Pairwise Ranking

OK.
I noticed on uscho's pairwise site that if you add in the "bonus" factor, Cornell moves higher. The "bonus" helps Cornell, I guess because of games vs Notre Dame & Ohio State. It's kind of ironic I think, that this quality win bonus factor helps an ECAC team.

ANYway...

Didn't someone figure out the likely approximate bonus factor last year after the tournament teams were announced, and if anyone remembers this, can they advise?

Thx.

Also, did anyone notice that MAAC regular season games last weekend had higher attendance than ECAC preliminary round playoff games?
Holy Cross had crowds over 2000, that's kind of ecac-ish, hmmmm.....

I don't remember what the conclusions were about the bonus points last year. I expect peopel did make guesses. Keep in mind that there's no guarantee that they will keep the same numbers this year though. Since they refuse to tell us what the numbers are I can't help but imagine that we're back in the smoke filled room with the committee adjusting the numbers to give them the outcome that they want to see.

In fairness, it's probably just as likely that the idiots in charge just want to "add some mystery" to the process so that people will bother to watch the selection show. Ah yes, mystery over transparency. Brilliant.

Here is the text re: the adjustments to pure PWR on the USCHO site [www.uscho.com] : The current selection criteria is now affected by two additional RPI tweaks. ... 1) The RPI "bonus". The bonus, added in 2002-03, is awarded for "quality wins," defined as nonconference wins against a team in the Top 15 of RPI. More bonus points are awarded for neutral-site wins than home wins, and even more for road wins. Since the amount of bonus points is kept secret, we can't calculate a definite RPI. However, USCHO allows users, through the form on the right, to enter their own guesses on how much the bonus is worth, and based on that, see the effect on RPI and PWR. ... 2) If a team defeats a weak opponent, it is possible for its RPI to go down despite the win. Starting in 2003-04, if this happens in a team's conference tournament, that game can be thrown out of the RPI calculation for that team. Any team where this has occured has been noted below.

No good reason has ever been given for keeping the RPI bonus a secret, which is suspicious. Given the poor understanding that committee members have shown of their own selection criteria, it's likely (to the cynical among us, which I'm afraid includes me) that they include a fudge factor to cover their asses in the event that they make a dumb mistake. I doubt it's kept open to favor a premeditated outcome. That implies way too much planning and logic.

It seems like the whole idea of the RPI Bonus is a fudge factor to make sure that teams like Michigan and Minnesota make the tournament. Sure, they'll likely make it into the tournament on their own accolades (or, y'know, by virtue of just being in the WCHA, in Minnesota's case), but heaven help us if teams like Colgate make the tournament and lead to less revenue dollars. (No offense to the real Colgate fans, but let's face it, Colgate doesn't have the biggest student fan following in the college hockey world--free, projectile pizza is needed to get the students into Starr, after all.) For instance, if the season ended today, it would be no surprise to me if Colorado College and St. Cloud got in and Notre Dame and Michigan State didn't, especially since CC is the Western Regional host this year and thus should attract a good amount of fans there locally. One wonders if Cornell could benefit from this sort of policy, though--I think it's obvious that Cornell fans made an impact in the last two years' tournaments.

The bonus is not a fudge factor. That much is sure. Though, their secrecy opens them up to this sort of speculation - so, in that case, they deserve it.

But, in regards to the comment re: St. Cloud/Colorado College ... neither team qualifies for any bonus points this year (just go to USCHO's bonus pages and put in any number, and you'll see). ... So the Committee could do no amount of fudging to help those guys. After all, no matter what you multiply by 0, it's still 0.

If you look at this [www.uscho.com] page on USCHO, it shows the number of quality wins based on the current top 15 teams.

[Q]chrisatpc Wrote:

If you look at this [www.uscho.com] page on USCHO, it shows the number of quality wins based on the current top 15 teams. [/Q]

I just noticed this on that page:
[q] If a team defeats a weak opponent, it is possible for its RPI to go down despite the win. Starting in 2003-04, if this happens in a team's conference tournament, that game can be thrown out of the RPI calculation for that team. Any team where this has occured has been noted with an asterisk beside their RPI. [/q]

It's only helped RPI's RPI so far.

