The general consensus around the NCAA bracket-projecting world is that Marquette is in, and Missouri State is out. In the Bracket Matrix Marquette is a 10-seed, and that's before last night's win over Providence. Missouri State is one of the first four out in brackets across the country. With no games left and potential conference tournament upsets shrinking the "bubble," the Bears will be relegated to the NIT again if the committee reaches the same conclusions as the majority of prognosticators.
Comparing Marquette to Missouri State can give us an idea of what people believe the committee is looking for in a tournament team. The first statistics most people see are win-loss record, and conference record. Here Missouri State's 25-8 and 15-3 numbers far outclass Marquette's 19-13 and 9-9, but of course the Golden Eagles play in the nation's toughest conference while the Missouri Valley is a mid-major that is having an off-year. In fact, the MVC seems to be considered a minor conference this season, or else even Wichita State would be in serious discussion for an at-large.
For the sake of simplifying the argument let's consider the teams' records to be equal given the disparity of their conferences; it's also consistent with the oft-repeated idea that the committee doesn't look at conference record or conference position (Missouri State finished 1st in their conference, Marquette tied for 9th in theirs).
The next thing basketball fans look at is the team's RPI, or ratings percentage index. Why? Because whether they know anything about it, fans know it's what the committee looks at, in some way, shape, or form. The RPI considers a team's winning percentage, its opponents' winning percentage, and its opponents' opponents' winning percentage. This means it goes three "levels" deep in determining what constitutes a "good" team. In the most recent RPI, Missouri State is #42, Marquette #63.
If RPI is all you knew about a team, then #42 should give a team an 11 seed while #63 would keep a team out of the tournament. But as the committee watchers say, RPI isn't the end-all of what is considered. The RPI is used more, it is said, to evaluate a team's competition, in order to derive Strength of Schedule (SOS) and track the number of Top 100 RPI wins a team has. These two metrics become very important.
Obviously Missouri State fans know that a good RPI isn't enough. In 2006 the Bears had an RPI of #21—a 5-6 seed level—and weren't selected as an at-large. That same year the Missouri Valley had a record four teams in the tournament. Was that a factor? The committee says no. But was it in their minds, despite what they claim? If so, Marquette may encounter some resistance as the 11th Big East team, when the previous selection record for a conference is 8 teams. But we digress.
In order to encapsulate a team's season of wins and be able to compare those wins in significance among all teams, SportsRatings uses the RealWins algorithm. It basically takes the total number of Division I wins and redistributes them to teams that had more significant wins. It doesn't even bother with losses; those are unimportant to it. It just looks at who you beat, and who they beat, and who those teams beat.
Like the RPI, it goes three levels deep, since it's designed to reflect what the selection committee considers. A better algorithm would go "infinite" levels deep to refine the numbers maximally, but three levels is not surprisingly what matches up best with NCAA selections. Last year the RealWins algorithm lined up with nearly all at-large selections, missing just two teams; giving it leeway and using just the top 34 teams (the number of at-large selections available until this year) it is almost 100% accurate.
Missouri State is #27 in the current RealWins rankings, a level which should indicate an extremely high probability of selection. Marquette, on the other hand, is just #59, which should mean they are unlikely to be selected. Remember, this is an unbiased look at the aggregate significance of a team's collection of wins. But as we can see, humans hate composites; they want to see small numbers that have a definitive meaning, such as wins over top 50 teams, and strength of schedule ranking.
The Bears have no games remaining, and therefore their RealWins ranking can only go down. It's true that if they teams they beat win more games, their score can increase, but the majority of their opponents are no longer garnering additional wins. Teams currently behind them can pass them, and they might fall into the low 30s. Teams at that level are still almost always selected, though in 2009 a team at #26 was passed up; that team was Creighton, also in the Missouri Valley.
Conference biases aside, though, what would make the committee take Marquette over Missouri State? As mentioned before, Top 100 RPI wins and Strength of Schedule are what is killing the Bears, and what is saving Marquette. The teams both have 3 wins over teams in the 51-100 range, but Missouri State has no Top 50 wins while the Golden Eagles have 4. In SOS, Missouri State ranks just 123rd. Marquette is #34.
