RebKell's Junkie Boards
Board Junkies Forums
 
Log in Register FAQ Memberlist Search RebKell's Junkie Boards Forum Index

RPI rankings
Goto page 1, 2  Next
 
Post new topic   Reply to topic    RebKell's Junkie Boards Forum Index » NCAA Women's Basketball - General Discussion
View previous topic :: View next topic  
Author Message
myrtle



Joined: 02 May 2008
Posts: 32326



Back to top
PostPosted: 12/10/19 1:49 pm    ::: RPI rankings Reply Reply with quote

First ones out. Really too early but interesting:
https://www.ncaa.com/rankings/basketball-women/d1/ncaa-womens-basketball-rpi



_________________
For there is always light,
if only we’re brave enough to see it.
If only we’re brave enough to be it.
- Amanda Gorman
Gamecock1



Joined: 02 Dec 2018
Posts: 55



Back to top
PostPosted: 12/10/19 3:09 pm    ::: Reply Reply with quote

That is a very confusing rating on so many levels.


Stormeo



Joined: 14 Jul 2019
Posts: 4701



Back to top
PostPosted: 12/10/19 3:16 pm    ::: Reply Reply with quote

Always interesting (if nothing else) in general. At this time of the year, it's also interesting to see just how long these mid-majors at the top of it can stay up there. Princeton at #5 seems like the biggest surprise? Good news for Bella Alarie fans, I suppose. Notable wins include FGCU (home) and Seton Hall (away), with their lone loss being in OT vs. Iowa (away).

Anyone know if WBB is ever gonna move to using NET rankings like MBB?


Coyotes



Joined: 28 Jan 2018
Posts: 1467



Back to top
PostPosted: 12/10/19 7:36 pm    ::: Reply Reply with quote

While it's EARLY, I don't know that I've seen a mid-major as high as #2 in Missouri State this early, followed by #5 Princeton, and #8 Gonzaga. It'll be interesting to see how their RPI holds up once we start hitting conference play. We might see the Ivy League have a higher than expected RPI this season, so come Selection Sunday, people might be hoping that Princeton keeps winning along with Gonzaga.

From a Stanford perspective, it's probably a little disappointing that Syracuse isn't living up to expectations at #115 early, but that's what happens when you lose to Green Bay, and your best win is arguably over #97 Ohio....in any case, hopefully Gonzaga and Mississippi State can stay reasonably high this season.


Missouri State has beaten #60 Minnesota, #54 Oklahoma, #65 Boise State, #14 South Dakota, in order to be #2. It's not a bad collection of wins, but I think they're high because they have only played one sub 200 team in Corpus Christi. Odd how this works out. Virginia at #20 also stands out as an oddity despite playing a pretty tough schedule.


myrtle



Joined: 02 May 2008
Posts: 32326



Back to top
PostPosted: 12/10/19 9:22 pm    ::: Reply Reply with quote

It's kind of too bad when a mid plays a decent preseason schedule but their RPI gets ripped to shreds when they get into league because the other teams in their league are so bad. It's one of the reasons why (I think) Geno tries to keep playing good teams even after their league starts. one of the (many) things wrong about using RPI anyway.



_________________
For there is always light,
if only we’re brave enough to see it.
If only we’re brave enough to be it.
- Amanda Gorman
calbearman76



Joined: 02 Nov 2009
Posts: 5152
Location: Carson City


Back to top
PostPosted: 12/11/19 1:43 am    ::: Reply Reply with quote

The Princeton example is both a good example of how flawed the RPI is, particularly this early in the season, but also as a case study of what could happen as the season goes on. Princeton dropped from 5 to 11 as a result of their win tonight. Hartford, a team that had been good a few years ago, is now 0-9 this season, so Princeton's win did little to help their winning %(WP) (.889 vs. .875) but dropped their opponents' winning %(OWP) significantly (.571 vs. .667). Since the OWP is weighted twice as much as WP, the impact is magnified further.