To be accurate, it didn't really help their RPI, it just didn't hurt their RPI, which is fair. Now if they can do something when BU's RPI goes up for losing to BC this weekend, that would help too. But .500 RPI and .500 winning percentage for TUC would help currently bouncing three teams from TUC , all in Hockey East (Mass.-Lowell, BU and Northeastern). That Northeastern can be a TUC but be the only team ineligible for the conference playoffs is just wrong.

[Q]Tom Pasniewski 98 Wrote:

To be accurate, it didn't really help their RPI, it just didn't hurt their RPI, which is fair. Now if they can do something when BU's RPI goes up for losing to BC this weekend, that would help too. But .500 RPI and .500 winning percentage for TUC would help currently bouncing three teams from TUC , all in Hockey East (Mass.-Lowell, BU and Northeastern). That Northeastern can be a TUC but be the only team ineligible for the conference playoffs is just wrong. [/Q]
I'm still not convinced that losing a game should ever raise your RPI. Yes, it should go down less for losing to BC than to Bentley. (In the TUC criterion, however, you're better off losing to Bentley. That's a real puzzler.) But having your ranking improve because you lost? Huh?

In a sensible system your ranking would always go down when you lose and always go up when you win, though the amounts are ay vary significantly depending on competition. But everyone who's paying attention knows that RPI is not a sensible system...

Hey, do you think we could get more bonus points for beating the only teams who could give us bonus points? That is, we didn't lose to any bonus point teams or putting it another way, we didn't have any quality losses that we didn't also beat.

[Q]Jim Hyla Wrote:

Hey, do you think we could get more bonus points for beating the only teams who could give us bonus points? That is, we didn't lose to any bonus point teams or putting it another way, we didn't have any quality losses that we didn't also beat.



Edited 1 times. Last edit at 03/10/2004 6:23pm. [/Q]
We get more bonus points than UND!

It appears that according to this week's bracketology, Cornell is the last team to not make the NCAA tournament. And for those following bracketology since it started about a month ago, Colgate has moved up significantly from having to play the #3 seed in the first round (following the CHA and AHA auto-bids) to playing the #8 seed, Wisconsin. With just a few games remaining, I imagine that Colgate will have a pretty good idea of their at-large status. I think this might work in Cornell's favor somewhat, is that if in the back of all the Colgate players' minds, they know they're going to the tournament, they might not play with quite the same urgency as Cornell in the ECAC playoffs. On the same hand, Cornell must realize that if they want their season to continue, they must win. LGR!

[Q]chrisatpc Wrote:

It appears that according to this week's bracketology, Cornell is the last team to not make the NCAA tournament. [/Q] Actually, St. Cloud is the last team to miss the tournament. The first list on the page has the top 15 (including SCSU at 15) plus HC and Bemidji.

The point is the same; we almost certainly have to win out. (Though now I could see a loss in the final to 'Gate coupled with early upset losses by the other bubble teams pushing us in - especially since our RPI can't be hurt by winning against Clarkson.)

So Let's GO RED!!!

As discussed above (too lazy to check) ... does the bonus system put us ahead of St. Cloud?
Thus, we would be the last team out but I could be wrong ...

Disregarding the home and away bonus points, we need at least 0.0057 of bonus added to our RPI per neutral win (OSU and Notre Dame) to get ahead of St. Cloud.

Of course the only reason this is all necessary is if Colgate wins ECACs. Anyone else besides us or them and they take up another at-large spot. Similar if there's an upset in the CCHA or Hockey East (pretty sure that a WCHA team that's already in will win their tournament).

[Q]ben03 Wrote:

As discussed above (too lazy to check) ... does the bonus system put us ahead of St. Cloud?
Thus, we would be the last team out but I could be wrong ...
[/Q]

Oooo. Good point. I may have spoken too soon. Maybe the amount of the bonus matters, but the hypo numbers Jason Moy ran keep SCSU as the first team out.

