowboys matchup in week 16 say with the NFC East on the line – possibly, than a Browns vs the Bengals matchup with no consequence) you may find a different result. You may find at this point you get more evidence that suggests that line makers (who in a lot of cases set the opening line for the bookmakers to work from), will set up a line that will be skewed to exploit the public bias for a specific team (as they will take more action on that team and as a result they are keen to either limit the loss or make more profit by exploiting this bias – through a higher line, whichever way that you want to look at it).

There is a paper floating around that has a bit more on this “Why are Gambling Markets organised so differently from Financial Markets”, Steven.D.Levitt (The Economic Journal 2004) where the point of lines being set to exploit biases is touched on. The data that they refer to in it though, I think it pulled from the Hilton Supercontest from around 2002 or 2003 so you could argue that it isn’t clean data to a certain extent (as competitors may at points have tried to look for the alternative games and approach to differentiate themselves from others who were picking the public teams week in and week out until they didn’t cover).

• Chase Stuart

So how would you exploit the inefficiency?

• George

I guess it comes down to looking at the opportunity and looking at the two teams in question (I think?); just off the top of my head Jacksonville vs Seattle (-20) this year (should it have been more than 3 touchdowns? – I haven’t checked the ratings back) I think we all felt comfortable with, whereas I don’t think any of us realistically expected Denver (-26) vs Jacksonville to be a cover (even though they had the talent on paper, especially when it was in the 27 to 28 range earlier in the week). Looking back on it now with hindsight, it looks a lot clearer, Seattle were the stronger package of the two teams, and they have a significantly above average home field advantage over the last 10 years. Denver just took people out offensively, but leaked points through their defence all year (and Jacksonville weren’t historically bad at the end of it – even though it looked that way at one point and had enough going for them to cover).

• SparkyATS

I find this terribly interesting and appreciate the topic. I would be curious to see how opening lines perform vs. closing lines. There may be a difference in prices available, and thus, results.

An informative website, with some stuff behind the paywall, is Sports Insights. I know we should be wary of touts but they are more of an information service than a picks service. They provide betting data (number of tickets per side) on a real-time basis. This is a very helpful service, yet costly. http://www.sportsinsights.com/blog/80-20-nfl-betting-system/

I think that’s the next step to this analysis is to follow up this article. It would be nice to see someone independent like you test it. The number of tickets for each team is publicly available from most books. However, all tickets are treated the same. \$5 on the Team A counts as 1 ticket while \$5,000 on Team B counts as 1 ticket. Issues like reverse line movement are able to be identified and followed then. (Note: Reverse line movement is when ticket counts are high on one team yet the line moves in favor of the other team. The most logical explanation is that the bets on the other side, with low ticket counts, are actually very high in dollar amount).

I hate personal anecdotes but I have found this betting strategy to be far better than most. Whether it overcomes the vig is a different story.

• Chase Stuart

Thanks, Sparky. Are you affiliated with Sports Insights? If so, can you send me the raw data behind that article?

• SparkyATS

I am not affiliated with Sports Insights and never have been a subscribed. I follow on Twitter, and read their free content. A good interim source for side bet ticket numbers is http://www.sportsbookspy.com

The Sports Insights CEO and Founder Dan Fabrizio is very engaging and I see you reached out to them on Twitter. Knowing them, I wouldn’t be surprised if they let you have a peak at the data. I’ve been impressed with their transparency.

• Michael Terry

Vegas has blind spots. Before the season began I posted in various places online about Seattle’s incredible record against the spread (including preseason and play-offs) when Russell Wilson started. Seattle went on to once again have the best record in the NFL against the spread this season. The silly Super Bowl line was the capstone of my claim. It will be interesting to see if Vegas finally corrects for it this year.

• Chase Stuart

You must be a very wealthy man, Mr. Terry.

• Richie

hehe

• Kibbles

I had read in the past that home underdogs cover at a rate just high enough to turn a very small profit against the vig if you bet nothing but home dogs. This inspired me to Google, upon which I discovered that home dogs have covered just 47% of the time over the last decade after covering at a 54% clip the two decades before that. So, again, Vegas is smarter than us.

