Gaming the RPI?

[longtimefan]

I think we have known this intuitively, but Luke Winn--who does good work--has compiled some interesting numbers to support the position.

http://sportsillustrated.cnn.com/20.../27/schedule-strength/index.html?sct=cb_wr_a2

 

[reedny]

It's a good article .. 75% of the RPI is based on SOS, and two thirds of that is based on a raw winning percentage for non-conference games. I think he's right that Kenpom's adjusted formula for non-conference SOS is better: http://kenpom.com/index.php?s=RankNCSOSPythag

 

[Room 914]

None the less. If your Pitt or Syracuse you should be playing good teams out of conference. It makes your team better. I would care more about that then the RPI.

Check Pitts non conference...it's a joke. Not in agreement that you should game the system. You should play good teams.

 

[Knicks411]

None the less. If your Pitt or Syracuse you should be playing good teams out of conference. It makes your team better. I would care more about that then the RPI.

Check Pitts non conference...it's a joke. Not in agreement that you should game the system. You should play good teams.

Except their goal is going to be to make the tournament and get a good seed, and for as long as the committee looks at the RPI, you gotta keep that in mind.

 

[AlpharettaCuse]

It's a good article .. 75% of the RPI is based on SOS, and two thirds of that is based on a raw winning percentage for non-conference games. I think he's right that Kenpom's adjusted formula for non-conference SOS is better: http://kenpom.com/index.php?s=RankNCSOSPythag

I agree that the RPI is too simplistic, however RPI did a better job ranking the teams than kenpom did. No way was Wisconsin a Top 5 team (ahead of Syracuse and North Carolina), and no way Memphis or Wichita State were Top 10.

 

[Knicks411]

I agree that the RPI is too simplistic, however RPI did a better job ranking the teams than kenpom did. No way was Wisconsin a Top 5 team (ahead of Syracuse and North Carolina), and no way Memphis or Wichita State were Top 10.

And I don't think Colorado State was one of the 30 best teams in the country either. No ranking system is perfect; I'd take Pomeroy over the RPI on the whole every day of the week though.

 

[longtimefan]

I agree that the RPI is too simplistic, however RPI did a better job ranking the teams than kenpom did. No way was Wisconsin a Top 5 team (ahead of Syracuse and North Carolina), and no way Memphis or Wichita State were Top 10.

You're comparing apples and oranges. The article was about NCSOS, and how that metric is badly measured by the traditional RPI as opposed to Ken Pomeroy's method.

You are referring to his overall team ratings, of which NCSOS is only a small part.

 

[AlpharettaCuse]

You're comparing apples and oranges. The article was about NCSOS, and how that metric is badly measured by the traditional RPI as opposed to Ken Pomeroy's method.

You are referring to his overall team ratings, of which NCSOS is only a small part.

Right. But what's the point of the rankings? The NCAA uses RPI, and it appears it does a better job ranking the teams correctly.

 

[AlpharettaCuse]

And I don't think Colorado State was one of the 30 best teams in the country either. No ranking system is perfect; I'd take Pomeroy over the RPI on the whole every day of the week though.

Agree, Colorado State was not a Top 30 team either, and agree that no ranking system is perfect. But why Pomeroy over RPI?

 

[Knicks411]

Agree, Colorado State was not a Top 30 team either, and agree that no ranking system is perfect. But why Pomeroy over RPI?

I think the RPI is too simplistic. It doesn't take margin of victory into account (I grant you MOV isn't perfect, there are sometimes when teams let up at the end of a game when they are winning by a lot, etc, but if 2 teams play a common opponent and one team beats them by 20 and the other team beats them by 2, then I'm going to assume the team that won by 20 is better, RPI says differently). It has a rather crude way of dealing with the home/road issue. I'd rather have more information put into the pot than less. I'm not sure Pomeroy is better than Sagarin, or the ESPN system they introduced last year, but I b et all 3 are better than RPI. But to me, the biggest thing is the MOV. There's no reason to throw that info out; it's extremely predictive.

(trying to think; is there a way we could test this?)

 

[AlpharettaCuse]

I think the RPI is too simplistic. It doesn't take margin of victory into account (I grant you MOV isn't perfect, there are sometimes when teams let up at the end of a game when they are winning by a lot, etc, but if 2 teams play a common opponent and one team beats them by 20 and the other team beats them by 2, then I'm going to assume the team that won by 20 is better, RPI says differently). It has a rather crude way of dealing with the home/road issue. I'd rather have more information put into the pot than less. I'm not sure Pomeroy is better than Sagarin, or the ESPN system they introduced last year, but I b et all 3 are better than RPI. But to me, the biggest thing is the MOV. There's no reason to throw that info out; it's extremely predictive.

