Thursday, March 10, 2005

Breaking RPI News

After seeing the 'team sheet' that Luke Winn posted, the alarm bells sounded in my head (see previous post). But then I read Gregg Doyel's piece this morning, and noticed this...

Until Wednesday, the Pac-10 had been the country's No. 2-rated RPI league for months. Bowlsby himself thought the Pac-10 remained the No. 2 league, referring to that league as second behind the ACC, until Hancock informed him of the change.

The ACC remains No. 1. The Pac-10? It's now fourth. In between are the Big 12 (second) and the Big East (third), which had been third and fourth behind the Pac-10 until Wednesday.

In the computation you have been seeing on my page, the Pac-10 has remained number two all week, and comfortably so. But now I am fairly certain this computation is wrong based on the above the information.

Back in January, when the RPI changes were announced, I mentioned that there were two possible ways for the changes to be applied.

Now that the ingredients have been revealed I will be posting this version of the RPI on the main page. There is one possible bit of confusion. Will they do it the way I tested in that earlier post, or will they do it the way that makes sense? The way I talked about before is to only apply these adjustments to Part I of the RPI. The sensible way is to apply the adjustments to all three parts.

I had no way of knowing which one was correct. I chose to apply it the way that made sense. ESPN and others have done it the same way. It turns out all the lemmings have gone over the cliff on this one.

The RPI is actually being computed the other way. Without independent confirmation from the NCAA - which will never happen - I can't be totally certain. But based on the two pieces of information I have come across over the past 12 hours, I am certain enough to now post data based on the other version of the formula. The way the Pac-10 would be fourth in overall RPI is by applying the home/road adjustment only to the winning percentage portion of the RPI, and not to the SOS factors. [Note: the Pac-10 is still #2 in the average non conference RPI]

So for the last five days of the season, my RPI will differ significantly from what ESPN uses, but it will be nearly identical to what the NCAA people are looking at this week, based on the glimpses that they have let the fans see over the past couple of days.

Tourney Talk

Andy Katz has an article detailing the hopes of bubble teams after talking to some people close to the committee. (But I think it's hard to believe Notre Dame, with an RPI of 92, still has a bid after a stinker of a performance against Rutgers). Katz knew the field before CBS did last March, and if you want to know who is in and who is out before anyone else on Sunday, you would be advised to listen to your local ESPN affiliate as I am sure Andy will be dropping some hints Sunday afternoon.

Elsewhere, Luke Winn has a comprehensive look at the selection process. What caught my eye about this is the 'team sheet' example presented for Illinois. All of the data looks factual...except that it lists Illinois as having played two home games against Ohio State and two road games against Purdue. And the RPI ranks of Illinois' opponents are suspiciously close to the old RPI formula as opposed to my computation of the new formula. For instance Indiana is listed as 49 (vs. somewhere in the 70s on my page). I'm just going to assume that this is a bogus document that the NCAA floated out there because they have to be so darn secretive. But it is a little troubling to me that all aspects of Illinois' schedule are correct (scores and opponents) except for the sites of the two games mentioned above.

I'm headed to the Mountain West tourney, so there will be no post tomorrow. I have updated the stats page and included total rebounding as the extra stat. Congrats to Washington State for taking over the honors for top defense in the nation. I'll update the page again after Sunday's action.

Wednesday, March 09, 2005

Cants Win It All Without Rashad

UNC's Rashad McCants has missed the Tar Heels last four games due to a mysterious intestinal disorder. UNC has won all four games, but not surprisingly, they have not been as dominant without him. Most of the minutes vacated by McCants have been snapped up by the offensively-challenged and smaller Melvin Scott. How has McCants' absence affected his team? Let's look at how UNC performed in the four games without him, and against those same opponents earlier in the season when McCants played.

               With McCants     Without McCants 
-Opponent-    Tempo  OE  DE      Tempo OE  DE  
NC St.         67.9 140 105      62.9 129 113 
Maryland       89.4 122  84      72.6 117 114 
Florida St.    68.0 119  88      73.4 124 103 
Duke           71.5  98  99      68.0 110 107 
Average        74.2 120  94      69.2 120 109 

If you have been reading this blog for a while then you're familiar with what these numbers mean, but for those who aren't...

Tempo: Number of possessions in the game. It's estimated from box score data, so that's the reason there aren't round numbers listed.

OE: Offensive Efficiency. It's 100 multiplied by points scored divided by possessions. Higher numbers are better.

DE: Defensive Efficiency. It's 100 multiplied by points allowed divided by possessions. Lower numbers are better.

From this simple analysis, UNC's change has taken place on the defensive end where they have been unable to keep any of their last four opponents under a point per possession. We're only talking four games here, but the difference is significant enough to take note of.

As a few of the stats page zealots have noted, Carolina's defense has been more exceptional than their offense this season. It's a fact you're not likely to hear in any tournament previews, which will mainly gush about how many points the Heels score. But their adjusted defensive efficiency ranks first nationally, while in the same stat on the offensive side they rank fourth.

I don't think it takes much of a leap to say that it's because of that defensive prowess that UNC is able to score consistently and score quickly. UNC has enough scorers to fill the McCants void, but they aren't getting the fast break opportunities that they used to get, because they aren't forcing as many missed shots and turnovers as they used to. This is reflected in the decrease in tempo, although in fairness, most of the decrease is attributable to the difference in the Maryland games.

The reduction in defense is one reason why UNC is less upset-proof without McCants, but the decrease in possessions is also important because it gives UNC fewer opportunities for their talent to demonstrate its superiority. So if UNC is going to make a run at a title, they need to get McCants back.

Or alternatively, try to suit up the other McCants that will be in the tournament.