RPI and SOS

[ciaomichael]

Since RPI and SOS have come up on several threads here (with some controversy here and there), I though it might be useful to discuss what these indexes really mean.

It is my understanding the RPI is calculated as:

1. Team winning percentage x 25%, plus

2. Opponents' winning percentage x 50%, plus

3. Opponents' opponents' winning percentage x 25%

It is my understanding the SOS is simply the sum of #2 and #3 above.

So, I don't quite get why SOS is always discussed as a separate factor in evaluating a team's strength. It's already incorporated into the RPI.

Please correct me if I'm wrong and discuss.

[wufan]

Well put! All statistical evaluators use SOS in their calculation (even the eye test). To discuss SOS independent of win/loss is pretty silly.

According to RPI, we are the #6 team in the country.

According to Warren Nolan we are the #1 team in the country.

According to BPI (ESPN) we are the #4 team in the country.

According to KenPom we are the #5 team in the country.

According to the eye test (AP/Coaches poll) we are the # 2 team in the country.

Take your pick!

[Jamar Howard 4 President]

Yes RPI incorporates SOS into its formula. However, does it factor it in too heavily? Too lightly? Those issues are not agreed upon by everyone, so it often helps to back up and talk about SOS, W/L record, etc. separately.

[wufan]

Arguments about whether or not the SOS or W/L are weighted correctly has nothing to do with arguing about if SOS matters. It's a part of the formula, not a characteristic of the team's ability.

[1979Shocker]

The RPI also factors in whether the game is home, away, or neutral, too.

[ciaomichael]

Well, just my own opinion, but I think the RPI should be a multiplicative formula rather than additive. In other words, you have to actually beat a team to pick up the benefit of the other team's winning percentage. Sure, you could schedule and play a bunch of teams with a good winning percentage, but if you don't beat them, why should you get the benefit of their wins?

[The Mad Hatter]

Yes RPI incorporates SOS into its formula. However, does it factor it in too heavily? Too lightly? Those issues are not agreed upon by everyone, so it often helps to back up and talk about SOS, W/L record, etc. separately.

Since SOS is 75% of RPI, I can't see a strong argument for it being factored too lightly.

[Jamar Howard 4 President]

Since SOS is 75% of RPI, I can't see a strong argument for it being factored too lightly.

It is more complicated than you think. SOS as a decimal number (.4876, .5678, etc.) does not vary nearly as much as winning % (below 0.1000 all the way up to 1.0000). SOS needs to be a heavy percentage or else it would hardly factor in at all.

[richardj43]

Reminds me of a Professor I had when in college. He taught Economic Statistics and frequently would advise the class to remember that "figures don't lie, liars figure." I, for one would really like some of these guys to add factors that include "chemistry of team" as a significient factor, perhaps as a multiplier on wins and for the one and done wonder teams a negative multiplier attached to their wins. Just saying fair is fair cause team work counts which I believe is why the coaches poll has us #2 -

[The Mad Hatter]

It is more complicated than you think. SOS as a decimal number (.4876, .5678, etc.) does not vary nearly as much as winning % (below 0.1000 all the way up to 1.0000). SOS needs to be a heavy percentage or else it would hardly factor in at all.

Point taken, but since KU (who have the best SOS) have been consistently higher in RPI than in other statistical metrics, I would say that the 75% number does plenty to overcome their much higher loss total than other teams.

I think there is a good case that RPI doesn't handle SOS in a very sophisticated manner, but not that it doesn't value it enough.

[shockmonster]

It is more complicated than you think. SOS as a decimal number (.4876, .5678, etc.) does not vary nearly as much as winning % (below 0.1000 all the way up to 1.0000). SOS needs to be a heavy percentage or else it would hardly factor in at all.

I know how important the knowledge of statistical analysis is. It is literally a strong avenue to monetary success and I know that each of you that enjoys this type of discourse is a great guy. In fact, some of my best friends are mathematicians. However, besides my vision becoming blurry, my mind fogging over, getting dark circles around my eyes, and a HUGE headache, this is what I think of when I see your posts arguing about these issues.

http://www.bing.com/videos/search?q=revenge+of+the+nerds+youtube&qpvt=revenge+of+the+nerds+youtube&FORM=VDRE#view=detail&mid=F2C2511BC62A48A70CADF2C2511BC62A48A70CAD

[shox1989]

For what it is worth, the Shockers are #1 and it isn't even close in the Nolan Power Index (Warren Nolan's power index): http://warrennolan.com/basketball/2014/npi

In my humble opinion, this is the only index the NCAA should use!

[ciaomichael]

It is more complicated than you think. SOS as a decimal number (.4876, .5678, etc.) does not vary nearly as much as winning % (below 0.1000 all the way up to 1.0000). SOS needs to be a heavy percentage or else it would hardly factor in at all.

True, but you can think of it this way also: you can actually lose to a team, but if that team is just 1 game above .500, you can improve your RPI in the process. (I'm ignoring opponents' opponents' winning percentage just to simplify.) That just doesn't make sense to me. How are you better than you were the day before by losing?

[Jamar Howard 4 President]

ciaomchael, I'm not sure what you are saying. If WSU lost to a team "1 game over .500", say 16-15, their RPI would not improve.

[jdmee]

I agree with JH4P, losing to team doesn't improve the RPI.

For the 30 games that count on WSU's RPI.

30-0 = 1.000 winning percentage or 0.2500 towards the RPI
29-1 = 0.966 winning percentage or 0.2417 towards the RPI.

or a drop of 0.0083 on the RPI.

Our SoS currently is 0.5329. I am assuming each opponent played 30 games each. I also am assuming the OWP and OOWP is equal to the SoS.

So 30 opponents * 30 games/opponents = 900 games.

900 * 0.5329 = 480 wins and 420 losses.

If you take the MSU game that would change the OWP to 482 and 418. (MSU was played twice)

So our OWP would then be equal to 0.5355 or a difference of 0.0026
0.0026 *.5 = 0.0013.

0.0083 - 0.0013 = 0.0070 net loss on the RPI.


But if you notice the numbers would be the same if the loss came to MSU or Loyola.