Going forward Princeton will be helped by the overall success of the Ivy League this season. The Ivy League has the 6th best non-conference record (65.7%), only behind the 5 majors. (The Missouri Valley is 7th at 65.4%) Even though the Ivies rolled up their record primarily against the really weak eastern conferences. If Princeton can go through the season with only one or two more losses they could stay in the top 20, but without a better win than Florida Gulf Coast the Tigers would probably still be looking at an 8 or 9 seed. And I don't think they deswerve better.


linkster



Joined: 27 Jul 2012
Posts: 5408



Back to top
PostPosted: 12/11/19 3:48 pm    ::: Reply Reply with quote

My suspision, after reviewing the evolution of the NCAA field to 64, is that there has always been a battle between the small conferences and the majors about who gets into the NCAA's. And the compromise was that 32 conference champs get in but that the at-large teams would be majors. The RPI/SOS is the tool that accomplishes that.


PickledGinger



Joined: 04 Oct 2013
Posts: 1350



Back to top
PostPosted: 12/11/19 4:54 pm    ::: Reply Reply with quote

Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


Shades



Joined: 10 Jul 2006
Posts: 63713



Back to top
PostPosted: 12/11/19 5:20 pm    ::: Reply Reply with quote

PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI’s are not an unflawed ranking system, especially this early in the season. That explains why Missouri St is at #2 and Creighton is at #8. I think those are farther from reality as compared to Iowa at #24.



_________________
Nnekalonians 1:14 - Thou shalt not accept that which is not earned
pilight



Joined: 23 Sep 2004
Posts: 66773
Location: Where the action is


Back to top
PostPosted: 12/11/19 5:27 pm    ::: Reply Reply with quote

PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.



_________________
Let us not deceive ourselves. Our educational institutions have proven to be no bastions of democracy.
PickledGinger



Joined: 04 Oct 2013
Posts: 1350



Back to top
PostPosted: 12/13/19 8:30 pm    ::: Reply Reply with quote

pilight wrote:
PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.


Wait, WHAT?!? I thought that was the whole point of the algorithm! Is there a rating system that does?


pilight



Joined: 23 Sep 2004
Posts: 66773
Location: Where the action is


Back to top
PostPosted: 12/13/19 8:33 pm    ::: Reply Reply with quote

PickledGinger wrote:
pilight wrote:
PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.


Wait, WHAT?!? I thought that was the whole point of the algorithm! Is there a rating system that does?


RPI just takes your winning percentage and multiplies it by your SOS. It doesn't matter which games you win, just how many.



_________________
Let us not deceive ourselves. Our educational institutions have proven to be no bastions of democracy.
Conway Gamecock



Joined: 23 Jan 2015
Posts: 1881
Location: Here


Back to top
PostPosted: 12/13/19 10:51 pm    ::: Reply Reply with quote

pilight wrote:
PickledGinger wrote:
pilight wrote:
PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.


Wait, WHAT?!? I thought that was the whole point of the algorithm! Is there a rating system that does?


RPI just takes your winning percentage and multiplies it by your SOS. It doesn't matter which games you win, just how many.


If it involves your strength of schedule (SOS) - or how strong your schedule is - then I guess by definition it DOES matter which games you win, or who you beat. The determination of just how "strong" your schedule is, by definition, is directly conditional on just who you have scheduled, and whether or not you beat any of them. The strength of those teams you schedule, is determined by the strength of THEIR schedules, and whether or not they beat any of THOSE teams...

The Rating Percentage Index (RPI) is based on a formula where 25% is based on the team in question's wining %, plus 50% of their scheduled opponents' winning % against their opponents, plus a final 25% of winning % of their opponents' opponents.

The idea is not just to look at Team A's win-loss record, because any particular team could just schedule a bunch of weak pasties, or be a part of a particularly weak conference. So Team A could end up having a fantastic winning %, but still not be that great of a team because they only played a ton of cupcakes.

