[Lax] PWR for lacrosse?

Hey JTW-

How hard would it be to run your PWR scripts on this season's lacrosse results to date? Obviously, L16 would have to be adjusted to L8 or L10 given the shorter season.

They're designed to be pretty generic (in fact, the '16' in last 16 is actually a function argument). What I'd need would be a list of teams with corresponding abbreviations, and a text file containing all the results. Do any of the poor substitutes for USCHO have an easily-parsed score database?

Not that I could find. I thought about switching over the stats at the top, but I couldn't find anything that would work or had even been updated for yesterday's game yet.

The lacrosse world still operates on suth'n time.

Okay, thanks to Matt Carberry, I got hold of composite scores and was able to kludge together my full suite of hockey ratings for the lacrosse season at
[slack.net]
It probably still says "hockey" in a lot of places, and I can't vouch for the KPWR (since my scripts assume all the teams are KRACH-comparable to each other, which holds halfway through a 30-game hockey season but not a 12-game lacrosse season), but everything else is probably right.

Note that according to Bradley-Terry, Cornell is #2 in the nation, since our only loss was to undefeated Georgetown. Syracuse, Hopkins, and Virginia are dead-even at #3-5, since they're each 1-1 within that group of three, and thanks to Syracuse's loss to us, they're all considered infinitely worse than us and hence also Georgetown. Loyola's lone loss to Syracuse puts them #6, and then most of the rest of D1 is in one big group. (But look at the range of KRACH values, from less than .15 to over 100,000!) Such funny structures are common when only a few games have been played.

In terms of the hockey criteria, Cornell is #4 in RPI and #2 in PWR. Syracuse is #5 in RPI and #6 in PWR. (I used John's suggestion of "last 8"; actually "last 5" might be more appropriate.)

JTW --

1.) Just out of curiosity, I'd like to know where the composite scores came from (one of the usual lax websites?). (I also thank Matt; and thank you, John, for all the... uhh... "kludging"...) )

2.) I don't know who Bradley and Terry are, but if they've got Cornell at #2, then they're OK with me!

3.) So, if "RPI" is supposed to be "the most important of the NCAA's selection criteria", then Cornell's #4 should put 'em on course for a tourney bye!

4.) As for converting hockey's "last 16" to lacrosse: If my assumption is correct that "last 16" is based on one-half of a typical 32-game hockey season, then the typical 12-game lax season would suggest a "last 6". (But, then, I'm no eckspurt on all these ratings systems, so...)

I can answer #1, since I gave the scores to JTW:
[www.laxpower.com]

(which still says SU scored 12 against us)

I removed the games played against non-D1 opponents from the data I sent to John (I believe there were only six).

Two things I can think of right now. One is the RPI distribution. The championships handbook (see link below) makes no mention of the formula to be used, so it's possible that it could be the so-called "general purpose" 25/50/25.

Also, hockey has a "must play 20 games against D-1 opponents" requirement to be a TUC - lacrosse has a similar requirement (10 games in this case). This generally does not affect teams in the top half or so, but could affect some at the bottom (like VMI and BC, who each have only 8 games against D1 opponents). The way the NCAA handled this in hockey (with respect to Air Force two years ago) was to say that the games would be considered in the selection criteria, but that a team had to meet the minimum or it would not be a TUC. This is irrelevant with respect to the two teams I mentioned above, since neither of them can finish above .500.

The lacrosse championship handbook is at [www.ncaa.org] . The selection process for Division I is detailed on page 12.

Does the NCAA even use RPI for Lacrosse? If they do, it's almost certainly the standard 25/50/25 weighting, with a bunch of unspecified NCAA fudge factors folded in. As far as I know, hockey is the only sport to use the modified weighting, and the only one to use the RPI undoctored.

Also, the Air Force 19-game solution (which turned out to be irrelevant since AFA ended up with 20 D1 games once the CHA playoffs were over) was actually an unusual solution to the problem. Back in the old days when a bunch of teams like Canisius and UConn were division I but played largely non-D1 schedules, games involving those teams were left out of the calculations entirely. (This meant that then-independents Army and Air Force which played a number of games against such teams had less than 20 games which counted towards the selection criteria.)

Thanks John!

I think

1Georgetown (EC)
2Cornell (IL)
3Virginia (AC)
4Johns Hopkins (Indy)
5UMass (EC)
6Syracuse (Indy)
7Loyola (CA)
8No. Carolina (AC)
9Maryland (AC)
10Hofstra (CA)

is far more reasonable than the laxpower modified BCS, which has:

1 Syracuse
2 Virginia
3 Georgetown
4 Johns Hopkins
5 Maryland
6 Loyola
7 Cornell
8 Hofstra
9 Princeton
10 Massachusetts

What do you say to the Bradley-Terry RRWP version:

 1. Georgetown     1.0000 
 2. Cornell         .9815  
 3. Johns Hopkins   .9444  
    Syracuse        .9444  
    Virginia        .9444  
 6. Loyola          .9074 
 7. No. Carolina    .8456 
 8. Princeton       .8441 
 9. Maryland        .8434 
10. Duke            .8327

Princeton gets a big boost from the fact that three of their four losses are against teams which are "infinitely better" than them. (Part of the drawback of being absolutist in a short season with no ties.)

From the handbook Matt referenced. Look at the last two paragraphs, which seems to provide a massive amount of wiggle room to the committee.