It's these numbers that everyone is looking at, and taking into heavier consideration than the team's RPI. Missouri State's numbers are poor, and Marquette's are at least reasonable for a selected team. So if Marquette is taken and Missouri State isn't, we can conclude a few things:
- Who you play is as important as who you beat.
Strength of schedule, in and of itself, is important, while win-loss record, in and of itself, is not. This doesn't make a lot of sense. The two have to be used in conjunction with each other. The RPI is supposed to provide this balance, but apparently no one even believes the notion that it does. - After weighting a team's wins, you should then really give weight to wins over top teams.
In focusing on wins over Top 25, Top 50, and Top 100 teams, the committee is making up for a huge deficiency of the RPI, namely that it doesn't know whether you won or lost an individual game. That's why these have to be factored in. More than anything, it seems that the bracketologists penalize teams without these big wins, and reward the bubble teams that have at least a few of them.
- The RPI doesn't do its job—yet is used in the corrections that make up for its shortcomings.
The RPI works best as a strength of schedule indicator. When a team plays a tough team, its RPI can increase whether they win or lose; likewise playing poor teams drags one's RPI down. In the end it gives a general balance between how well a team has done and how tough their schedule is.
But the RPI's conclusions aren't trusted. Teams with very good RPI rankings still might have very easy schedules. In those cases the RPI rating is essentially disregarded—because of SOS, which is measured by using the RPI rankings of opponents.
The RPI also doesn't give any indication of how well a team did against the good teams it faced, so again this is corrected—by using the RPI as a judge of how good those teams are. We look at top 25, top 50, and top 100 RPI opponents—all after judging the RPI to be an inadequate composite of a team's performance.
The irony is that if Marquette beat Missouri State, they'd rack up another Top 50 RPI win. But if the Bears won, they'd still have a zero there; when evaluating the teams, Marquette would win that measure 4-0.
The RPI has been attacked enough over the years that you'd think the NCAA would decide to use something better. It's not impossible to make a formula that would balance the things the committee wants to see more appropriately. One shouldn't have to use additional measures to re-evaluate teams because your formula has produced results you don't like for certain teams. And to use the same, flawed formula when deriving your new statistics?
Currently our Dance Chance algorithm has Missouri State an 11-seed. This is due to their overall record and conference record—while taking into account the conference in which they play—as well as their RPI and RealWins rating. We also give a bonus/penalty for SOS and a Top 100 wins composite. All of this hasn't been enough to put Marquette into our brackets; they're still in the first four out. In other words, we have things the opposite of what almost everyone else is saying.
If, and where, these two teams get seeded will tell us a lot about the priorities of the selection committee; we can already see that the bracket pundits are emphasizing SOS and top 100 wins over RPI and other overall performance indicators. It will be an interesting test for the balance of all these forces.
And of course, we can't discount bias. Missouri State has been a major recipient of ignoring a solid RPI number many times. In the 2006 case, the Bears' RealWins ranking was just #52. Compared to that year, Missouri State is actually in better shape if you add up their RPI and RealWins rankings; their total this season is 69, compared to 73 in the year they weren't selected. And that year there were only 34 at-large slots available compared to 37 this season, so there still may be hope for the Bears.
The bottom line, as always: if you aren't in a power conference, don't dare lose in your tournament. Lower-conference teams can never count on getting the benefit of the doubt, even from an allegedly "unbiased" system.
The move to the "Super Conference" is only going to make the RPI rankings even more hokey I think. You're definitely right though in that if you are a Mid Major, your best bet is to make sure you win your tournament!!
Posted by: 2012 NCAA Tournament | November 05, 2011 at 08:44 PM
As i was looking for more information related sports, luckily i made a visit to this post, which seems to be really excellent.
Posted by: Funny College Stories | December 05, 2011 at 06:48 AM