But the idea goes, that if each opponent on Team A's schedule ALSO has a great winning % of their own, then they may - to a degree - be a competitive team, and that adds value to Team A's schedule. Then to take it another dimension, if the formula looks at the winning % of the opponents of Team A's opponents, and if THOSE opponents have a solid winning %, then that adds even more value to the opponents on Team A's schedule.

The intent was to apply incentive to programs seeking to schedule games with consistently strong non-conference opponents. They can't really control their conference schedules, and how strong or weak their conferences are. But if everyone in the AAC - as an example - scheduled OOC schedules like Connecticut does annually, then the overall RPI of the AAC would rise, and perhaps then Connecticut wouldn't have to go schedule such a touch OOC schedule every year.

But the final element of the RPI equation, is Team A has to win games. If Team A schedules a lot of high RPI opponents, but loses to all of them, that won't help their RPI very much. A team that typically has a decent RPI, but a poor SOS, is a team with a LOT of wins. Oregon St. is #1 in RPI, and also has the #4 SOS, and is undefeated thus far (8-0). Half of their wins are against top 50 RPI opponents, including #2 Missouri St. and #6 DePaul.

Missouri St. has the #2 RPI as well as a slightly higher (#2) SOS then Oregon St.'s. They are 8-1, with their lone loss to Oregon St. As stated above, they have a number of solid wins against near-top 50 or top 100 RPI opponents.

That's the big issue with RPI: early in the year with a lot of teams playing a lot of cupcakes, it's hard for the formula to get an accurate reading. A team could have a big win against a relatively high RPI opponent, or a loss against a relatively poor RPI one - just one single game - and it can fling a team way up or way down the RPI rankings. It's an average that again, includes the team's winning %, which means it includes the team's wins, or games. So the more games a particular team plays and wins, the more accurate the RPI rating will become. By the very formula of the thing, the RPI will be it's most accurate for any team, only AFTER that team has concluded it's schedule. Even one game shy of ending it's season, and the RPI for that team will have a degree of inaccuracy to it, by design.

So, we look at Virginia, which Coyotes described as an "oddity" by comparing their #20 RPI. Oregon State as we've shown has the #1 RPI, but UVA has the #1 SOS at this point. So why are Virginia only #20 in RPI? Because they are 4-5 overall, while Oregon St. is 8-0, and Missouri State is 8-1. You can play a killer schedule, but if you go 0-30, you shouldn't get too high of a RPI rating (your SOS ranking will be killer, though).

The RPI is a formula to help the NCAA Tournament committee pick between which 28-4 team should rate the NCAAT, and which 28-4 team should not, if it came down to one spot. If the first 28-4 team played a solid to stout SOS, where the combined record of it's opponents was above .500, and the overall combined record of THEIR opponents were above .500 as well, while the second 28-4 team did not, where their opponents' combined record was below .500, then the first team should warrant the invite, and the selection committee has a plausible explanation for why the 2nd team did not.

If Virginia was even 7-2 or perhaps 6-3, they would be challenging Oregon St. and Missouri St. for top spot in the RPI ranking, for sure. But you not only can just schedule tough teams: you got to beat them too. Virginia has played 7 opponents with top 75 RPIs, and they are 2-5 against them, and 0-4 against the top RPI opponents Connecticut, Rutgers, Kentucky, and UCLA. Just like it's easy for Team A to have a great win-loss record if all they do is beat cream puff opponents, it's also easy for Team A to play the toughest schedule in the land, if they aren't required to win any of them. The RPI is looking for the best balance of both...


pilight



Joined: 23 Sep 2004
Posts: 66773
Location: Where the action is


Back to top
PostPosted: 12/14/19 8:01 am    ::: Reply Reply with quote

Conway Gamecock wrote:
pilight wrote:
PickledGinger wrote:
pilight wrote:
PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.