(trying to think; is there a way we could test this?)

I totally agree that RPI is too simple, and I think MOV should be used (although teams may run up the score) and that road wins and neutral game wins should count for more than home wins (doesn't RPI do this?). Sagarin seems more accurate than kenpom, comparing to the 2nd to last Top 25 polls and final polls, but RPI looks to be the most accurate. I'm all for more information, but if that information concludes things that we know are not really true (Wisconsin > Syracuse as an example) it's hard to find credible. On the other hand, even the final polls are not always accurate in my opinion. Syracuse was the 2nd best team in the country last year, beat Louisville twice, and was still ranked behind them, despite 7 fewer losses. Guess it's just a reflection of the best teams in the country at that particular time, and not all season.

 

[longtimefan]

I totally agree that RPI is too simple, and I think MOV should be used (although teams may run up the score) and that road wins and neutral game wins should count for more than home wins (doesn't RPI do this?). Sagarin seems more accurate than kenpom, comparing to the 2nd to last Top 25 polls and final polls, but RPI looks to be the most accurate. I'm all for more information, but if that information concludes things that we know are not really true (Wisconsin > Syracuse as an example) it's hard to find credible. On the other hand, even the final polls are not always accurate in my opinion. Syracuse was the 2nd best team in the country last year, beat Louisville twice, and was still ranked behind them, despite 7 fewer losses. Guess it's just a reflection of the best teams in the country at that particular time, and not all season.

USAToday/coaches automatically rate the Final Four teams 1 thru 4 (there may have been exceptions to this--maybe VCU two seasons ago?).

 

[reedny]

I agree that the RPI is too simplistic, however RPI did a better job ranking the teams than kenpom did. No way was Wisconsin a Top 5 team (ahead of Syracuse and North Carolina), and no way Memphis or Wichita State were Top 10.

I agree with this... but it's not because of the NCSOS.

 

[reedny]

I think the RPI is too simplistic. It doesn't take margin of victory into account (I grant you MOV isn't perfect, there are sometimes when teams let up at the end of a game when they are winning by a lot, etc, but if 2 teams play a common opponent and one team beats them by 20 and the other team beats them by 2, then I'm going to assume the team that won by 20 is better, RPI says differently). It has a rather crude way of dealing with the home/road issue. I'd rather have more information put into the pot than less. I'm not sure Pomeroy is better than Sagarin, or the ESPN system they introduced last year, but I b et all 3 are better than RPI. But to me, the biggest thing is the MOV. There's no reason to throw that info out; it's extremely predictive.

(trying to think; is there a way we could test this?)

I'll let you do the regression analysis (MOV as the independent variable?)

 

[Knicks411]

I totally agree that RPI is too simple, and I think MOV should be used (although teams may run up the score) and that road wins and neutral game wins should count for more than home wins (doesn't RPI do this?). Sagarin seems more accurate than kenpom, comparing to the 2nd to last Top 25 polls and final polls, but RPI looks to be the most accurate.

The polls are probably better than the RPI, but I don't think the polls are all that good either.

I'm all for more information, but if that information concludes things that we know are not really true (Wisconsin > Syracuse as an example) it's hard to find credible.

The Wisconsin one was weird; the Pomeroy ratings had them really high all year; I thinl they killed a few bad teams that really boosted their rating. (Though to be fair, Wisconsin did play us in the tournament at a site that favored us and had a chance to beat us on the last shot. Of course we were missing Fab, obviously. Just throwing it out there)

But no system is perfect. Plus there's no real independent way to rank the teams, and its not like we can have everyone play a best of seven against everyone and figure out who the best really is. The RPI will have some bad ones as well. Probably the best would be some sort of weighting of many systems, which I believe someone (Moqui?) has linked to in tthe past.

 

[AlpharettaCuse]

The polls are probably better than the RPI, but I don't think the polls are all that good either.



The Wisconsin one was weird; the Pomeroy ratings had them really high all year; I thinl they killed a few bad teams that really boosted their rating. (Though to be fair, Wisconsin did play us in the tournament at a site that favored us and had a chance to beat us on the last shot. Of course we were missing Fab, obviously. Just throwing it out there)

But no system is perfect. Plus there's no real independent way to rank the teams, and its not like we can have everyone play a best of seven against everyone and figure out who the best really is. The RPI will have some bad ones as well. Probably the best would be some sort of weighting of many systems, which I believe someone (Moqui?) has linked to in tthe past.

