What if we let a computer select the NCAA Tournament field?

[jocoshock]

What if we let a computer select the NCAA Tournament field?

http://mweb.cbssports.com/ncaab/eye-on-college-basketball/24511062/what-if-we-let-a-computer-select-the-ncaa-tournament-field

Based on RPI

The fourth No. 1 seed would have gone to Kansas instead of Virginia. Wichita State is the 4th overall seed, but gets a much easier bracket.

Final Four participant Kentucky is a No. 5 seed, which is may be a bit high, but more accurate than the committee's No. 8.

There`s more...

[ShockTalk]

http://mweb.cbssports.com/ncaab/eye-on-college-basketball/24511062/what-if-we-let-a-computer-select-the-ncaa-tournament-field

[Jamar Howard 4 President]

Please, please, please don't let them ever use the RPI (in its current form) for direct seeding. Terrible, terrible, TURRIBLE idea.

Good for grouping teams (top 50, 51-100, 100-200, etc.)

Bad for choosing an exact order.

[Cdizzle]

Please, please, please don't let them ever use the RPI (in its current form) for direct seeding. Terrible, terrible, TURRIBLE idea.

Good for grouping teams (top 50, 51-100, 100-200, etc.)

Bad for choosing an exact order.

Being good for grouping but bad for choosing an exact order doesn't work out when you draw exact brackets around a group. Team 50 and 51 are ordered exactly and grouped exactly.

[Jamar Howard 4 President]

True cdizzle, but the effect of a team slipping a group (into the top 50 when they should be #51 or vice versa) is minimal when looking at a resume with 34ish games on it.

If you are 5-4 vs the top 50 and 7-1 vs 51-100
or
if you are 6-4 vs the top 50 and 6-1 vs 51-100

not much changes. Sure it would be nice to be even more exact, and I'm all for using better computer models (like KenPom), but my point is that we are best off using computers to group teams in general and then letting humans do the final ranking.

One last small point. Let's focus on WSU. Say one of their opponents slipped into the top 50 of the RPI even though we believe them to really be the 51st best team. Maybe the 49th best team slipped out of the top 50 in the RPI. It can go both ways and generally will even itself out. Sure there will be outliers, but no system is perfect.

We can debate the initial value of using the RPI at all for sorting teams into groups. I'm fine with that. What I can't understand is why anyone would argue that the RPI could actually seed teams better than the committee currently does. I can find countless examples where the RPI would have been horribly unfair to teams.

[RoyalShock]

I like using the selection committee to determine what teams get into the field. I wouldn't mind then using a predictive system like Pomeroy to decide seeding.

[Jamar Howard 4 President]

I like using the selection committee to determine what teams get into the field. I wouldn't mind then using a predictive system like Pomeroy to decide seeding.

Just a few of the changes we would have seen if Pomeroy was used directly for seeding this year:

WSU falls from a 1 to a 2 (shockernet goes into full meltdown mode and pleads for the selection committee to come back into power)
Michigan falls from a 2 to a 5
Tennessee jumps from an 11 to a 3 (shockernet refuses to complain about a 12 loss team as a 3 seed because WSU beat Tennessee, and any argument against Tennessee would be equal to hatred toward WSU)
Pittsburgh (AKA garbage) jumps from a 9 to a 5 (ShockerFever spontaneously combusts in anger)
Iowa State falls from a 3 to a 6

[MadaboutWu]

Weren't we 4 in Pomeroy at the time of seeding and fell during the tournament? TN and Michigan are pretty striking--same question about the former: did you take the data before they went 3-1 against good competition?

I agree that any straight computer model shouldn't be the sole determiner of seeds, but I might like some more objective metrics given more weight.

[AndShock]

If Pomeroy was given any weight in seeding, Wisconsin and Pitt would get a 3 seedline boost every year.

[Jamar Howard 4 President]

Weren't we 4 in Pomeroy at the time of seeding and fell during the tournament? TN and Michigan are pretty striking--same question about the former: did you take the data before they went 3-1 against good competition?

I agree that any straight computer model shouldn't be the sole determiner of seeds, but I might like some more objective metrics given more weight.

Data was from Pomeroy's rankings on selection sunday. WSU has stayed steady at #5 overall. I've got a screenshot for reference.

If you go by today's rankings, Tennessee would be a 2 seed (#7 overall)
Michican making the Elite 8 still only bumps them up to a 3 seed (#10 overall)

[WSUwatcher]

I like using the selection committee to determine what teams get into the field. I wouldn't mind then using a predictive system like Pomeroy to decide seeding.

They've used a combination of human evaluation and computer ratings to decide the participants in the football championship for years now. Obviously, there have been those who complained (mainly teams who finished third and therefore just missed the big game) -- but it certainly isn't obvious that there's ever been some gross miscarriage of justice that could have been avoided by relying solely on a selection committee. I think a combination of human evaluation and computer rankings for basketball would almost certainly improve things, not make them worse.

[RoyalShock]

Just a few of the changes we would have seen if Pomeroy was used directly for seeding this year:

WSU falls from a 1 to a 2 (shockernet goes into full meltdown mode and pleads for the selection committee to come back into power)
Michigan falls from a 2 to a 5
Tennessee jumps from an 11 to a 3 (shockernet refuses to complain about a 12 loss team as a 3 seed because WSU beat Tennessee, and any argument against Tennessee would be equal to hatred toward WSU)
Pittsburgh (AKA garbage) jumps from a 9 to a 5 (ShockerFever spontaneously combusts in anger)
Iowa State falls from a 3 to a 6

It would need to be explained, and people would have to understand, that your season-long resume is what gets you into the tournament. But once there, a team's performance within those games, as evaluated by whatever system is adopted, defines seeding. I see no problem with WSU getting a 2-seed in that type of system (using KenPom), especially considering UK would have been a 5-seed (far more in line with how they were playing than being an 8-seed) and our 2nd game would have been against UNewMex or Providence.

The argument is always going to be in which predictive system to use. Each one has its merits and its faults.

[ShockTalk]

This isn't about using computers for seeding, but what about only having "at-large" teams competing in the "Round 1" games?

My reasoning is that every conference's automatic team got that bid by the rules. Either winning their conference tournament or, if no tournament, by winning the conference title. They should not have to do a play-in game to make the field of 64. Only those teams that didn't get an automatic bid, at large teams, should be included in Round 1.

Another reason for this is making Round 1 more relevant by only having bigger conference teams involved, probably drawing more TV and fan interest in that round. I wouldn't even mind if they expanded the number of teams to 10 or 12 ending most of the controversy over a "deserving" team not making the field. You now could have up to 6 teams in the field that would have been left out under the old 64 cut off line. Between those 12 teams you will have covered any team that would have had a reasonable argument that they should be in had there only been 64 teams. There would be little doubt of the 71st team that didn't make this field actually having a case that their resume was good enough to make the old 64 team cut off. Now you would have 6 games where teams can earn their way in since they couldn't get it done otherwise.

[Ricardo del Rio]

Paging Nate Silver. Paging Nate Silver. The NCAA is on line 2.

[Ricardo del Rio]

What if we let a computer select the NCAA Tournament field?

http://mweb.cbssports.com/ncaab/eye-on-college-basketball/24511062/what-if-we-let-a-computer-select-the-ncaa-tournament-field

Based on RPI

The fourth No. 1 seed would have gone to Kansas instead of Virginia. Wichita State is the 4th overall seed, but gets a much easier bracket.

Final Four participant Kentucky is a No. 5 seed, which is may be a bit high, but more accurate than the committee's No. 8.

There`s more...

Pregnant idea.