That’s the biggest problem with markets like Vegas in the real world. I would call Vegas an efficient market, but when I say that, I don’t mean that there are not any biases or blind spots to be exploited right now. I mean that by the time anyone discovers those blind spots, Vegas will have, too, and they will have adjusted. Similarly, I would consider the NFL draft to be an efficient market. This doesn’t mean I don’t think there are biases or errors of reasoning in the draft process right now, it means that the draft process is very, very good at uncovering those errors and correcting them. Did you discover that short quarterbacks are historically undervalued? Guess what- so did the NFL, and now they’ll be valued properly going forward. In the ’90s, Pittsburgh discovered that the league underrated tweener DE/LB types, and the resulting correction led to the rapid revival of the 3-4 across the league. When Lewin came up with the Lewin Career Forecast, he might have really been on to something- he might have really discovered a class of quarterbacks that were systematically undervalued by the league- but the league is not stupid, and ever since the LCF was published its predictive power has been nonexistent, because the league doesn’t undervalue that class of players anymore.

I am always skeptical of anyone who claims they can model complex markets like Vegas or the Draft, identify inefficiencies, and exploit them going forward. I do not question that people can find inefficiencies in the past… I would just be unwilling to bet any money that those inefficiencies would remain quite so inefficient in the future, too.

• SparkyATS

All are very valid points, Kibbles.

I would add that for such efficiencies to occur enough bettors, and thus information, must enter the market to be efficient. NFL Side wagers are the #1 type of U.S. wager so that’s a deep market. I agree that one single trend usually doesn’t last. I do think some market bias does exist, though, especially toward home favorites, that can be exploitable long-term. There’s just too much dumb money out there at times.

Go off the beaten path though, even in NFL, and venture into quarter/half bets and best of all, proposition wagers, and there’s a lot of chum out there.

• George

Totally agree with this. If you go out there enough (and going off tangent – away from the NFL for instances) you will find an inefficiently handicapped market (as if it has minimal action from the other side, oddsmakers won’t be being a lot of attention to pricing it). In the UK (where I am from) at the last Election, one of our current MP’s (think Congressman or Senator – I think), was priced to lose his seat. Locally we knew this was wrong and a total miss-price, but one TV debate later and the price flipped totally and he went from being a significant underdog to a significant favourite.

With the NFL I think it is as you say that there are probably large inefficiencies in the prop markets (e.g. price of the first score being a safety in the Superbowl the last couple of years, or the price on Devin Hester being the first touchdown scorer in the Colts vs the Bears Superbowl etc.).

• George

Chances are you also read the Levitt article (or possibly picked up the reference to it in Mathletics by Wayne Winston as I did originally); there’s a nice article about how that trend changed in the 10 years after that article was published (and basically what would have turned a profit previously, would now have been a way to guarantee a loss);

http://www.inpredictable.com/2013/08/home-underdogs-are-no-longer-good-bet.html

• Ty

This post isn’t the NFL (though it would apply very much to it), Haralabos Voulgaris, probably the best NBA bettor, ever, has mostly stopped betting the NBA (from what I have read). He said that his ROI in the past few years has consistently decreased, and that market has just about figured out his model (which he has spent millions of dollars on).

The market might be naive initially, but as it gains more information, it becomes much smarter and can adapt relatively quickly. What might have worked one year ago, might no longer have success in the future.

• prowrestlingisstrong

I have no evidence to support this but i always felt the best valued underdog plays were not just straight dogs, but mostly dogs against very sexy, very public offensive teams. These are the favorites that always seem to be over valued. I know this gets into some serious data mining but I would be curious to know against the spread records of teams ranked in the top five of scoring offense (excluding games where two top five offenses face off) in the last ten years. My gut says they would have a losing record as those teams tend to have the most public support and would be the most over valued. But I very well could be wrong.

• jojo

Cantor has their NFL win totals out (http://www.wagerminds.com/blog/nfl-football/cantor-gaming-releases-2014-nfl-win-totals-2-7911/). Is there any value in betting for teams projected to win less than X games (say, 6 games), or betting against teams with lines over, say, 10 wins?

• A different criteria of good vs. bad, but from 1990 – 2013 (using all weeks) underdogs of 10.5 points or more were 55.5% against the spread, up 39.4 units based on -110 vig. Reproducible R code here (with waterfall chart at the bottom):

http://rpubs.com/andylytics/historical_dogs

• Peter

Sort of unrelated, but I’d love to see an analysis of ATS winning or losing streaks, i.e. do teams that win 3-4 games in a row ATS get an inflated value and become more likely to lose ATS in their next game? (and the same phenomenon for teams that lose 3-4 in a row). For example Jacksonville early this year, they just kept losing and becoming bigger and bigger underdogs, until the line got so big (Denver -28) that the “Jags suck” bubble popped, so to speak.

• Dave

NFL seasons are already pretty short. Add in a qualifier like that and the sample becomes pretty small.

That said under the right conditions yes, bad teams do eventually cover at an increased rate. They are difficult to play though. Many of the wins come by 0.5 or a single point while the losses often look quite ugly.