While we're in the down time before the tourney resumes, I thought Jay's piece on ESPN Insider with suggestions for improving the NCAA selection process was very interesting.
Aside from some things he and others have suggested before (e.g., more “basketball people” doing the selection/seeding process and replacing RPI as an “organizational tool” with Sagarin/Ken Pom data), he also suggests making a much bigger structural change to how the season now ends, namely to move "Selection Sunday" up by one week, so that it immediately follows the end of the regular season.
Jay explains his proposal, in part, as: ". . . the committee [would] select the 68 best teams based upon the regular season's body of work, without concern for automatic qualifiers. [This would] make conference tournaments about the automatic bids and conference pride only, rather than earning late brownie points.”
The mechanics would be: “select the best teams and rank them on the S-Curve, and then the automatic qualifiers [who win bids in conference tournaments] would knock out the lowest-ranked teams on the curve, one by one.”
I assume, under Jay's proposal, the seeding/bracketing would then still be done and announced at the same time it is now.
This proposal takes some of the drama out of the selection announcement and, obviously, substantially reduces the bubble speculation.
But, it has the great advantage of letting clubs like Va Tech and Colorado and Bama know in advance exactly what they have to do in their conference tournaments in order to get in (or else hope for very few conferences to be won by bid-stealers).
I think this is a very good proposal - of course, it will never happen as it would entail far too much transparency (and allow for lots of political pressure, lobbying and griping about who should be where on the S-curve).
Also, what about those last handful of teams fighting for the at large spots? Shouldn't their tournament performance be a consideration? Had two closely-rated teams, such as Va Tech and Clemson, met in the ACC tourney, shouldn't the outcome of that game be considered in the equation?
Seems to me that every issue that this proposal addresses adds more equally controversial ones. The only real difference is that we'd see Seth Greenberg whining a week earlier.
Bottom line in my view is that, with the number of at large teams in the tourney, there just isn't much sympathy for those three or four that feel that they were left out improperly. It's not like the committee has kept the steps a team can take to raise its prospects some big secret.
I do think the incorporation of kenpom or some of the other more sophisticated statistical models over the use of the RPI is an idea well past due.
I also think I would like it better if the committee worked this way, but they still announced it on the real Selection Sunday. Honestly, I think the most important thing that the committee could do, even if everything else is the same, is release the entire S-curve. That way, we would know what teams were given higher/lower seeds because of procedural bumps instead of perceived ineptitude. It would help the media ask better questions of the seeding and perhaps open doors to fewer generic answers from the committee chair when he appears on TV.
Sorry but I think this is a bit of a silly idea. There is great drama and speculation leading up to Selection Sunday and this would eliminate a lot of it. Additionally, it would greatly devalue the conference tournaments.
Take UConn this year for example-- how would the committee adjust for them winning the BigEast tourney by winning 5-games-in-5-days? Under Jay's scenario, they would have simply been one of the teams on the S-curve high enough to make it regardless of the outcome of their tournament and would not have had their seed adjusted at all to reflect the awesome quality of their play over that week. How does that make any sense?
What about a team that was in the mid-40s on the pre-conference S-curve that beat a couple really good teams but then failed to win a conference tourney? What about a team that was on the bubble that bombed out in their conference tourney? Is Jay saying that we should totally the 90% of conference tourney games except for the ones that determine an automatic bid? That just seems silly to me.
The job of the committee is to pick the 64 best teams in the land (allowing for automatic bids). The only way to do that is to make the selections at the latest possible moment. Otherwise, you are taken very recent, very relevant data and totally ignoring it. What sense does that make?
-Jason "I usually agree with Jay, but not this time" Evans
Don't ask me why, but my mother is making me Tweet. Says it will be good for my career. So, follow my ramblings, mostly on the film industry, @TVFilmTalk
Yeah, I like Bilas but hate this idea. There is zero motivation to do anything in the conference tournament for teams like Duke who know they're in. Which means you'd see a ton of upsets because the higher ranked teams just don't care. What would be the point of even having the conference tournaments?
And as an alternative, couldn't you just do the selection at the time it's done now but tell the selection committee to disregard performance in the conference tournaments? Why do it twice?
I don't like it either. As one of the articles posted on DBR lays out, in conferences with unbalanced schedules the regular season results may not be indicative of the true strength of a team. Look at the Big East, where the top programs in effect got to pad their schedules with the likes of Rutgers and Seton Hall (and so far have gone 0-fer in the NCAA tournament against schools from other conferences). Don't undermine the conference tournaments.
No soup for you!
I like part of the idea, which is to release the list of the top 68 teams as it exists after the regular season. However, the list can change due to conference results. In addition, the tournaments would still mean a lot because of seeding implications. Duke would still want to win the ACC because it is a championship and a number one seed is at stake. As far as this year goes, UConn probably got a #3 seed because they won the Big East tournament but would have been seeded lower if they had not won.
I like one (and only one) part of this idea: making the S-curve rankings deterministic.
This would let everyone -- coaches, analysts, fans, and ultimately the Selection Committee -- run their own what-if scenarios of equal validity. Everyone would still assign their own probability to the inputs (eg, "Clemson beats VPI by 5" or "Duke is ranked #3 in the final AP poll"), offering plenty of room for healthy debate and speculation. Nobody's fate would be predestined before the traditional start of the postseason.
But in a fundamental sense, there would be no more surprises. Coaches would know how scheduling a cupcake affects their distribution of likely tourney seedings. As the NCAAs approached, players would know what each win meant for their selection chances. Fans would know why their team was seeded a certain way, or left with a popped bubble. Since ranking systems are highly interdependent I won't claim these things can be known "exactly" -- but I think all of the potential audiences could be satisfied to well beyond the system's intrinsic margin of error / luck factor, given reasonable assumptions and a big enough Monte Carlo.
The only losers in this scenario are the Selection Committee, who'd lose the ability to make arbitrary choices. Their function would be relegated to assigning the pre-ranked teams into pods & regions, while enforcing any additional constraints the NCAA deems necessary (such as today's rule against early intra-conference matchups). Of course, any such loss of authority is likely to doom the proposal once NCAA politics come into play...
One could also argue that all deterministic systems can & will be gamed. To them I say
(a) designing a multi-factor, skew-resistant system is really not that hard
(b) the RPI is already far, far easier to game than any serious proposal I've seen in recent years