Wait, WHAT?!? I thought that was the whole point of the algorithm! Is there a rating system that does?


RPI just takes your winning percentage and multiplies it by your SOS. It doesn't matter which games you win, just how many.


If it involves your strength of schedule (SOS) - or how strong your schedule is - then I guess by definition it DOES matter which games you win, or who you beat. The determination of just how "strong" your schedule is, by definition, is directly conditional on just who you have scheduled, and whether or not you beat any of them. The strength of those teams you schedule, is determined by the strength of THEIR schedules, and whether or not they beat any of THOSE teams...


My point was that if Iowa had beaten Washington and instead lost some other game on their schedule, their RPI would be the same as it is now.



_________________
Let us not deceive ourselves. Our educational institutions have proven to be no bastions of democracy.
Conway Gamecock



Joined: 23 Jan 2015
Posts: 1881
Location: Here


Back to top
PostPosted: 12/14/19 5:40 pm    ::: Reply Reply with quote

pilight wrote:
Conway Gamecock wrote:
pilight wrote:
PickledGinger wrote:
pilight wrote:
PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.


Wait, WHAT?!? I thought that was the whole point of the algorithm! Is there a rating system that does?


RPI just takes your winning percentage and multiplies it by your SOS. It doesn't matter which games you win, just how many.


If it involves your strength of schedule (SOS) - or how strong your schedule is - then I guess by definition it DOES matter which games you win, or who you beat. The determination of just how "strong" your schedule is, by definition, is directly conditional on just who you have scheduled, and whether or not you beat any of them. The strength of those teams you schedule, is determined by the strength of THEIR schedules, and whether or not they beat any of THOSE teams...


My point was that if Iowa had beaten Washington and instead lost some other game on their schedule, their RPI would be the same as it is now.


I understand your point, but only 25% of the formula depends on the team's winning %. It's true that FOR THAT 25%, it doesn't matter if a team is 12-0 in their non-conference schedule, and only played against 12 cupcakes, or 12 really good, tough teams. Their winning % would still be 100% either way, so again for that 25% portion of the equation, the team's RPI stays the same regardless.

But then the other 75% comes into play: it's very doubtful that the 12 opponents that are cupcakes also play as tough of a schedule as the 12 opponents that are really good, tough teams. And further, it's even more doubtful that had they actually played similar schedules, that their own winning % would be as good as the winning % of those other really good, tough teams, because they are cupcakes. They most likely play in cupcake conferences, and still may struggle to win games.

The 12 opponents that are really good, tough teams may play in better conferences, and even play against several really good, tough non-conference opponents, but still end up with better winning %'s than those other 12 opponents. And likewise, if they do, then their tougher opponents will most likely also play tougher schedules, and have solid winning %'s. The first sentence covers 50% of the RPI algorithm, and the second sentence covers the final 25%.

So by your example, YES it won't matter to Iowa's RPI if they beat Washington and lost to a lower opponent, as opposed to losing to Washington and beating that lower opponent, because both are still on Iowa's schedule together, and Iowa would still end up 1-1 either way. Doing one or the other doesn't impact any part of Iowa's overall RPI rating, based on just that aspect of the formula.

But if it did matter, what would give Iowa the better RPI: beating Washington and losing to a lower team, or beating the lower team and losing to Washington? Should just one game out of the schedule - an impressive win against a higher-quality opponent, make the overall schedule RPI? What about that loss to that lower-level opponent: wouldn't it more likely make the win a fluke, and bring Iowa right back down?

The best indicator is to win BOTH games against Washington and the lower team, and be 2-0. That's the first 25% of the formula. THEN add the other 75% of Washington's opponents' winning %, and their opponent's winning %.

