(Too much) time on my hands....

Cornell sits at 12-3 (.800). Realistically, I have to think that we’re going to lose a couple more to finish at 24-5 (.828). I’ll take it. This contributes .25*.828 = .207 to our end of season RPI (+.007 from its current contribution of .200).

Right now, the Cornell’s ECAC opponents are 31-54-10 in non-conference games (.379). There are still 10 non-conference games left, so the record could go as high as 41-54-10 (including Dartmouth over UNH and a Beanpot championship for Harvard) or as bad as 31-64-10 (which would include Iona AND Holy Cross over Brown. Please, dear God, no….). Looking at the matchups, I think the ECAC goes 6-4 in the remaining non-conference games to end up at 37-58-10 (.400). And that includes a few non-conference wins against themselves - bleah.

Fortunately for our strength of schedule, this is offset by all those wonderful league games, where our opponents are guaranteed to average .500. Between our 11 conference brethren, there are 110 games, each of which is won and lost, or tied (twice). Therefore, it matters not what the actual records end up, the final record of our conference opponents will end up at the equivalent of 147-168-10 (.4677). Yea conference games!

The current composite record of our 7 non-conference opponents is 92-48-12 (.645), taking into account that we played OSU, BU, and WMU twice and ME only once. Let’s assume that they continue on at this pace (I think ME and BU will do a bit worse than this with their remaining schedules, but I think OSU and WMU stand a shot at doing a shade better and we played the latter pair 4 times to the former’s 3) and finish their regular seasons at this (weighted) average win percentage.

Therefore, our final opponents’ winning percentage will be (22*.4677 + 7 *.645)/29 = .510, so the final opponents’ record contribution to our end of season RPI will be .5 * .510 = .255. Our current opponents’ winning percentage is .555 (contribution of .278), so our RPI will change by -.023.

I don’t feel like doing anything with our opponents’ opponents’ records because it’s too much work, so let’s assume that it stays about the same at it’s current value of (though with all the upcoming ECAC action, it could actually drop a little bit). Therefore, our final RPI would be our current value of .616 +.007 - .023 = .600. This would put us in 4th or 5th place currently (depending on my roundoff errors and ignoring opp opp pct) even with a couple of losses. Winning a few playoff games would help, too, especially if we end up beating someone in the 7-9 range for the quarters (as opposed to 10-12). OTOH, if we run the table (right…), our final RPI would be .617, almost exactly where it is now.

Of the teams that currently have and RPI better than .600, we will almost certainly lose the PWR comparisons with Maine and CC (unless they totally tank). UND and SCSU are another story. I think UND is going to lose enough games to drop their RPI below .600. Don’t forget that the in-league-play effect that will *raise* the ECAC win percentage will be *dropping* the average WCHA win percentage (since they're currently above .500). So we may keep the comparison with UND. Even if SCSU sneaks ahead of us in RPI (and they certainly have some losses coming), we should be able to keep the comparison based on TUC and RPI. Our only common opponent with SCSU is RPI (the school), who split with SCSU (yea Engineers!). If we can take 3 points from RPI (the school), we should win the PWR comparison regardless of RPI (the ranking), based on TUC and COp. Therefore, in the long run, I’m now optimistic that our imminent drop in RPI will be small enough that it may not cost us *any* PWR comparisons and we’ll finish 3rd in PWR, with comparison losses only to Maine and CC. Whew.

Of course, if our boys don't play to their potential then we'll lose plenty of PWR comparisons and we'll deserve what we get in (or out of) the NCAAs...

Terrific post, Robb!

I know that must have taken a lot of time because I was trying to do something similar, figuring out how Cornell's RPI might change based on its five upcoming games. Once I got to Opponents' Opponents' Winning Percentage, I realized that it was all beyond me. (I DID major in ILR, after all... )

Anyway, thanks for all the effort...I hope it all comes true...Cornell finishing #3 in the PWR and getting a Number One seed would be terrific!

LET'S GO RED!

David Klesh wrote:

I know that must have taken a lot of time because I was trying to do something similar, figuring out how Cornell's RPI might change based on its five upcoming games. Once I got to Opponents' Opponents' Winning Percentage, I realized that it was all beyond me.

And yet everyone says that KRACH could never replace RPI because RPI is so simple and KRACH is so complicated.

OTOH, KRACH should never replace RPI because RPI tends to put Cornell higher. That's solid reasoning to me!!!

Greg wrote:

OTOH, KRACH should never replace RPI because RPI tends to put Cornell higher. That's solid reasoning to me!!!

Comments on your argument aside, that may not be true with the new weighting; remember, this is the old RPI which has been known to penalize teams for playing weak opponents even if they beat them (see BGSU/OSU in 1995); don't draw conclusions based on experience with the RPI of the past five years or so, which rewarded MAAC teams for monster winning percentages, regardless of their strength of schedule.

To this observer, PWR has gone too far in giving weight to strength of schedule. Having all nine Hockey East teams as TUCs along with a 7-12-3 Wisconsin team that's 1-8-3 in its league is a little bit much.

