|
Post by lonewolf on Oct 27, 2011 10:50:10 GMT -5
Good information Bofa. I think the 2 issues that often come up is how RPI is sometimes used, and that people interpret RPI as a ranking of the strength of teams in a comparison of who would win.
|
|
|
Post by The Bofa on the Sofa on Oct 27, 2011 10:53:39 GMT -5
what kind of astounds me is that a 12-7 team (Minnesota) has such a high rating. their opp win% must be off the charts. (not saying they aren't a good team, b/c they are. it's just odd to see 7 losses that far up on the RPI chart) The RPI system is too convoluted and should be dropped or seriously revised. I would vote for a team's wins at 75%,...team's opponents' wins at 25% and thats it!!!!!!!!!!! OUT! If you want it to more accurately reflect team strength and head-to-head outcome, you are actually going the wrong direction. Empirically, the oppOpp% is a better reflection of team quality. The problem is that it is a correlation, but not a causation. Being a better team does not make the OppOpp% better, but that's the way it turns out. OTOH, keep in mind that while the absolute RPI calculation uses 25/50/25, the relative values do not mirror that. That's because whereas RPI only uses 25% of a team's W/L percentage, the range of W/L percentage is huge, basically from .95 or higher down to 0. That means that the contribution to RPI due to a team's W/L varies from 0 - .24 or so. While Opp% has a bigger weight in RPI, it also comes with a smaller range. Ignoring freaks of nature like Minn, the usual range of Opp% is about 0.3 - 0.7, which means the contribution to RPI goes from 0.15 - 0.35, or a range of 0.2ish. Meanwhile, the OppOpp% ranges from about 0.4 - 0.6, and taken by 0.25 gives a contribution of 0.1 - 0.15, a max difference of only 0.05. As a result, W/L percentage ends up being the single biggest factor to contribute to RPI _differences_. If you look at the ranges possible, W/L percentage is about half, opponents% is about 40%, and oppOpp% is about 10%.
|
|
|
Post by lonewolf on Oct 27, 2011 10:59:08 GMT -5
Would they not understand that the WCC being 4-1 against the ACC debunks any argument that the ACC is a better and more 'deserving' conference than the WCC?? That's a very faulty argument. Note, I'm not saying that one conference is better than the other (though Pablo does have a higher ranking for the WCC), just that the argument of the WCC being 4-1 against the ACC in pre-conference play means that the conference is better is faulty.
|
|
|
Post by jayj79 on Oct 27, 2011 11:10:44 GMT -5
what kind of astounds me is that a 12-7 team (Minnesota) has such a high rating. their opp win% must be off the charts. (not saying they aren't a good team, b/c they are. it's just odd to see 7 losses that far up on the RPI chart) The RPI system is too convoluted and should be dropped or seriously revised. I would vote for a team's wins at 75%,...team's opponents' wins at 25% and thats it!!!!!!!!!!! OUT! I haven't done the math, but I would think that would make it even more in favor of the teams that play "weak competition", and the Westies would complain even more about "east coast bias". I haven't done the math, but a 75/25 break would probably move UNI even higher than 4th.
|
|
|
Post by The Bofa on the Sofa on Oct 27, 2011 11:15:32 GMT -5
Would they not understand that the WCC being 4-1 against the ACC debunks any argument that the ACC is a better and more 'deserving' conference than the WCC?? That's a very faulty argument. Note, I'm not saying that one conference is better than the other (though Pablo does have a higher ranking for the WCC), just that the argument of the WCC being 4-1 against the ACC in pre-conference play means that the conference is better is faulty. Absolutely. It all depends on who plays whom. If the top teams in the WCC are playing the bottom half of the ACC, then who cares? Of course the WCC would win. Without knowing the balance of the competition, these head to head comparisons are meaningless. Similarly, even if you know the balance of the competition, it has nothing to do with conference strength as measured by RPI. Notice in the Big Ten/Pac Ten comparison, no mention is made of Colorado, for example (since Colorado didn't play anyone in the Big Ten). How convenient! Because one of the main reasons that the Pac Ten is ranked below the Big Ten in RPI (and Pablo, for that matter) is because Colorado at 161 is dragging down the overall conference RPI. The next lowest Pac 12 team I find is Arizona St at 114, with Wash St at 113. Both are below Iowa at 112 (and Iowa beat Arizona St) The lowest team in the Big Ten is Indiana at 126. The issue is that RPI looks at ALL the teams in the conference, and doesn't just focus on the top teams, or those in head to heads. FWIW, the conclusions regarding Pac 12 and Big Ten are exactly the same in Pablo. Using the old alignment, the relative conference strengths would be exactly as we have seen in the past. However, the Pac 12 is getting hammered by the addition of Utah and Colorado, who are far below the standard for the rest of the league, whereas the Big Ten is being helped immensely by adding Nebraska. And as I have noted a couple of times, the Big 12 is actually having a net benefit, because as much as it hurts to lose Nebraska, they helped even more by getting rid of Colorado.
