NCAA Lax Selections

[KeithK]

[quote jtwcornell91][quote KeithK]Any system that puts an undefeated team behind teams with losses is ill-conceived.  I don't care what the SoS is.  There's just no data to indicate that an undefeated team isn't the best team, only incomplete data saying that it might be.[/quote]

Tell that to Boise State. :-/[/quote]
But everyone knows the BCS system is ill-conceived, at least in terms of determining a true and worthy national champion.  It's probably pretty well set up to provide a near optimal revenue stream, however.

[Al DeFlorio]

[quote Hillel Hoffmann]With our luck, they'll change the system after Princeton and Syracuse return to form and when Colgate, Dartmouth and Brown are noob powerhouses, just in time for Cornell to be punished for our ridiculously strong fixed schedule.[/quote]
The only way you can be "punished" for a strong schedule is if you fall below .500, as Virginia did in 2004 and Syracuse this year.  Virginia responded by replacing Penn State and Air Force (both 2004 losses) with Mount St. Mary's and Manhattan in 2005, and then with Bellarmine and VMI in 2006 and 2007.  These schools, of course, have no effect on their strength of schedule, as only the top ten teams are considered, but they do assure two wins and the padding of individual stats.  Syracuse played a solid schedule top to bottom this year.  I hope they don't resort to scheduling Wagner and Lafayette as a curative.

[Hillel Hoffmann]

[quote Al DeFlorio][quote Hillel Hoffmann]With our luck, they'll change the system after Princeton and Syracuse return to form and when Colgate, Dartmouth and Brown are noob powerhouses, just in time for Cornell to be punished for our ridiculously strong fixed schedule.[/quote]
The only way you can be "punished" for a strong schedule is if you fall below .500, as Virginia did in 2004 and Syracuse this year.  Virginia responded by replacing Penn State and Air Force (both 2004 losses) with Mount St. Mary's and Manhattan in 2005, and then with Bellarmine and VMI in 2006 and 2007.  These schools, of course, have no effect on their strength of schedule, as only the top ten teams are considered, but they do assure two wins and the padding of individual stats.  Syracuse played a solid schedule top to bottom this year.  I hope they don't resort to scheduling Wagner and Lafayette as a curative.[/quote]
You know, until you posted that, I had not focused on how few creampuffs there are on Syracuse's schedule. You're right, it will be interesting to see whether Siena or Providence or some such team finds a way onto their schedule next year.

[nshapiro]

Does anybody have the ability to take the Lax season results and generate a pairwise ranking.  It would be interesting to see what this well known set of criteria would generate.

[Chris '03]

[quote nshapiro]Does anybody have the ability to take the Lax season results and generate a pairwise ranking.  It would be interesting to see what this well known set of criteria would generate.[/quote]

If you had a few hours, you could probably format the season's results to plug them into JTW's script. You'd have to do things like assign Hopkins the abbreviation for CC or something...

All I know is Cornell would be PWR 1.

[jtwcornell91]

[quote nshapiro]Does anybody have the ability to take the Lax season results and generate a pairwise ranking.  It would be interesting to see what this well known set of criteria would generate.[/quote]

I could probably do this, since I'm planning to do a Bayesian Bradley-Terry with a proper prior, and I already have methods to do most of the pairwise criteria.  (Although now that I think about it, the actual PWR was done in each script separately.)

[jtwcornell91]

I've known for years that the lax criteria were retarded.  I mean...

Criterion #1 counts how many wins you have against good teams, regardless of how many games it took you to get them.  So going 3-6 against the top ten would be better than going 2-0.

Criterion #2 compares strength of schedule, paying no attention to how many of those games you actually won.

Criterion #3 finally considers your won-loss record, as part of RPI.  Which is supposed to include strength-of-schedule.  Which was already its own criterion.

People say KRACH overweights strength-of-schedule because it actually uses the strength of your opponents to evaluate all of your wins and losses.  Well, the lax criteria actually do weight strength of schedule separately and more strongly than your actual game results.

[DeltaOne81]

[quote jtwcornell91]I've known for years that the lax criteria were retarded.  I mean...

Criterion #1 counts how many wins you have against good teams, regardless of how many games it took you to get them.  So going 3-6 against the top ten would be better than going 2-0.[/quote]

Actually, the actual wording of the criteria says "record" against top 10 teams, not just wins. However the committee has treated it as wins only for a long time. What is needed here is to emphasize record instead of wins.

[redhair34]

This article is worth a read.