Selection Criteria. The Division I Men’s Lacrosse Committee will
select and seed teams based on the criteria listed in Bylaw 31.3.3 (wonlost
record, strength of schedule and eligibility and availability of student-
athletes.) When evaluating teams’ won-lost records and
strength-of-schedule, the committee has received approval from the
Division I Championships/Competition Cabinet to consider the following
primary factors as determined by the Rating Percentage Index
(RPI) in selecting at-large teams for the field (in priority order):
( 1 ) Results against teams in descending order, as determined by
the column entitled normal RPI rank that is used during selections
(e.g., vs. teams 1-5; vs. teams 6-10; vs. teams 11-15,
e t c . ) .
( 2 ) Strength-of-Schedule Index, based on the opponents’ success
ranking in the RPI that is used during selections.
( 3 ) The column entitled normal RPI rank, based on the entire
Division I schedule. The RPI includes the Division I winning
percentage, opponents’ success, and the opponents’ strengtho
f - s c h e d u l e .
If an analysis of the primary criteria does not result in a decision by
the committee, the following secondary criteria will be evaluated (not
in priority order):
• Head-to-head competition.
• Results against common opponents.
In order to be considered for at-large selection, teams must have a
.500 or better record against all opponents. Division I teams must play
at least 10 games against Division I opponents.
Strength of Schedule Index/Power Rating. In addition to the criteria
listed earlier, the Men’s Lacrosse Committee has developed a
strength-of-schedule index and a power-rating system that will be
used only as an aid during the Division I selections. Each regional
advisory committee member shall submit a ranking of all teams in his
division and all their opponents on a one-through-five basis, with a
“one” being the highest rating. Opponents’ total rating points shall be
divided by a team’s total number of games to arrive at a strength-ofschedule
index. Power ratings will be developed using the same onethrough-
five ratings. Both the strength-of-schedule index and the
power-rating systems shall be used as selection guidelines only, and
each divisional committee shall reserve the right to deviate from these
figures, based on the criteria listed and in the championships’ best
interest. The committee will consider only the 10 highest ratings on
each team’s schedule.

If you want lacrosse computer rankings, results, league standings, etc, visit laxpower.com

They have not yet updated for yesterday's win vs Dartmouth (as of right now), but prior to that game, the computer ranking had Cornell #9.

Note: this also has high school rankings, pretty comprehensive.

Petunia-

Thank you for you input. We are actually aware of the laxpower power rating and consider it amazing flawed.

Here are my thoughts as I posted them over at laxpower.

[forums.laxpower.com]

That having been said, laxpower is still a great place for scores and discussion about lacrosse.

Did you guys used to read collegelacrosseusa.com which now seems to be defunct?

Petunia: I did. CLUSA was the USCHO of lacrosse... then it changed into the pathetic mess known as 360Lacrosse.com. There are much better sites now (LaxPower, Lax.com, etc.), and their links are available on the thread titled "[Lax] LINKS!". Happy hunting! (Unless, of course, you already knew about all those sites and were simply wondering about what happened to CLUSA... OK... I'll shut up now...)

I updated the lacrosse PWR tables (thanks to Matt for the score conversion script!). Hopkins and UVa pass us by virtue of improved records vs TUC and by Colgate dropping below .500.

[slack.net]

[EDIT: I was missing two ACC tourney games when I ran the numbers for this message. See revised ranking in my reply to this post]

With Georgetown's loss to UMass, everyone's Bradley-Terry rating can be expressed on the same scale except for the eight teams at the bottom, who can be considered to have a KRACH of 0 and ignored. This also means that the three-way tie among Syracuse, Hopkins and UVa is resolved, since their other games enter the mix. Cornell drops from #2 to #4 in the Bradley-Terry [slack.net] (and also the KPWR) with yesterday's events:

1. Johns Hopkins
2. Virginia
3. Syracuse
4. Cornell
5. Georgetown
6. Princeton
7. Maryland
8. North Carolina
9. UMass
10. Loyola

Cornell is 6th in the RPI and PWR; all the hockey rankings for lacrosse are summarized at [slack.net]

It would be hard to argue with the top six teams in John's ranking above, IMHO.

Correction:

I was missing the two ACC tournament games; Duke's win over Maryland and UVa's win over UNC change the BT top ten as follows:

1. Virginia
2. Johns Hopkins
3. Syracuse
4. Georgetown
5. Cornell
6. Princeton
7. Loyola
8. UMass
9. Yale
10. North Carolina

Updated ratings are at [slack.net]

But the same top six, in different order.

How Maryland continues to be ranked in the top six in polls and in the LaxPower computer rankings is a great mystery to me. Hofstra seems to be overranked, too, IMHO.

Since the lack of context seemed to be confusing people outside the eLF who'd come across the link (the preview of the Brown game quoted my RPI without noting (or perhaps realizing) that it used a different weighting than the one the NCAA presumably uses for lacrosse), I clarified some of the patter at [slack.net] and also distinguished "Hockey RPI" from "NCAA RPI".

The good news is that even after the Brown game, we're still #6 in the RPI, PWR and KRACH. The bad news is that in the non-hockey-weighted RPI, we're #7. The good news is that this is because of a flip-flop with Princeton, who will hopefully get the Ivy AQ anyway. I'll let those more in the know analyze this RPI along with the more nebulous criteria described in the lacrosse championships handbook.

[Q]Age Manocchia '98 Wrote:

Not that I could find. I thought about switching over the stats at the top, but I couldn't find anything that would work or had even been updated for yesterday's game yet.

[/q]