From the lines at Pinnacle, we can derive both the chances each team has of winning on Saturday, and the chances they have of winning the championship. Using these two numbers, we can then determine what percentage of the time each team will win the title if they reach the championship game. I have calculated all of these in the chart below.
Kansas' percentage for Monday seems right, as I think they will be favored if they win on Saturday.Carolina's odds make no sense. In the last sentence Kansas was established as the second best team in the tournament. Yet Carolina has a 58.8% chance of winning against them on Saturday, and then their % decreases when facing supposedly inferior on Monday?
More confusion comes when we contrast the odds for Memphis and UCLA. Memphis is favored by 2 points in the semis, thus seemingly establishing them as superior to the Bruins. Yet the "Monday" column shows them winning the championship game only 40.5% of the time they get there, while that number is 48.5% for UCLA. This really makes no sense.
Something is wrong here. Either the futures lines are off, or the lines for Saturday's games are incorrect. This is not limited to Pinnacle- all of the other sites I have checked have the same odds.
I believe that the lines for Saturday's games are off. Let me explain.
For huge events, such as the Super Bowl and Final Four, there books take much more action from casual gamblers than during the regular season. For these games, the ratio of money from casual gamblers compared to sharps is much higher than usual. Because of this, the books are forced to adjust accordingly, and set the lines so they will receive an equal amount of money on both sides. Because these casual gamblers tend to not be particularly bright, this can cause the line to be significantly different than it would be in the regular season.
UNC and Memphis are the popular teams right now, since each played so well last weekend. This is why they are favored on Saturday, and are both getting a lot of action. But let's look back a couple weeks ago, and see where the oddsmakers had each of these teams ranked:
1t. KansasNow, things can certainly change over the course of a couple weeks, but they can't change much. After all, each of these teams are 4-0 over that span. According to these rankings, a KU-UNC game would've been a pick'em two weeks ago, and now the Tar Heels are favored by 3. That is an enormous jump. This is pretty clear proof that the oddsmakers don't actually think UNC is three points better than the Jayhawks. But they know that if they'd set the line at UNC -1, or whatever it should be, they'd get a ridiculous amount of action on Carolina, and would be putting themselves at risk to lose a huge amount of money.
1t. North Carolina
3. UCLA
5. Memphis
The same can be said for the other game- UCLA was two spots ahead of Memphis before the tournament, and now the Tigers are suddenly 2 points better? I don't think so.
It's worth noting that, despite the inflated spread, Wagerline shows 71% of bets coming in on UNC. This may at first seem to refute my argument, but it actually supports it. If the books could set a split action line, they'd do so. But in this case, that number would be so absurd- say, UNC -5, when they should be favored by 1- that anybody with a clue ("sharps", as they are called) would place a large sum of money on Kansas. At this point, setting the line so high would be completely counterproductive- there would end up being more money on Kansas than UNC, which would be terrible for the books, since the Jayhawks would cover that spread the vast majority of the time.
It's a different story with the futures odds. Those lines receive way less action than the sides, so it's likely that the books set them at what they think is a fair price, and adjust them slightly when large bets come in. Because of this, they are much more indicative of what we should actually expect to happen than the game lines.
The percentages above are telling us that UNC is the best team remaining, but they're not three points better than Kansas. The Jayhawks are the second best team left, and will be (deservedly) favored should they reach the title game. And, despite the spread, the Memphis-UCLA game is essentially a toss up, as neither the futures lines nor the Vegas rankings indicate that Memphis is anywhere near 2 points better.
(Thanks to Money Line for helping me get my head around this whole mess. Make sure to check out his completely premature 2009 college hoops previews for the Big Ten and ACC.)
Edit: I switched up some things in the middle, because people were bitching, and it's not an argument I feel like having right now. Wasn't really relevant to the conclusion anyway.
14 comments:
Great analysis. I have KS and UCLA advancing in everyone of my pools...but besides that, it's good to see that in games that are essentially pick'ems that there is an edge.
Excellent post. Are you going to lay any money down on these games? I thought KU and UCLA were excellent wagers this weekend and this seems to confirm that.
I was actually thinking of taking Carolina big, and I couldn't figure out the Memphis - UCLA game but I was leaning to Memphis small.
I'm guessing you would not approve of this action.
UNC and KU I think Vegas has right. I thought that UCLA would be a 1 or 2 pt favorite over Memphis... this does seem a little off. One thing you have to remember is that two teams equal, one team in the west and one team in the east, w/ UCLA and Memphis, the line will always seem slightly partial to the team in the east. More press coverage in the east and a higher population, thus slight favoritism to the team from the east. Remember, Vegas' only intention is to split bets 50/50, not necessarily to set the line correctly
Matt, did you read the whole post? Vegas' only intention is to maximize profits, which may or may not involve trying to split the action 50/50.
Sorry it took me so long to respond here. Corporate Finance exam.
Bobby S- I already have a significant futures stake in Kansas. I still need to figure out what I need to do on Saturday, actually.