The RPI wasn't designed to point out teams that were able to beat one particular good opponent here or there, like Iowa beating Washington. It was designed to help out the post-season tournament selection committees figure out which 28-4 team is the better, more stronger 28-4 team for selection and seeding purposes. And the post-season selection committees wouldn't care less about Iowa beating Washington, if Iowa finishes 14-18 on the season.

They are looking at the whole picture, not a per-week perspective after only 8-10 games in. That's what fans do. And the RPI formula is the most accurate at the end of the season, when those committees need to determine who will play in their tournaments....


Conway Gamecock



Joined: 23 Jan 2015
Posts: 1881
Location: Here


Back to top
PostPosted: 12/14/19 5:54 pm    ::: Reply Reply with quote

pilight wrote:
Conway Gamecock wrote:
pilight wrote:
PickledGinger wrote:
pilight wrote:
PickledGinger wrote:
Someone please explain how Iowa can be ranked 24 with a loss to 156 Washington on their record, who themselves have lost to 157 Hawai’i and 180 Tulane?


RPI doesn't care you beat or who you lose to. All that matters is records.


Wait, WHAT?!? I thought that was the whole point of the algorithm! Is there a rating system that does?


RPI just takes your winning percentage and multiplies it by your SOS. It doesn't matter which games you win, just how many.


If it involves your strength of schedule (SOS) - or how strong your schedule is - then I guess by definition it DOES matter which games you win, or who you beat. The determination of just how "strong" your schedule is, by definition, is directly conditional on just who you have scheduled, and whether or not you beat any of them. The strength of those teams you schedule, is determined by the strength of THEIR schedules, and whether or not they beat any of THOSE teams...


My point was that if Iowa had beaten Washington and instead lost some other game on their schedule, their RPI would be the same as it is now.


But I do agree that the RPI system is not the most accurate system to use, for purposes of comparison during the season, or for a league that is very competitive and where parity rules the day. They are using NET for MBB because that league is far more competitive and has far more parity. The WBB lags a good deal behind MBB in that regard, but is consistently making strides forward to catch up.

One looks at Olympic sports as a reference: the competitiveness (and political-ness) of Olympic sports competition has changed how they score participants. Used to be it was a simple 1 to 10 score system from the panel of judges. Nowadays - and for some time now - competitors win Olympic gold medals because they have a performance that earns a score of 9.8155, and the next competitor scores a 9.8152 score.

WBB ratings and rankings haven't got THAT anal quite yet. But as sports continue to become more competitive at the top, the systems being used will constantly become more accurate as well.....


pilight



Joined: 23 Sep 2004
Posts: 66773
Location: Where the action is


Back to top
PostPosted: 12/14/19 6:44 pm    ::: Reply Reply with quote

The RPI formula is not very good at figuring Strength of Schedule either.

Compare these two non-conference schedules from last season and decide which one you think is tougher:

Schedule A

Chadron State
@ St Mary's
Colorado Springs U
North Texas (neutral site)
@ Pacific
Montana St
Denver
Idaho
South Dakota St

Schedule B

SE Missouri
@ Virginia
Lamar
Coppin State
Furman
Jackson State
@UALR
@ Texas
Marquette
@Southern Miss
@ Oregon
@ Washington
Louisiana


According to RPI, schedule A was much tougher despite featuring no ranked teams, no major conference teams, and only one opponent in the top 50 of RPI. It ranked #26 in the country for non-con SOS while schedule B ranked #76.



_________________
Let us not deceive ourselves. Our educational institutions have proven to be no bastions of democracy.
SpaceJunkie



Joined: 10 Sep 2012
Posts: 4241
Location: Minnesota


Back to top
PostPosted: 12/14/19 9:32 pm    ::: Reply Reply with quote

pilight wrote:
The RPI formula is not very good at figuring Strength of Schedule either.

Compare these two non-conference schedules from last season and decide which one you think is tougher:




According to RPI, schedule A was much tougher despite featuring no ranked teams, no major conference teams, and only one opponent in the top 50 of RPI. It ranked #26 in the country for non-con SOS while schedule B ranked #76.