Let's wait and see what happens to SCSU's RPI if/when they bow out of the WCHA playoffs this week at Minnesota.

[q]Let's wait and see what happens to SCSU's RPI if/when they bow out of the WCHA playoffs this week at Minnesota. [/q]

I doubt it'll go up, but I really can't see it dropping all that much.

Many people are forgetting that just about every single team is going to lose in the next two weeks. If Cornell sweeps this weekend, and goes 1-1 next weekend, with Harvard upsetting Brown weekend, it is very possible to get an at large big. I think I'd rather just win the ECAC than have to worry about it though.

[Q]calgARI '07 Wrote:

Many people are forgetting that just about every single team is going to lose in the next two weeks.[/q]Yeah, but most teams are either (a) so far ahead that they don't really matter or (b) playing higher quality teams than we are. We need the bubble teams to lose now.

[q]I think I'd rather just win the ECAC than have to worry about it though. [/Q]
Amen, brother.

Random playoff factoid. We've played four CCHA teams this, two of which are bubble teams, and thr four are facing off this weekend.
- Bowling Green at Ohio State
- Western Michigan at Notre Dame
The two bubble teams we need to lose are part of our SoS, but so are their opponents so these games are a wash wrt our SoS.

I don't know why I found this so interesting, but there it is.

So Keith, in layman's terms, is that good for us, bad for us, or does it not matter (sorry, I still get easily confused by all the pairwise stuff... but I'm working on it!!)?

[Q]dss28 Wrote:

So Keith, in layman's terms, is that good for us, bad for us, or does it not matter (sorry, I still get easily confused by all the pairwise stuff... but I'm working on it!!)? [/Q]We want the bubble teams to lose, so losses by OSU and Notre Dame are good for us.

But we played OSU and Notre Dame, so if they lose our SOS goes down - which is bad for us.

But we also played their opponents - WMU and BGSU - so if they win, our SOS improves, which is good for us.

So the SOS is a wash which is, neutral.

Therefore we return to a situation in which losses by Notre Dame and OSU are good for us.

Thank you and good night.

Okay, I'm following what you're saying... But since we beat both OSU and ND in the Everblades and early in the RS we did not beat both WMU and BGSU, wouldn't it be better for us for OSU and ND to win? Or does the "order" in which we played the teams not matter... ie, we "didn't win" at the beginning of the season when we played MWU and BGSU, but then as the season went on and we began to gel as a team we beat other teams (OSU, ND)... do later-in-the-season wins "count" more? (not including playoff games, which obviously count more than RS games )

I'm not sure if I'm making sense... sorry.

you have severely managed to over-think this wins help ... losses hurt ... order does not matter.

Haha... gotcha.

I'm bored and waiting for 7pm to come around... this overanalysis is the work of an extremely idle brain

But, my question was also: if a team (or teams) we beat managed to beat a team we didn't beat, doesn't that help us? Or does it not matter?

The winning percentage of the teams we played against figures into our RPI. Since we played WMU twice and NDame once, it helps our RPI if WMU wins. Also, since we are close behind NDame in PWR, our only shot a passing them is probably to have WMU win. On the other hand, if NDame stays in the top 15, we get bonus points, which might help us in comparison to some other teams. So, the answer is, it depends on what all the other results are also. I think rooting for WMU is the better bet.
Since we played Ohio State twice and BGSU once, OSU victories help our RPI and BGSU wins hurt it, but on the other hand, we may need to pass OSU in PWR to get in also, so rooting for BGSU might be the better bet. Again, it depends on everything else.
Between Ohio State, NDame and St. Cloud, we'd have to pass 2 of 3 for sure, (and if Colgate doesn't win the ECACs make that 3 of 4 including Colgate) so if both ND and OSU win, we've got no chance without an autobid.
At this point, just about everything would have to break right for us to have a remote chance at sneaking in without winning the ECACs, including a necessary sweep of Clarkson. So, let's just win our games and not worry about the rest.

