Wednesday, March 26, 2008

Washington State versus UNC

Washington State travels to Charlotte as a number four seed to face UNC, the number one seed in a Sweet 16 match-up that pits virtually the fastest playing squad in college basketball against the Cougars, who are near the bottom in tempo among college teams.

Thus far in the tournament, both teams have won their games easily, leading some to pick WSU as an upset possiblity, while a few others (very few), such as Ken Pomeroy, see the game as a virtual pick 'em match-up.

Let's look a few things about this game to get an idea of some of the factors. First, of all, neither team has defeated many top twenty-five teams. Carolina has four or five such wins, if you count the Tar Heels win at Davidson, while Washington State beat USC twice and won at Gonzaga, and now Notre Dame, three teams sort of on the fringe of the top twenty of the various polls and rating organizations.

In terms of elite wins, say among the top 15, Carolina won at Duke, while WSU doesn't have any elite wins, having lost 3 times to Stanford, and twice to UCLA. Of these games, only their loss to Stanford at home was close, coming in over-time. They also lost to Cal and barely beat rival UW at home. They also have beaten Baylor, Oregon, Winthrop and ASU.

Surprisingly, Pomeroy believes that Washington State is better than Stanford. I know that styles make match-ups and all that, but Stanford has beaten them 3 times, the last 2 convincingly, and finished two games ahead of them in conference and also went farther in the Pac Ten tournament, so once again this year, Pomeroy's ratings have some rankings that make very little sense. One wonders exactly what Stanford would have to do to be ranked ahead of WSU by Pomeroy if beating them 3 times and finishing two games ahead of them in a conference where everyone plays each other twice, is not enough!

Sagarin's ratings are much more likely to be in line with reality this year than Pomeroy's. Why this is so, I do not know, but something has gone badly askew this year with the Pomeroy ratings. While no one knows his exact ranking formula, there appears to be some system bias in favor of teams that play in conferences in which the play is slower, in that he has the two fastest-paced teams in the tournament rated far more lowly than other ratings services, while UMass, another team that plays at lightning pace is also rated less highly. On the other hand, plodding teams, such as WSU, Wisconsin and Illinois, all of whom play at a snail's pace, are rated more highly by Pomeroy than by fans and the experts in the field.

While his data collection is still top-notch, Pomeroy probably needs to tinker a bit more with his ratings. He truly may be the only single person in the entire United States who believes that Washington State is better than Stanford and essentially as good as UNC.

In terms of defense, Pomeroy has WSU slated as a far better defensive squad. In terms of offense, he has UNC as the better squad. However, there may be some other clues as to why UNC is more likely to win.

First of all, UNC has much better talent. Almost everyone agrees that for WSU to have any chance, it probably needs to be a game in the 50's or 60's, to keep Carolina from having extra possession and a fast game in which it can exploit its depth. This is a fancy way of saying that Washington State doesn't have all that many good players so they have to try to take the air out of the ball to have any chance to win.

I have to admit that I am a bit more biased against this team than perhaps against other foes, because they pride themselves on their plodding way of playing. They are pretty much the slowest-playing team in the entire NCAA Division I, brandishing a style of basketball that, if further adopted, could be a new non-narcotic form of sleep aid. Their coach is already trying to turn this into some sort of grudge match against Roy Williams, who he claims, was disrespectful to his father back in the 1990's, and apparently, said father also taught players to play as though they had on muddy workboots.

Well, as pathetic and boring as WSU's style of play might be, it is still within the rules, so let's look at some other factors.

Looking at something Pomeroy calls consistency rating, we see that UNC is the 59th most consistent team, which is pretty good--only Memphis and Western Kentucky are ranked ahead of them among teams still in the tourney. WSU is 311th, which is very poor--only West Virginia is less consistent among tourney teams.

Now, I suppose consistency can cut both ways when looking at a one game match-up, as opposed to winning the entire tournament, where it is essential to be consistent. We can expect North Carolina to be pretty good, just like they have been all season. But WSU, will it be the team that beat Notre Dame badly, or the team that lost badly to a mediocre Arizona team twice.

Pomeroy also has a figure called luck rating in which he rates UNC as being one of the luckier teams left in the tournament, which is most likely due to a slew of close games they had while Ty Lawson was hurt. He also has WSU as being relatively unlucky this year, which is truly puzzling. A glance at their results indicates that they split two overtime games this year and that most of their losses have been by more than three points. Their schedules appear to be relatively even, with UNC playing tougher defensive teams and WSU playing tougher offensive teams.

There have recently been some articles published that state that Vegas uses the Pomeroy ratings to set its lines. Vegas is currently predicting a 142.5 over/under, which is 2.5 points higher than the Pomeroy prediction, so that could possibly be the case here.