The best way to help your SOS is never ever play those severe cupcake teams. Coppin St and a lesser extant Louisiana are probably way schedule B is ranked so much lower. RPI thinks the difference between a terrible P5 team and Coppin St is more than the difference between UConn and a bubble team.

Of course since RPI is 50% opponent's record and 25% opponent's opponent's record, playing teams from cupcake conferences with records much better than the records of arugably better teams with harder schedules helps too (St. Mary's, Idaho, Boise St)


pilight



Joined: 23 Sep 2004
Posts: 66773
Location: Where the action is


Back to top
PostPosted: 12/14/19 9:50 pm    ::: Reply Reply with quote

SpaceJunkie wrote:
pilight wrote:
The RPI formula is not very good at figuring Strength of Schedule either.

Compare these two non-conference schedules from last season and decide which one you think is tougher:




According to RPI, schedule A was much tougher despite featuring no ranked teams, no major conference teams, and only one opponent in the top 50 of RPI. It ranked #26 in the country for non-con SOS while schedule B ranked #76.


The best way to help your SOS is never ever play those severe cupcake teams. Coppin St and a lesser extant Louisiana are probably way schedule B is ranked so much lower. RPI thinks the difference between a terrible P5 team and Coppin St is more than the difference between UConn and a bubble team.

Of course since RPI is 50% opponent's record and 25% opponent's opponent's record, playing teams from cupcake conferences with records much better than the records of arugably better teams with harder schedules helps too (St. Mary's, Idaho, Boise St)


I get why the RPI rates A as harder. I'm saying the RPI is wrong and schedule B is harder, actually much harder.



_________________
Let us not deceive ourselves. Our educational institutions have proven to be no bastions of democracy.
SpaceJunkie



Joined: 10 Sep 2012
Posts: 4241
Location: Minnesota


Back to top
PostPosted: 12/14/19 10:22 pm    ::: Reply Reply with quote

pilight wrote:
I'm saying the RPI is wrong and schedule B is harder, actually much harder.


Yeah. Winning the msot games in B much harder than A.


elsie



Joined: 08 Apr 2016
Posts: 271



Back to top
PostPosted: 12/15/19 1:33 am    ::: Reply Reply with quote

good info....since you seem to understand all this, tell me, does the SOS or RPI take into account playing on your opponents home court?.....it seems to me it should....


willtalk



Joined: 13 Apr 2012
Posts: 1088
Location: NorCal


Back to top
PostPosted: 12/16/19 3:56 am    ::: Reply Reply with quote

The idea of RPI is a good one. it just is sort of archaic and certainly could be improved with a few slight adjustments. I believe it was initially meant only as a rough measurement for people who did not have all the records available that are now at fingertip access due to the internet.

I even suggested a simple one, which would eliminate any team a certain number of RPI points lower from the team unless it resulted in a loss. It would eliminate all the lower teams that now can bring down a teams otherwise higher RPI. The wins wouldn't count either. That would, at a glance give a better perspective to a team's won-loss record in respect to their actual SOS.

All the teams would be factored into the general RPI pool. Teams would just be eliminated after the fact to give a better SOS metric using RPI. Therefor if a team played a large number of difficult games it would be reflected as a strong schedule with the really low teams not pulling it down. The cut-off point for teams would be any reasonable difference that would make it unlikely for that team to have a chance to win. The cut of point for eliminating teams would be adjustable throughout the season because the main stats would always include everyone.



_________________
No one one is ever as good as their best game, nor as bad as their worst.
Phil



Joined: 22 Oct 2011
Posts: 1258



Back to top
PostPosted: 12/18/19 6:43 pm    ::: Reply Reply with quote

> The idea of RPI is a good one.