What everyone has said is pretty much right. FWIW I think we root for ND and OSU to lose because they're bubble teams and we need bubble teams to drop for us to have a prayer. Jeff's probably right regarding the multiple game effect with OSU and WMU, but the negative RPI effect of OSU losing prob is washed out by the more direct effect on their RPI.

This whole excercise would be a lot more worthwhile if we'd just beaten WMU twice and BGSU. I haven't checked to see if Whelan has his What if? script running or not, but I suspect that if we had those extra four points we might be solidly on the bubble instead of hoping to get on the bubble.

The order of games no longer matters at all. It used to matter before they abolished the Last 16/Last 20 category in the PWR. But ti wouldn't have mattered here anywya, since all of our RS non-conf games will be more than 20 games old by the end of the season.

I don't think the transitive property doesn't help in the rankings (A > B > C > A)

If Western loses twice, are they still a TUC? If we eliminate a TUC loss and a TUC tie from our stats, how much does that help us?

It's a slippery slope. Their RPI is .5060, just over TUC status. If they lose 2 that could lower them below .5000. But wait! If Cornell sweeps (a must if we are talking about at-large bids), we might crack the RPI top 15. In that case, Western now has a "quality road win" that could push their RPI back over .5000.

[Q]KenP Wrote:

It's a slippery slope. Their RPI is .5060, just over TUC status. If they lose 2 that could lower them below .5000. But wait! If Cornell sweeps (a must if we are talking about at-large bids), we might crack the RPI top 15. In that case, Western now has a "quality road win" that could push their RPI back over .5000. [/Q]
Seems to me there was a situation a few years ago involving a midwestern school (Michigan?) where if they swept a playoff series in two games their opponent would end their season as a non-TUC, thereby hurting the first school's PWR. But if the first school took three games to win the series, their opponent would then remain a TUC, thus enhancing the first school's PWR.

So, conceivably, there can be situations where you should even root against your own team in order to improve its chances for an NCAA bid.

[Q]KenP Wrote:

It's a slippery slope. Their RPI is .5060, just over TUC status. If they lose 2 that could lower them below .5000. But wait! If Cornell sweeps (a must if we are talking about at-large bids), we might crack the RPI top 15. In that case, Western now has a "quality road win" that could push their RPI back over .5000.[/Q]
It doesn't work that way. You decide who the top 15 are using the pre-bonus RPI, then add the bonuses for quality wins agaisnt those teams.

[Q]jtwcornell91 Wrote:

It doesn't work that way. You decide who the top 15 are using the pre-bonus RPI, then add the bonuses for quality wins agaisnt those teams.
[/Q]

For argument sake, say Western is swept, and their RPI is .4995. Let's also say that Cornell is a top 15 team according to (pre-bonus) RPI at selection time. Western's quality win(s) will push their RPI back over .5000, making them a TUC once again.

[Q]jtwcornell91 Wrote:

[q]KenP Wrote:

It's a slippery slope. Their RPI is .5060, just over TUC status. If they lose 2 that could lower them below .5000. But wait! If Cornell sweeps (a must if we are talking about at-large bids), we might crack the RPI top 15. In that case, Western now has a "quality road win" that could push their RPI back over .5000.[/q]
It doesn't work that way. You decide who the top 15 are using the pre-bonus RPI, then add the bonuses for quality wins agaisnt those teams.
[/Q]I think KenP's example still works, John. In the hypo Cornell is top 15 RPI pre-adjustment, while WMU is sitting on the fringe. Post adjustment, WMU would bump over .500 and become a TUC. (Or is TUC also judged pre-adjustment?)

[Q]ugarte Wrote:

I think KenP's example still works, John. In the hypo Cornell is top 15 RPI pre-adjustment, while WMU is sitting on the fringe. Post adjustment, WMU would bump over .500 and become a TUC. (Or is TUC also judged pre-adjustment?)




Edited 1 times. Last edit at 03/12/2004 10:40am. [/Q]
Oh, sorry, I read too fast and thought you were talking about wins against WMU (which we don't have anyway) becoming "quality wins". What you say looks right.