|
|
|
Post by jake on Oct 27, 2011 11:34:49 GMT -5
The RPI system is too convoluted and should be dropped or seriously revised. I would vote for a team's wins at 75%,...team's opponents' wins at 25% and thats it!!!!!!!!!!! OUT! I haven't done the math, but I would think that would make it even more in favor of the teams that play "weak competition", and the Westies would complain even more about "east coast bias". I haven't done the math, but a 75/25 break would probably move UNI even higher than 4th. I haven't don't the math either.But, putting my team's RPI number in the hands of my opponents' opponents seems a little preposterous. We need to allow quality teams in weaker conferences a level field (or court). Sure, I realize OOC scheduling can help,...but, that is sometimes easier said than done.
|
|
|
Post by dawgsfan on Oct 27, 2011 11:36:45 GMT -5
Since it hasn't been brought up, and a few in this thread so far seem to not have looked too deeply into the actual RPI computations I guess I'll post it. RPI, is a mathematical computation. The general formula is thus (there are some known and unknown weightings depending upon sport): 0.25 x Team record + 0.50 x Team Opponent record + 0.25 x Team Opponent record e.g. If team A has a is 20-10, their opponents have a record of 150-150, and the Opp, opp have a 60% win%(didn't feel like putting actual large numbers): .67x.25 + .50x.50 + .60x.25 =.5675 (then add weightings) Do this for all the teams and rank from highest to lowest. So basically in the end, teams that are looking to improve their RPI should schedule some of the upper teams from cupcake conferences. Playing a team like UNI, Northern Illinois, etc help your RPI even if you lose because of the fact that they will win 95% of their conference games. If this type of computation is even remotely true, then I would expect the SEC teams to be dropping very soon because of the lack of solid opponents throughout (talent and record wise).
|
|
|
Post by lonewolf on Oct 27, 2011 11:38:33 GMT -5
As a result, W/L percentage ends up being the single biggest factor to contribute to RPI _differences_. If you look at the ranges possible, W/L percentage is about half, opponents% is about 40%, and oppOpp% is about 10%. Very cool, thanks for posting these numbers, I've been meaning to look into that, but keep forgetting to work it out.
|
|
|
Post by siddhartha on Oct 27, 2011 11:42:18 GMT -5
I haven't done the math, but I would think that would make it even more in favor of the teams that play "weak competition", and the Westies would complain even more about "east coast bias". I haven't done the math, but a 75/25 break would probably move UNI even higher than 4th. I haven't don't the math either.But, putting my team's RPI number in the hands of my opponents' opponents seems a little preposterous. We need to allow quality teams in weaker conferences a level field (or court). Sure, I realize OOC scheduling can help,...but, that is sometimes easier said than done. Opponents winning % makes a lot of sense. A measure of your success is your opponents success. How else can you determine if a successful team in a weak conference is a "real" quality team?
|
|
|
Post by lonewolf on Oct 27, 2011 12:16:25 GMT -5
Since it hasn't been brought up, and a few in this thread so far seem to not have looked too deeply into the actual RPI computations I guess I'll post it. RPI, is a mathematical computation. The general formula is thus (there are some known and unknown weightings depending upon sport): 0.25 x Team record + 0.50 x Team Opponent record + 0.25 x Team Opponent record e.g. If team A has a is 20-10, their opponents have a record of 150-150, and the Opp, opp have a 60% win%(didn't feel like putting actual large numbers): .67x.25 + .50x.50 + .60x.25 =.5675 (then add weightings) Do this for all the teams and rank from highest to lowest. So basically in the end, teams that are looking to improve their RPI should schedule some of the upper teams from cupcake conferences. Playing a team like UNI, Northern Illinois, etc help your RPI even if you lose because of the fact that they will win 95% of their conference games. Pretty much. Though as Bofa points out, you're own win% is the biggest variant in the mix, but after you control that, you're OPP win% can make your standing a lot higher or lower. However, you don't know how those teams will fare in their own pre-season scheduling, and you would want to avoid all lower conference ones as the OppOpp could come into effect.
|
|
|
Post by dawgsfan on Oct 27, 2011 12:30:36 GMT -5
With all of these great posts about the RPI and how it is computed, it would be interesting to see what the RPI rankings looked like heading into the playoffs last year (haven't been able to find any archived RPI rankings).
I think last year six Pac-10 (12) teams made the tourny, with four of them making the elite eight (and eventually beating each other up until Cal lost to Penn State in the championship). I know 3 of the 4 were ranked in the Top 16 of the bracket rankings, but would be interesting to see if those were their RPI rankings or a little mixture of RPI and AVCA. Obviously UW was not deemed as strong by the RPI last year as they may have actually been.