FWIW, Wisco also has an RRWP above .500 (KRACH>100), so they'd be a TUC according to the not-yet-calculated nouveau KPWR. The other differences between the two are that Huntsville (#32 in RPI with a .5006 and #35 in RRWP with a .4575) would not be a TUC, while Brown (#34/.4940/#30/.5443) and NMU (#35/.4927/#27/159.4/.5773) would be.

At least 1-12-7 Anchorage is no longer a TUC (in either system apparently). At like 1-10-7 they were just a tick before 14-4-1 QU... which just doesn't seem right to me.

I agree with the sentiment that this current system seems to underweigh win % and overemphasis S.O.S. - even if the old one did the opposite.

Perhaps someone who studies this stuff more carefully than I do will comment, but it also bothers me when I see people write: Well, it's better for Harvard that they lost to Princeton rather than a TUC, because it will hurt their PWR less. If this is true, its just nuts. A team should never be penalized more for losing to a top team than for losing to a slug.

I'm still uncomfortable with the whole TUC business. If a team is 3-0 against a team with a .500000000001 RPI and 0-3 against another team that's .4999999999999, why should it get great marks for the former and yet the latter counts not at all in the TUC criterion?

Well, obviously, in that scenario, it would be better had they won those 3 games, because their win % would be up, so they're RPI would be up, so it would count for *something*, but what you're saying is what would the difference be had those 3-0,0-3 records been reversed.

RPI: minimal - same win %, opp's win % changes but in an unpredictable way, and probably not significantly especially after a 29/34 game season.

Record against TUCs: Yup, the three games against the 0.4999... teams count for nothing, so in this category you'd rather beat the good teams than the poor ones.

Record against COp: Obviously this changes, but again unpredictably.

Head to Head: Your record only matters in the direct comparisons, so against you'd rather beat the good teams and lose to the bad ones.

So, yup, this works out.

People have talked about ways that repair for the strict cut-off a bit, ie. counting all games in a sort of "record against TUCs" (more like record weighted for S.O.S.) but weighting the big wins heavier by RPI or such. Similar idea for Record against COp.

Of course, there's probably 'paradoxes' in those situations too, and the fact is change in statistical models are slow in coming, but at least they're talking about it in the summers.

One of the biggest problems with KRACH is its name. It's a really, really unfortunate acronym.

I am all for rewarding sheer winning percentage and not caring about schedule strength. To a small extent this is because you can't beat a team that you don't play, and WCHA teams don't come east often enough to justify going out there to serve as sacrificial lambs in their tournaments.

But to a larger extent it's pure self-interest -- the ECAC tends to roll up bad winning percentages in NC games, so I'm for depressing the influence of those percentages. In the long run the thus "artificially" (whatever that means) inflated ratings of ECAC teams will lead to slightly better tourny placement --> better results --> more attractive recruiting --> better ECAC hockey.

I realize there is a contrary argument that we should highlight the ECAC NC record, get killed in tourny selection, and thus force the ECAC to do something about the structural handicaps of scholarship bans and short schedules. It's a matter of taste, but I think this is the same terrible logic as Naderites who suggested throwing the election to the Bushies, letting them destroy the country, and thus inspire the electorate to wake up, have a moment of enlightenment and vote for more radical change.

Um, yeah, that'll work.

I'm all for trying to address the underlying structural issues, but taking concrete steps backwards deliberately in order to leap forward in some fantasy future just doesn't work.

[Q]I am all for rewarding sheer winning percentage and not caring about schedule strength.[/Q]
There'd be 100 times more bitching about that than the current system... especially when Quinnipiac beats out us and Maine for the Eastern #1 seed, and when Wayne State knocks Minny out of the NCAA tourney... god, that's a terrific way to piss off the entire college hockey world .

In the current systems there are a few things that don't seem right. In that situation, there'd be a ton. What you're essentially saying is "screw even trying to figure out who's better than whom" and that just not how college sports work. Self-interest is no excuse for blatant unfairness... well, except in Bush's tax cuts...

[Q]I'm all for trying to address the underlying structural issues, but taking concrete steps backwards deliberately in order to leap forward in some fantasy future just doesn't work.[/Q]
Agreed, but what concrete steps backwards are you talking about? I don't think anyone's proposing any such step.

The point is "fair" is in the eye of the beholder. Why isn't "fair" just having a six team tournament field with the six conference champions? What's so fair about watching Michigan and Michigan State beat up their opponents year after year after year after year and get to play NCAA tourny games on home ice? Just because it's the status quo doesn't make it reasonable.

The key to compromise is to put your toughest position forward first and then fight for every inch. That way maybe you'll get a little of what you wanted. The problem with the "fairminded" approach is you put an even proposal on the table first, the moron across from you counters with a ludicrous proposal, and then you compromise your way to a solution that is excessively against your interests.

No thanks. Tell the NCAA we think the rating system should be winning percentage times average SAT score.

That might be a decent description of negotiations, Greg, but we aren't actually deciding on the policy. We are only discussing out loud what a reasonable policy would be. For this discussion you can take your ECAC hat off; you are among friends.

My response exactly... in what way is this a negotiation?