[Q]jtwcornell91 Wrote:

ugarte Wrote:

I think KenP's example still works, John. In the hypo Cornell is top 15 RPI pre-adjustment, while WMU is sitting on the fringe. Post adjustment, WMU would bump over .500 and become a TUC. (Or is TUC also judged pre-adjustment?)




Edited 1 times. Last edit at 03/12/2004 10:40am.
Oh, sorry, I read too fast and thought you were talking about wins against WMU (which we don't have anyway) becoming "quality wins". What you say looks right.

[/Q]

1. I'm pretty sure you have to be a TUC to benefit from an RPI bonus, so in the hypothetical it wouldn't do WMU any good.

2. More importantly, even if swept, WMU won't drop out of TUC. Do the math yourself. They benefit as much or more from adding Notre Dame's winning % and their Opp % as they lose from adding 2 to their own loss column.

3. No way Cornell get's high enough in the PWR to earn a spot that way except if they win out, and probably not even then. Of course if they win out, an at-large is meaningless for them, but it will probably drop Colgate out of the top 14.

[Q]Ken '70 Wrote:1. I'm pretty sure you have to be a TUC to benefit from an RPI bonus, so in the hypothetical it wouldn't do WMU any good.[/q]Rule citation please? This doesn't make a lot of sense to me.
[q]3. No way Cornell get's high enough in the PWR to earn a spot that way except if they win out, and probably not even then. Of course if they win out, an at-large is meaningless for them, but it will probably drop Colgate out of the top 14.[/Q]Keep hope alive!

[Q]Do the math yourself. They benefit as much or more from adding Notre Dame's winning % and their Opp % as they lose from adding 2 to their own loss column.[/Q]
No they wouldn't benefit as much. Here's the math, with some liberal rounding and simplification. Two losses added to a .500 team, would drop their win percentage by .026 to .474. Assuming ND's winning percentage finishes at .600, and the current average of WMU's oppnents winning percentage is .500 (I don't have the data handy but it can't be too far off), the average would rise to .505. The effect on opponent's oppenents percentage is likely to be negligible. Thus they lose 25% of .026 and gain 50% of .005, for a net loss of .004. It does look like WMU would still be a TUC however.

I'm sure this has been mentioned somewhere else, but I haven't been able to go through all the messages.

After last night's results, Cornell is tied with Colgate for the #15 spot.


Edit: corrected 14 to 15

It looks like Cornell and Colgate are actually tied for the 15th slot, meaning that there would not be an at large bid for either team, given the autobids for the non-major conferences. Cornell currently wins the comparison with Colgate so that's a little help. Harvard is now a TUC and Brown is out of the TUC range, so that helped Cornell rise, despite the loss last night. Of course, Cornell has to win tonight and get more in Albany to have a chance....winning the tournament is still likely the only way to get in.

Wow... we flipped the Colgate comparison? Uhhh... Thanks, Harvard? Feels weird saying that.

Win out and forget about it or....

Speaking of flipping, it looks like the Cornell-Notre Dame comparison is flippable say if Ohio State beats Notre Dame at the CCHA championships which would give Cornell the common opponents piece of that comparison. A Cornell win against a TUC in the ECAC's would bring them to .500 against TUC's for the Notre Dame comparison since the Cornell-Notre Dame counts in the H2H and not the record against TUC's. Cornell could then win COp and TUC's against Notre Dame and win that comparison.

Just kidding...

But like Jayson Moy likes to do in his Bracketology, the above is mostly moot because Notre Dame would beat a TUC and a common opponent in Western Michigan as has been mentioned, to get to the CCHA championships and that would not make the comparison flippable and Notre Dame is helped by the bonus. I'd like to think that we are close enough to Notre Dame in the comparison and have a win over them that the committee would look at the comparison carefully if a bubble situation arose between both of us.

In the present PWR, Notre Dame, Colorado College and an ECAC team can not all make the NCAA's. It's kind of sad though that the WCHA has six, and the CCHA five of the top fourteen spots in the PWR and the ECAC currently has none.