Would love to see a mixture of some sort that takes into account RPI and AVCA. This type of system would be much like the BCS in football, but without the problems of only picking the Top 2 teams in the nation to play for a championship. I'm sure there would also be many that would see using the AVCA as an unfair system because of coaches voting certain teams higher than what their actual records or level of play represent (prestige effect). However, the same can be said about the RPI and its "bias" towards certain conferences and/or teams. As one poster already mentioned, you would assume the coaches are pretty knowledgeable about which teams are really that good out there (based off playing them, watching film, hearing from other coaches, etc). A mixture of the two systems would give a much more fair representation of who the top teams are because it is taking in the human factor (that can see a team is more talented than their record, or that a team lost a few games without a star player, etc) as well as the computer factor, which takes into account factors from every week and every match that coaches may omit from their memory because of how long ago it was.
A BCS type system would spit out rankings that may look like this: 1) Nebraska 2) Illinois 3) Stanford 4) Texas 5) Hawaii 6) Southern California 7) Northern Iowa 8) Iowa State 9) Penn State 10) UCLA 11) Minnesota 12) Purdue 13) Florida 14) California 15) Florida State 16) Pepperdine 17) San Diego 18) Tennessee 19) Kentucky 20) Michigan 21) Oklahoma 22) Washington 23) Oregon 24) Texas A&M 25) Miami 26) North Carolina 27) Western Kentucky 28) Tulsa 29) Cincinnati 30) Northern Illinois 31) Dayton 32) Michigan State 33) Ohio State 34) Ball State 35) Colorado State 36) Kansas State 37) Arizona 38) Western Michigan 39) Duke 40) Marquette 41) Missouri 42) Long Beach State 43) Houston 44) Missouri State 45) Kansas 46) Wichita State 47) Baylor 48) Notre Dame 49) Wisconsin 50) Louisville
Still see some flaws in this type of system, but appears to more accurately rank some teams (in my mind at least).
|
|
|
Post by mikegarrison on Oct 27, 2011 12:45:48 GMT -5
Opponents winning % makes a lot of sense. A measure of your success is your opponents success. How else can you determine if a successful team in a weak conference is a "real" quality team? This only makes sense on the surface. Why not just use a team's winning percentage? Well, OK, because they may have played all cupcakes. Their winning percentage may not reflect their real quality. OK, so let's judge them by their strength of schedule. And to do that, we judge the teams they play by ... winning percentage! If winning percentage is so unreliable to judge one team, what makes it a good thing for judging the teams they play against? Conversely, if it really is a fair test of strength of schedule, then why not just use it directly on the team in question? Fundamentally, the RPI is a stupid idea. As with many stupid ideas, if you tweak it enough you can kind of make it work. But that doesn't mean it's not a stupid idea. The pablo-type scheme is much better.
|
|
|
Post by stand on Oct 27, 2011 13:02:23 GMT -5
With all of these great posts about the RPI and how it is computed, it would be interesting to see what the RPI rankings looked like heading into the playoffs last year (haven't been able to find any archived RPI rankings). You asked for it... Attachments:
|
|
|
Post by Barefoot In Kailua on Oct 27, 2011 13:05:21 GMT -5
I think we can all agree that the criteria the NCAA Women's Volleyball Championship Committee relies on to seed tourney teams needs to be chamged. While RPI can be part of the equation, I don't belive it should be "THE" equation. But no one has really talked about is another obvious problem with the Chapionship tournament: Regionalization. Take the University Park Regional last year for example: Penn State enter the tourney as the National 4 seed with UNI (5th) Duke (12th), and LSU (13th) as the other seeds. Both UNI and LSU lose in the first round and Penn State needs only to beat the 12th seeded Blue Devils to advance to the Final Four (along with Niagra, Viginia Tech, and Oklahoma along the way.)
|
|
|
Post by dawgsfan on Oct 27, 2011 13:12:53 GMT -5
With all of these great posts about the RPI and how it is computed, it would be interesting to see what the RPI rankings looked like heading into the playoffs last year (haven't been able to find any archived RPI rankings). You asked for it... Thank you VERY much. That is crazy when just looking at the Pac 10 teams that were in the elite eight. Cal was 10 in RPI, Southern California was 8, Stanford was 2, and Washington.......37!?!?!? UW was ranked lower than Arizona (36) and UCLA (27) also. Is this how good of a predictor the RPI is? I know there are teams that get on hot streaks and just match-up well against other teams (much like NCAA BBall Tourny), but you would expect that a thing that has such a large impact on the tournament selections would be a little more accurate in predicting who should be in and where they should be ranked.
|
|