PDA

View Full Version : Some interesting RPI tidbits



caduceus
03-11-2010, 08:51 AM
So, guess which teams helped our RPI the most this season? Some are obvious, others, not so much. Here's the breakdown of the top 15, per RPI forecast (http://www.rpiforecast.com/teams/Gonzaga.html):

Team
1. Gonzaga +16.39%
2. St. Mary's +4.35
3. Portland +1.69
4. Duke +1.11
5. Wisconsin +1.05
6. Michigan St. +0.93
7. Memphis +0.86
8. Loyola Marymount +0.59
9. Wake Forest +0.54
10. Illinois +0.30
11. Cincinnati +0.19
12. Purdue +0.13
13. Washington St. +0.12
14. Pacific +0.11
15. Kansas +0.09

And the 15 teams that hurt our RPI the most?

1. Pepperdine -2.25%
2. Santa Clara -1.45
3. San Diego -1.29
4. San Francisco -1.26
5. Cal St. Bakersfield -0.95
6. Eastern Washington -0.82
7. Mississippi Valley St. -0.75
8. Oklahoma -0.20
9. Portland St. -0.07
10. Sacramento St. -0.07
11. Cal Poly -0.06
12. Toledo -0.06
13. SIU Edwardsville -0.06
14. Colorado -0.06
15. North Dakota -0.06

The RPI is in part weighted by our opponent's opponents record, hence the reason some teams we didn't play affect GU's RPI. Obviously, conference opponents will carry more weight, since we play them more than once. The numbers may change slightly since there are still tourneys going on.

=caduceus=

ZagLawGrad
03-11-2010, 08:53 AM
Always nice to see our own conference helping to put a dent in things.

caduceus
03-11-2010, 09:00 AM
BTW, our RPI improved from 37 to 36 yesterday. Gotta hope our OOC foes do well in the tourneys, every bit helps. Could make the difference between a low 6 seed and a high 7.

sittingon50
03-11-2010, 09:11 AM
Go Cincy!

Ziggy
03-11-2010, 09:25 AM
Go Memphis! Go Illinois!


On, Wisconsin!

tinfoilzag
03-11-2010, 09:40 AM
I'm probably wrong but it looks like the WCC net RPI affect on GU is -.028%?

That surprises me but I'm assuming that SF, LMU, and SMC all had a jump in their RPI because they beat us meaning if we had run the table in the WCC the negative impact on our RPI would of been greater.

U Zig, I Zag
03-11-2010, 10:11 AM
http://www.globalgallery.com/prod_images/600/esc-e6.jpg

caduceus
03-11-2010, 10:23 AM
I'm probably wrong but it looks like the WCC net RPI affect on GU is -.028%?

That surprises me but I'm assuming that SF, LMU, and SMC all had a jump in their RPI because they beat us meaning if we had run the table in the WCC the negative impact on our RPI would of been greater.

Take a look at the link above for more details, as well as here (http://rpiforecast.blogspot.com/2009/12/rpi-broken-down-by-contributions-by.html). Each team is weighted differently, and a team's own contribution is also affected by home/away/neutral weighting.

Don't forget, the biggest influence is your own winning percentage. Kansas, for example, is +23.76% in helping their own RPI. If we had run the table in conference, the percentage on the Gonzaga line would be higher.

St. Mary's helped their own RPI to the tune of +17.13% (and GU's contribution was +3.72%).
San Francisco hurt their own RPI at -7.91% (GU gave them +2.98%).

zagzilla
03-11-2010, 10:30 AM
What jumped out at me is that getting hammered by Duke helped our RPI more than beating Wisconsin.

It's things like that which make me question the RPI as an effective measurement of a team.

Also let me rant again about CSB. Winning that game hurt us by almost a full point. As we are scraping for a 6 vs 7, that could be a difference maker. I understand the reason we put a game in the schedule there, just not why we didn't pick a DII school if we wanted a patsy.