However, if Vegas used the Pomeroy predictions to set its points plus or minus, then the public immediately moved the line far away from the Pomeroy setting.

Pomeroy predicts that UNC will only win 71-69, which seems frankly amazing. His system predicts that WSU will be playing in a site three time zones away, on a highly partisan "neutral" court, against a team that everyone agrees has far more talent and that UNC, with all those advantages, is only a 2-point favorite.

This is ludicrous and indicates something deeply wrong with the manner in which he is currently rating teams. WSU has an excellent shot at winning. In a one game and out tournament, unlikely events occur somewhat frequently. Davidson beat Georgetown, albeit while playing in Davidson's home state. Nevertheless, there is a huge difference between a team winning an upset and being deemed essentially an equally good team, which is what Pomeroy would have you believe is the case in the WSU-UNC match-up.

It is difficult to perceive how Pomeroy's system could predict such a thing, about as difficult as figuring out why he ranks them above Stanford. The current Vegas odds are Carolina by 8 points, which seems about right given that WSU will try to hold down the score as much as they possibly can.

Everyone seems to agree that Washington State is a very unselfish team that maximizes their chances to win, given the talent at their disposal. But is that likely to be enough? Pomeroy says almost, and he means this in general, seeing them beating Carolina 43 times out of 100.

While I am definitely not neutral in terms of my rooting interests, I am not sure any other neutral observer can agree with Ken Pomeroy.

Washington State has not beaten any of the supremely talented teams that they have faced this year, losing all 7 games against UCLA, Stanford and Arizona, a disappointing team with injuries, but top-notch talent. They may finally achieve that break-through win Thursday night in Charlotte, but if they do, it will be a huge, overachieving upset, not a battle between two even teams.

I have to believe that WSU has not shown much ability to beat teams that exceed them in talent. They have beaten decent squads like Arizona State, Baylor, Winthrop, Oregon and now Notre Dame. They have not, however, beaten anyone that would surprise you while reading your morning newspaper. With virtually all the intangibles leaning UNC's way, it's hard to see them doing that now.

14 comments:

Anonymous said...

You are off the mark on a few things here.

First of all, Pomeroy's ratings are off on Carolina because they don't account for the games Lawson missed/played on one foot in. That's how they have always worked. They aren't broken.

Secondly, UNC isn't being penalized for playing fast. They are rated below slower teams such as Wisconsin because they don't play defense nearly as well.

Thirdly, his projection is based off this game being played at a neutral site. Now is Raleigh a true neutral site in this instance? Probably not, but his formula does not know this.

The large gap between Pomeroy's prediction and the odds makers line is due to a) the Lawson injury b) the venue and c) public perception.

It has nothing to do with his numbers needing "tinkering".

Jane Elizabeth said...

Well, first of all the game is being played in Charlotte, which is probably a slightly better venue for UNC than Raleigh, as it has served as a home site for Carolina in the past.

With respect to the impact of Lawson's injury, we don't know exactly what impact that has had on the prediction. Although some of those games were closer and they did lose to Duke, UNC did go 6-1. Pomeroy hasn't addressed this, but he is still using the same base log numbers for his multiplicative predictions without any disclaimers for the faithful.

Third, I understand all about tempo-free statistics. My question is whether bias has crepted into the system.

I notice that you don't attempt to defend WSU being ranked above Stanford, which is a true anomoly and embarrassment to his rankings, although I do respect him for sticking with them until the termination of the year. He has adjusted his formulas before and I expect he will again.

Anonymous said...

So just because a team beats another team 2 or 3 times they are automatically better overall?

That's interesting.

I didn't know that Purdue was better than Wisconsin nor was I aware that Providence was better than Uconn.

Thanks for showing me the light.

Vegas Watch said...

"Pomeroy hasn't addressed this, but he is still using the same base log numbers for his multiplicative predictions without any disclaimers for the faithful."

This has been explicitly addressed.

http://www.basketballprospectus.com/article.php?articleid=152

This disputes both your points that the issue has not been addressed, and that Lawson's injury did not have an effect on the ratings. Clearly it has been mentioned. Would you like to to be brought up any time he posts any sort of numbers? I think he gives his readers a lot more credit than this.

They Heels were obviously significantly worse without him. The fact that they were 6-1 in his absence is essentially irrelevant, since the whole point of the system is to treat a one point win about the same as a one point loss. This is why it is good at predicting future events.

Why, exactly, do you feel he should adjust his rankings? Because Washington St. is above Stanford? I don't think the tournament games thus far have done anything to disprove that. Since he only has UNC beating Wazzu by 2 points? Obviously the system is not perfect, but it would be unreasonable, and counterproductive, to try to make arbitrary adjustments for injuries.