B&G- Correct, sir.
Matt- Read the post, please. I understand how the system works.
This might be off-topic a bit, as well as picking nits, but I don't think that the percentages are quite accurate. For instance, where does the 56.7% for Memphis on Sat come from? I presume it reflects an "average" money line of -131/+131 (56.71%) when compared with Pinnacle's line of -137/+125. But, this is not quite right.
If Memphis *really* wins 56.71% of the time and UCLA wins 43.29% of the time, then let’s assume that an equal amount of $$ is placed on Memphis at -137 and UCLA at +125. For every $1000 spent on Memphis, the public wins $729.93*.5671 = $413.94, and the public loses $1000*.4329 = $432.90, for a net loss of $18.96
And, if UCLA *really* wins 43.29% of the time, then for every $1000 spent on UCLA, the public wins $1250*.4329 = $541.125, and the public loses $1000*.5671 = $567.10 for a net loss of $25.975
So, this indicates that if the true odds are -131/+131, then the public is better off playing the team at -137 than they are playing the team at +125. But, this contradicts the premise that -137/+125 implies that -131 is a fair line. If that were true, then the public would lose the same amount at -137 as they lose at +125.
To find the true winning percentage(W) of the favorite based on the Vegas odds, you have to take the amount the Dog Wins on a $100 bet (D), and the amount the favorite wins on a $100 bet (F), and apply the equation:
W = (D+100)/(D+F+200)
So, given the Memphis/UCLA (-137/+125) example, W=56.598% (rather than 56.71%) Probably not material to your conclusion here, but thought I'd share.
I believe the line has changed slightly since I posted this. I do not average the lines, though. For +125/-137, I do (137/237)/(137/237+100/225), which brings me to .565337. With that, I get an EV of -$22.0083 for a $1,000 bet on either side.
I really like the blog and find it very interesting, but I am going to play devil's advocate here. I think that you should reach the opposite here. I think that because their is more money bet on them the lines on the individuals games are sharper than the future lines. The idea being the people that set the lines aren't the ones that make them sharp, the people that bet them are. And the more money there is the sharper the lines will be.
Even though there are more recreational bettors during big events the sharp bettors are still betting much larger amounts than the recreational bettors and thus have more impact on the lines.
Generally the larger max bet size the more sharp a line, thus individuals games will be sharper than futures, props and the like. Based on that I think that UNC is undervalued in the futures market and I would suggest UNC to win the tourny at +185 at the Greek.
But I guess that is just my assumption and I have no data to back that up.
Either way interesting post.
"The idea being the people that set the lines aren't the ones that make them sharp, the people that bet them are. And the more money there is the sharper the lines will be."
This isn't entirely true. The more *smart* money there is on the lines, the sharper they are. And in this case, the majority of the money being wajored is public money, rather than smart money, which is why the lines are inefficient.
You have reached essentially the same conclusing I have, though; in this case, the other side of saying that KU +3 and UCLA +2 are good bets would be to say that the futures odds for UNC and Memphis are good wagers.
"Now, things can certainly change over the course of a couple weeks, but they can't change much. After all, each of these teams are 4-0 over that span. According to these rankings, a KU-UNC game would've been a pick'em two weeks ago, and now the Tar Heels are favored by 3. That is an enormous jump."
I'm not so sure they can't. That 3-pt jump is less than the amount of ground the Heels have gained in the Pomeroy ratings. On March 17th, their ratings (Off/Def) were 126.6/83.6 for Kansas and 122.9/89.7 for UNC, which works out to an 8.0 point edge for Kansas in a 75 possession game. Now the ratings are 125.4/83.7 for Kansas and 127.2/89.6 for UNC, which works out to a 4.3 point edge for Kansas.
The devil's advocate would ask ... if Pomeroy's objective rating system says UNC has gained 3.7 points, why can't the Vegas guys believe they have gained 3?
And remember, the UNC March 17 rating was not being dragged down by Lawson's injury (see here for evidence), so I'm not sure you can blame the Pomeroy change on his system "catching up" to the Vegas guys' evaluation of the team WITH Lawson.
Not sure I believe myself, but thought I'd throw it out there.
The Pomeroy system works a lot differently than the lines. Because of the weight for margin of victory, and for recent games, it will change a lot more quickly than the oddsmakers' opinions of teams.
Also, regardless of their ratings with/without Lawson, I think it's pretty clear that UNC was very underrated by Pomeroy coming into the tournament, and they are catching up with that.
Great post.
My comment is a day after the semifinal games but the question should still be relevant since it is based on Vegas methodolgy:
Can the slightly lower odds for Carolina's Monday odds (57.3% vs 58.8%) be at least partially explained by matchups? Might Vegas perceive a potential Memphis/UCLA opponent to be more difficult for Carolina relative to its matchup with Kansas based on style of play, personnel, etc?
The Carolina line was like that because Vegas didn't actually think there was a 59% chance UNC beat KU. I think individual matchups are taken into account much more by the public than by the linesmakers.
Post a Comment