According to Kenpom, Memphis and Wichita State were Top 10. That's absurd. All anyone had to do was watch them play. That's why polls are better. Pollsters account for things like road wins and how impressive a team plays (mov) and doesn't just look at numbers.

While no system is perfect, RPI seems like a pretty accurate tool, used in conjunction with common sense (polls).

 

[AlpharettaCuse]

The polls are probably better than the RPI, but I don't think the polls are all that good either.



The Wisconsin one was weird; the Pomeroy ratings had them really high all year; I thinl they killed a few bad teams that really boosted their rating. (Though to be fair, Wisconsin did play us in the tournament at a site that favored us and had a chance to beat us on the last shot. Of course we were missing Fab, obviously. Just throwing it out there)

But no system is perfect. Plus there's no real independent way to rank the teams, and its not like we can have everyone play a best of seven against everyone and figure out who the best really is. The RPI will have some bad ones as well. Probably the best would be some sort of weighting of many systems, which I believe someone (Moqui?) has linked to in tthe past.

The fact we beat Wisconsin without our Starting center reinforces that Syracuse was a much better team than Wisconsin, in my opinion.

 

[skurey]

Wisconsin's style of play is always highly ranked by KenPom

 

[Knicks411]

Wisconsin's style of play is always highly ranked by KenPom

Yeah I guess its because a blowout in a low possession game is going to look more impressive because of the fewer opportunities.

I remember looking at this last year, th ey beat Wofford by 36 points in a 55 possession game and Missouri Kansas City by 46 points in a 62 possession game. Both are really hard to do, but I feel like that may have juiced their rating a bit.

 

[Orangemen]

I think the RPI is too simplistic. It doesn't take margin of victory into account (I grant you MOV isn't perfect, there are sometimes when teams let up at the end of a game when they are winning by a lot, etc, but if 2 teams play a common opponent and one team beats them by 20 and the other team beats them by 2, then I'm going to assume the team that won by 20 is better, RPI says differently). It has a rather crude way of dealing with the home/road issue. I'd rather have more information put into the pot than less. I'm not sure Pomeroy is better than Sagarin, or the ESPN system they introduced last year, but I b et all 3 are better than RPI. But to me, the biggest thing is the MOV. There's no reason to throw that info out; it's extremely predictive.

(trying to think; is there a way we could test this?)

We beat Louisville by 9, PC beat them by about 30. Which team was better?

At the end of the day, all these polls are flawed, and relying on only one is the real problem.

 

[Knicks411]

We beat Louisville by 9, PC beat them by about 30. Which team was better?

Agreed, that's why you want lots and lots of information. 35 games is gonna tell you a lot more than 1.

At the end of the day, all these polls are flawed, and relying on only one is the real problem.

No question. Especially in college sports, where teams play such different schedules and there can be huge variations in talent from one team to another.

 

[kirbivore]

I remember back in the mid 2000's the MVC made a concerted effort to game the RPI system.

http://www.washingtonpost.com/wp-dyn/content/article/2006/03/04/AR2006030401283.html

 

[kuethstheman]

Reminds me that last year Gottleib claimed that our RPI benefited from a deficiency in the RPI calculation. That we played teams we were almost sure to beat around 100-150. Other top teams had played teams 200-250 and their RPI took a hit that ours didn't. He said that for top teams, any opponent over 100 should count the same. If you did that, our RPI wouldn't be as good as lots of highly-ranked teams while our actual RPI was coming out higher.

I'm not sure I agree that those games were gimmes like a 200-level team would be but I see the point. You probably shouldn't get that much more credit for beating teams rated at 100 vs. 200. At least, not when you're in the top-20, yourself.

 

[Knicks411]

Yeah it's a good point by Gottleib, and its part of the reason our SOS have gone up.

To look around the midpoint, teams around 130 RPI were Vermont, Detroit, Richmond. (LOL @ 2 of those teams being the teams we had our worst NCAA losses against). If the 2012 version of SU played those teams at home 10 times, we're gonna win 9 or 10, I'm thinking. SO I agree there isn't much difference between that and say a 200 team.

Now if you aren't an elite team, then the equaation changes, or if the games are on the road, of course. But how many home games have the good SU teams lost to teams like that over the past 8-10 years? Not a whole lot.