David Hess said...

Come on Moneyline, you're twisting his argument here. He's not looking at the head-to-head wins in isolation, he's combining them with conference record.

Take out the games vs. eachother, and Stanford and Wash St both went 11-5 in the Pac 10. Nothing there to show Wash St is better.

Do the same for those other pairs of teams...
Wisconsin 16-0
Purdue 13-3
Uconn 13-3
Providence 4-12

Jane Elizabeth said...

I don't think those were my exact words, moneyline. I said nothing about two times. First, of all, you decided to lower the threshold to two times.

I also said that styles make match-ups and I know that sometimes a certain team is just difficult for another team. That was the case of Hofstra against GMU a couple of years ago and no, Hofstra wasn't better than George Mason.

But there was a lot more here than that.

I said that if a team beat another team 3 times in a league where all the teams play the same schedule, and that team also loses 2 fewer games in conference and that team also goes farther in the conference tourney(and that team also loses fewer games overall during the regular season,) it is difficult to know how that team can be ranked lower.

Did any of those things happen in the examples that you presented?

Pomeroy's ranking with respect to WSU and Stanford is wrong, period, and it will be even if WSU wins tonight and Stanford loses its game.

Both Sagarin and Greenfield have Stanford rated ahead of Washington State and I am at a loss as to how anyone could argue to the contrary at this point. If WSU makes it to the Final Four, perhaps you could argue that they are trending or something, but that is not really such a sound statistical argument either.

NC State beat Houston in 1983 and Houston was better, way better. That is why it was such a great upset. Villanova beat Georgetown in 1985 and Georgetown was better, way better. That is why it was such a great upset.

I find it ironic, that all of these people puffing up WSU are going to deprive it of a great upset, should they win, because of the argument that Pomeroy has them essentially equal. If WSU wins tonight, it will be a historic upset and they should be congratulated for a great win in this one and done caldron called the NCAA tourney.

David Hess said...

"I said that if a team beat another team 3 times in a league where all the teams play the same schedule, and that team also loses 2 fewer games in conference and that team also goes farther in the conference tourney(and that team also loses fewer games overall during the regular season,) it is difficult to know how that team can be ranked lower."

Dude, take out the head-to-head games, and none of your other arguments are true. In fact, then Washington State has less losses on the season.

If you want an argument why Wash St is better, they have less double digit losses and less bad losses (Siena?).

Oh, and one last thing. Sagarin's predictor rating has Washington State rated above Stanford.

I'm not saying Washington State is clearly better, just that the ONLY thing supporting Stanford being better is the head to head results.

Jane Elizabeth said...

And I appreciate all the comments, so thanks, in advance.

With respect to Lawson, Gasaway did note that the offense seemed worse, but the defense was the same. It is not clear to me if this affected the ranking substantially or not.

They did, counting Lawson's two missed games earlier in the year, go 8-1 without him, all in games against either teams that made the NCAA, the NIT, or against ACC teams.

In terms of the log summations, Pomeroy made a big point of stating that Kansas had apparently, 8 times the chance of winning it all, compared to UNC. That simply sounded strange to me and I don't recall him mentioning that his numbers were somewhat suspect due to injuries.

Some of you may recall that in the past he did rolling five and ten game ranking as well, which may have been just as instructive as these season long ones, but for some reason he has stopped providing those.

One reason, sadly to me, why I have begun addressing these statistical issues is that Pomeroy and Gasaway,with whom I have been friendly by email over the years, have lately been very unresponsive. I have trusted them in the past to explain certain anomolies and unfortunately, I am not getting what seem to me, to be sufficiently explicative answers.

I have also seen a couple of areas where they seem not to be following the lead of Dean Oliver, whose book Pomeroy touted and inspired me to buy.

Hey, it's their high time of year. Maybe they are really busy and having trouble getting use to the new format, but Ken's blog in the past, combined with Gasaways' Big Ten Wonk, seemed much more adept at answer questions and providing convincing feedback.

Jane Elizabeth said...

I guess if you delete WSU's three losses to Stanford, then they would have fewer losses than Stanford. That seems to me a truly remarkable way to make the argument on behalf of WSU.

Would Stanford get to delete three losses as well?

I haven't even argued about Stanford getting robbed in their second game against UCLA because it seems pretty clear to me as it is, but I find the arguments interesting. It is too bad that Pomeroy refuses to make them on his own behalf.

Stanford may simply be disfavored by fate as their reward for finishing ahead of WSU was getting Marquette, who is arguable a much better team than Notre Dame.

For WSU, their problem is not just Stanford, they didn't come close to any of the other top talent teams in the Pac Ten either, going a big 0-7 against Arizona, Stanford and UCLA.