"As a committee, we sit down and we're given a sheet that tells us the exact formula to follow," said Butler associate athletic director Jon Hind, who heads the men's selection committee. "People see the selections, and they are unaware of the criteria, and that starts up all the angst."

Of course, because the criteria are sacred.  It isn't possible that the criteria and the unpopular consequences of using the criteria are what upset people.

[Chris '03]

Can someone show me the SOS numbers that get Maryland the 2?

According to the numbers used by the committee, Duke had the toughest strength of schedule, followed by Maryland at No. 2, Virginia at No. 3, and Johns Hopkins at No. 4. Solid numbers by all accounts. Of the teams that made it into the tournament field, however, only Notre Dame and Providence had a lower SOS than Cornell. Among all DI teams, that SOS number dropped to No. 19 overall for the Big Red.

So only 20 teams under consideration? Where does that come from? I guess QWF only goes out 4 bands then...which pokes holes in some of the Princeton over Colgate arguments.

"They had a great year, and they were on the board until the bitter end," said Hind. "...If Ohio State wins there, which they did, Colgate has an RPI that is 16th out of the 20 that were being considered [for the tournament]. If Army wins, that jumps to 14. And how many teams make it?"

Which is it?

Secondly, the committee examines a candidate's strength of schedule (SOS) index based upon the 10 strongest teams on its schedule (again, according to those opponents' RPI ratings).

"We assess value to team's full schedule," said Hind.

This logic is ass backward. The t-10 metric only pads Hop, SU, and the ACC and lets them take on 5-6 nobodies to beat into the ground to ensure an NCAA ticket.

Hind explains that the 10 strongest teams provision was instituted several years ago to ensure that perennial powers are not punished for playing traditionally weaker teams, and so that stronger teams within a primarily weak conference are likewise not penalized.

So you can't count H2H unless teams are within a certain metric (like GT and MD)... but:

In Cornell's case this year, Hind said, those secondary criteria never really came into play, although the head-to-head result between the Big Red and Blue Devils was certainly addressed and discussed by the committee.

Long live PWR!

[Jeff Hopkins '82]

[quote Chris '03]
Long live PWR![/quote]

If it means anything (and it doesn't), a quick look suggests that if a hockey-type PWR was used, we'd win the the comparison with Duke

RPI  -  Duke
H2H  -  Cornell
TUC  -  Cornell

[Chris '03]

[quote Jeff Hopkins '82][quote Chris '03]
Long live PWR![/quote]

If it means anything (and it doesn't), a quick look suggests that if a hockey-type PWR was used, we'd win the the comparison with Duke

RPI  -  Duke
H2H  -  Cornell
TUC  -  Cornell[/quote]

Of course that presumes Cornell played the requisite number of TUC games, whatever that may be in a 13 game season... and however you define TUC for lax. The committee apparently considered 20 teams. I think there were about 32 teams with RPI over .500. COp is obviously a wash. We'd certainly win every other comparison.

[billhoward]

Maybe the 10-strongest-teams-played parameter ought to be reworked to still not penalize teams with poorer league opponents, but disallow credit for putting on the schedule teams with limited talents (besides sounding like a single-malt scotch, such as Bellarmine).

This is not a legitimate criterion, but every non-Ivy opponent Cornell booked was in or around the top ten (or twenty) at the time Cornell played them: Army, Notre Dame, Syracuse, Binghamton, Colgate. Our strength of schedule is hurt also by not being able to play the best Ivy team.

[Chris '03]

Anyone feel like reverse engineering ratios to see if it's 25/50/25 or something else?

http://web1.ncaa.org/app_data/weeklyrpi/2007MLArpi1.html?

[DeltaOne81]

[quote Chris '03]Anyone feel like reverse engineering ratios to see if it's 25/50/25 or something else?

http://web1.ncaa.org/app_data/weeklyrpi/2007MLArpi1.html[/quote]

It is. It exactly matches this:
http://www.vaporia.com/sports/collegelacrosserpi.html
Which uses the rules defined here, including 25/50/25:
http://www.vaporia.com/sports/rpi.html

(this is the 2nd week in a row that the NCAA exactly matched Wobus - which is 2 for 2 since that's how long I've been counting)

It is very close to laxpower which now uses 25/50/25 as well:
http://www.laxpower.com/update07/binmen/rpi01.php
Some small detail flips Princeton/Maryland, and some others.

Hymie is close as well, but has some known differences in his versions of the calculations:
http://lacrosse.homelinux.net/rpi