I disagree. It does have two redeeming aspects, one of which is valid the other of which is questionable. The valid aspect is that it is very easy to calculate, but that was a meaningful consideration before computers, people needed adding machines or simplified calculators. That positive attribute has been meaningless for decades.

The other important attribute is that it ignores margin of victory, which some people argue helps to discourage teams running up the score. It's difficult to prove or disprove this without a controlled experiment, but I think the argument is bunk. Note that I'm not saying that teams don't try to run up the score – did you pay attention to Baylor today? I'm saying that the existence of the RPI doesn't act as a deterrent.

RPI is very crappy early in the season, which is one of the reasons they don't release it too early, and gets less crappy as the season goes along but it is always deficient compared to most other computer or human-based rankings. It ought to be scrapped.


calbearman76



Joined: 02 Nov 2009
Posts: 5152
Location: Carson City


Back to top
PostPosted: 12/18/19 9:50 pm    ::: Reply Reply with quote

Phil wrote:
> The idea of RPI is a good one.

I disagree. It does have two redeeming aspects, one of which is valid the other of which is questionable. The valid aspect is that it is very easy to calculate, but that was a meaningful consideration before computers, people needed adding machines or simplified calculators. That positive attribute has been meaningless for decades.

The other important attribute is that it ignores margin of victory, which some people argue helps to discourage teams running up the score. It's difficult to prove or disprove this without a controlled experiment, but I think the argument is bunk. Note that I'm not saying that teams don't try to run up the score – did you pay attention to Baylor today? I'm saying that the existence of the RPI doesn't act as a deterrent.

RPI is very crappy early in the season, which is one of the reasons they don't release it too early, and gets less crappy as the season goes along but it is always deficient compared to most other computer or human-based rankings. It ought to be scrapped.


The margin of victory(MOV) is necessary in football because of the limited number of games (12). With 30 or more games in basketball there is sufficient data to make reasonable rankings without using MOV. I also believe that the goal should be to pick the teams who have played the best during the season (which should be a purely objective measure) rather than the teams that are the best at the end of the season. Allowing for the latter concept means that injuries or suspensions can be considered in order to bump up or move down individual teams.

The RPI is a really bad idea. The answer is to find improved methods of initially evaluating teams that focus on the quality of teams played rather than just the records of teams. It makes zero sense to get more credit for beating a conference winner from a poor conference than a mid range team from a major conference. Even if the math seems difficult it would be reasonably easy to come up with a formula to determine the likelihood of a team winning a game over a team ranked at a certain level. While there are good arguments for evaluating that based on a 25th ranked team or a 50th ranked team (or some other specific benchmark) all teams could then be ranked by actual wins over expected wins. This would provide an objective measure for determining which teams should make the NCAA tournament and how they should be seeded.

To get really geeky, my preference would be to use the 50th (or 45th) best team to determine who gets into the tournament, but then to use a higher standard such as the 25th best teams for purposes of seeding. If the NCAA would like to make some additional rules such as a minimum allowable conference record (.500 or .400) and an overall allowable record (.501 or .550) as a first cut. The Committee's sole responsibility would be to determine actual matchups.


pilight



Joined: 23 Sep 2004
Posts: 66773
Location: Where the action is


Back to top
PostPosted: 12/19/19 9:50 am    ::: Reply Reply with quote

They do have a minimum overall record limit of .500 for at-large teams.

The best ratings systems use MOV but with a diminishing return. The difference between winning by 10 and winning by 20 is much greater than the difference between winning by 30 and winning by 40.



_________________
Let us not deceive ourselves. Our educational institutions have proven to be no bastions of democracy.
Display posts from previous:   
Post new topic   Reply to topic    RebKell's Junkie Boards Forum Index » NCAA Women's Basketball - General Discussion All times are GMT - 5 Hours
Goto page 1, 2  Next
Page 1 of 2

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum


Powered by phpBB 2.0.17 © 2001- 2004 phpBB Group
phpBB Template by Vjacheslav Trushkin