Why, with the exception of one of the games against Stanford, was WSU unable to really compete with those three teams? You guys probably watch more Pac Ten ball than I do, although I try to stay up anytime UCLA or Stanford is playing and I tried to follow ASU to see how Sendek was doing.

David Hess said...

OK, a couple points...

I have the Pomeroy ratings from March 16 (Selection Sunday), and at that point Stanford was rated ahead of Washington St, though just barely (#10 vs #11). Pomeroy's ratings are weighted to emphasize more recent games, so you can blame them being rated ahead of Stanford today pretty much entirely on the performance in the tourney.

Also, I took those March 16th ratings and did a UNC injury-adjustment. I separated UNC's games into WITH or WITHOUT Lawson, then adjusted the individual game efficiency ratings for opponent strength (using a formula confirmed as correct by Pomeroy) and averaged those adjusted ratings.

Situation ... Tempo/Offense/Defense/Pythag
WITH... 74.7 / 127.6 / 93.2 / 0.9736
W/OUT.. 72.5 / 118.4 / 84.4 / 0.9801

What to make of this? Beats me. But it clearly shows that their non-Lawson games are NOT dragging their rating down. Their offense suffered, but their defense was great in his absence. I wondered if perhaps they could be even better now. Maybe their offense would be back to normal, but they would keep playing defense at their new level. So I looked at the WITH-Lawson numbers before and after his injury absence.

PRE... 76.6 / 123.6 / 92.2 / 0.9667
POST... 70.4 / 136.9 / 95.6 / 0.9842

Well, they're playing better, but NOT because of defense. I'm thoroughly confused now.

rolub said...

"I haven't even argued about Stanford getting robbed in their second game against UCLA because it seems pretty clear to me as it is, but I find the arguments interesting. It is too bad that Pomeroy refuses to make them on his own behalf."

Getting "robbed" of that W is basically the same as if they had won the game by a point. Arguing that if they had won that game, the Pomeroy numbers would be even more out of line suggests a lack of understanding of the analysis. I thought it was clear to most that the analysis is based on per-possession metrics and relative percentages, and not the absolutes of wins and losses.

Jane Elizabeth said...

That is a good point, Rob. I think we were sort of going in and out of using the ratings, in terms of the different arguments people were making.

I still have real questions whether the ppp analysis is better than pure points. My inclination is that pure points is very tough to beat, when you have a large sample size.

I hear all kinds of ludicrous stuff out there. For instance, UCLA will benefit from having won a close one, because most champs have to win a close one. I doubt it. I think you would always rather win by a lot and winning by a lot is probably the best individual predictor of future success. I don't remember UCLA having a lot of gut checks back in the 1960's and early 1970's, but of course they also played at a far faster pace, which may give a team more margin for error.

It seems like this year's UCLA basically always has close games. It might make Pomeroy feel good to say that 50 is to 40, as 100 is to 60, but that doesn't erase the fact that a couple of lucky three's are much more likely to erase a low scoring team.

And David, I appreciate all that work and those calculations. Apparently, I am not the only obsessed hoops fan today. My feeling was that the Heels were definitely worse the first couple of games without Lawson, against FSU and Duke and then in the beginning of the second Clemson game. They seemed a bit shellshocked but then after the incredible win against Clemson, they seemed back on track. Roy Williams has made no hint of hiding the fact that Lawson is far from his best defender and Thomas did seem to be doing some things better.

Ultimately, I think that Deon Thompson and Alex Stepheson are bigger keys than whether Lawson or Thomas is out there. When they play well, which not necessarily that often, Carolina seems to score at will.

Jane Elizabeth said...

Pomeroy Expects Tweaks to his System

"RealTimeRPI.com noted that two of his predicted winners lost in overtime. And Mike Gerstein, who developed a bracket using a numerical tool sponsored by Coke Zero, wrote of one of his successful upset picks, “Already Davidson has proven that numbers can lie — that’s why they play the games.”

That was the sentiment expressed by Ken Pomeroy when he declined to participate. “Filling out a bracket robotically based on who the better team is kind of misses the point of what makes this time of year so much fun,” Mr. Pomeroy said. After he was told his ratings nonetheless had outperformed the others, he responded with tempered enthusiasm. “Don’t get me wrong, I do like my system compared to others because it’s based on basketball things and not just final scores,” Mr. Pomeroy said. “But we both know that 48 games isn’t enough to judge it on. And even if it consistently does better than the others, I’m not convinced it is by much.” He added that he’s hoping to tweak his system so that it does a better job of ignoring events once games have already been decided."

http://blogs.wsj.com/numbersguy/number-crunchers-vs-experts-vs-the-masses-306/?mod=WSJBlog

Anonymous said...

Well written article.