ZZ

Once and Future Zag
03-11-2010, 10:39 AM
Looking at a particular team's RPI is not very precise - and the selection committee doesn't use it that way - but they do look at the RPI's of their wins/losses as it's a decent way to "group" categories of wins/losses as the imprecision is reduced by having a larger sample size.

That means we're a W or L in the 1-50 metrics of teams we've played, but the committee wouldn't care if we were #1 or #50 beyond that as far as it goes for us.

I think that the "powers that be" have really failed to educate the public - and even the sports pundits - what it means and how the committee uses it. Due to what it is useful for, people get way too hung up on a team's RPI... It's like if someone thought Blocks per cheerleader was a useful stat. It annoys me how many sites present it as if it "means something" - like Richard Dreyfuss' pile of mashed potatoes.

Now KenPom's stats ARE good evaluation tools for how precisely good a team is, and how they can be directly measured against each other - RPI is useless for that.

caduceus
03-11-2010, 10:59 AM
Looking at a particular team's RPI is not very precise - and the selection committee doesn't use it that way - but they do look at the RPI's of their wins/losses as it's a decent way to "group" categories of wins/losses as the imprecision is reduced by having a larger sample size.

That means we're a W or L in the 1-50 metrics of teams we've played, but the committee wouldn't care if we were #1 or #50 beyond that as far as it goes for us.

I think that the "powers that be" have really failed to educate the public - and even the sports pundits - what it means and how the committee uses it. Due to what it is useful for, people get way too hung up on a team's RPI... It's like if someone thought Blocks per cheerleader was a useful stat. It annoys me how many sites present it as if it "means something" - like Richard Dreyfuss' pile of mashed potatoes.

Now KenPom's stats ARE good evaluation tools for how precisely good a team is, and how they can be directly measured against each other - RPI is useless for that.

The RPI is definitely over-emphasized in the sports press. The committee has consistently stated that it is only one of many tools they use to evaluate teams, while the commentators seem to point out a team's RPI as the final word.

I think KenPom's site is another helpful tool, but should also be taken with a large grain of salt in some respects. I don't think for a minute that Washington is #31 while the Zags are #54 and St. Mary's at #43 (as of today's rankings).

Once and Future Zag
03-11-2010, 11:20 AM
The RPI is definitely over-emphasized in the sports press. The committee has consistently stated that it is only one of many tools they use to evaluate teams, while the commentators seem to point out a team's RPI as the final word.

I think KenPom's site is another helpful tool, but should also be taken with a large grain of salt in some respects. I don't think for a minute that Washington is #31 while the Zags are #54 and St. Mary's at #43 (as of today's rankings).

Yeah - my problem with the committee is they don't explain HOW they use it :)

While there are some always some problems with any statistical method... I can't find myself disagreeing much with KenPom (though I haven't seen the Little Sisters of the Dub play for comparison).

Note his consistency metric. - we're ranked 248 - when we're good, we're awesome, but we also swing wide the other way and can be decidedly mediocre, whereas St Mary's is a quite a bit more consistent at 139.

That really jives with what I've seen this year. I think people tend to see the best game we play, and then set that as the bar for "how good" the team is, and why they may be perceived as being ranked low.

Part of my job is data analysis, and Pomeroy does as good a job as there is in trying to make his numbers really MEAN something. Where I think he could use some improvement, from my understanding of how he comes up with his numbers - at least for ranking purposes, as opposed to their predictive purposes, is that statistical outlier games have undue weight, in the "overall" number. If he bell curved the games to weight them stronger, the more normative a game it was for the team, I think that would give a more accurate number for "ranking"

Putting my fan-hat aside, we still have a very good team, and one that can make the Sweet Sixteen with the right matchups, but I think we're more likely to end our season in the round of 32 than winning that additional game. It just depends which GU team shows up. I'd love to be wrong, and make a nice run this year as well...