RPI and SOS

[ciaomichael]

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.

Well, bear with me for a moment. Let's assume a 30 game schedule for everyone.

By taking a loss, that would drop our RPI by .00833 (1/30 or .033333 x 25% weighting.)

But, just by playing them that would increase our RPI by .008889 (16/30 or .533333 x 1/30 x 50% weighting.)

So, yes, our RPI would show a small improvement.

[jdmee]

Well, bear with me for a moment. Let's assume a 30 game schedule for everyone.

By taking a loss, that would drop our RPI by .00833 (1/30 or .033333 x 25% weighting.)

But, just by playing them that would increase our RPI by .008889 (16/30 or .533333 x 1/30 x 50% weighting.)

So, yes, our RPI would show a small improvement.

But win or lose we would have played them. The RPI difference needs to take into account the 1 win not the 16 wins.

If you take that into account then 15/30 or .5 x 1/30 * 50% = 0.00833

So the difference in SoS would be 0.008889 - 0.00833 = 0.00056.

This is vs the win % difference of 0.00833

[ciaomichael]

But win or lose we would have played them. The RPI difference needs to take into account the 1 win not the 16 wins.

But, the RPI DOES take into account the 16 wins! It takes into account your opponents overall winning percentage and gives it twice the weight of your own winning percentage! That's the whole point of this post. Now, if team x only had a 30% winning percentage, then obviously our RPI would drop if we lost to them. That's why you sometimes see anomalies where a team loses a game but their RPI goes up.

[shockmonster]

So what you are saying is that winning percentage of opponent, home win, road win, affects one's RPI. But bad losses don't affect the RPI?

[jdmee]

But, the RPI DOES take into account the 16 wins! It takes into account your opponents overall winning percentage and gives it twice the weight of your own winning percentage! That's the whole point of this post. Now, if team x only had a 30% winning percentage, then obviously our RPI would drop if we lost to them. That's why you sometimes see anomalies where a team loses a game but their RPI goes up.

We might be talking different points, but look at it this way, playing and beating a 15-14 (making them 15-15) team will help the RPI more than playing and losing to a 15-14 (making them 16-14) team.


The impact of the win/loss is .00833 (1/30 x 25% weighting.)

The impact of win/loss in our game to the SoS is .00056 (1/30 x 1/30 x 50% weighting) Our 1 game is only 1 of the 30 they played.

[jdmee]

So what you are saying is that winning percentage of opponent, home win, road win, affects one's RPI. But bad losses don't affect the RPI?

Yes, according to the RPI who you lose to does not matter. However, it does matter when talking about NCAA seeding, ranking, etc.

[Jamar Howard 4 President]

The above posts have my head spinning. Here's a simple explanation:

RPI =
1) 25% your record
+
2) 50% opponents record
+
3) 25% opponents opponents record

If you have played a good schedule and both #2 & #3 are good (say 0.600), then a game against a 16-15 team (0.516) who has played a bunch of 16-15 teams will drop #2 and #3. If you lose, #1 drops too.

All 3 components dropping obviously means your RPI drops. Simple as that. A loss to a 16-15 team in that scenario drops your RPI.

Now if your SOS has been really weak, a 16-15 team could potentially provide enough of a boost to overcome the hit to your record by losing, but that is a fairly rare case that would never come into play unless you were playing the weakest of schedules.

[Heinro]

The real issue with the SOS is that it completely ignores road/home. The road/home variant is only calculated in the 25% that is your own personal record. The SOS is purely 66% your opponents record and 33% your opponents opponents record. Thus, if 2 teams had the exact same schedule and one played all games on the road and one played all games at home, the SOS would still be the same even though the RPI's would be drastically different. However, logically, there is no doubt the team that had all road games would actually have had a much more difficult schedule.

In my opinion, the SOS is purely purposed to help inflate big schools who refuse to play road games. If for instance, you look at Southern Miss, they have an SOS of 154 which is absolutely awful for anyone considered bubble or better; however, they have an RPI of 39. Their RPI which is solid, takes into account that they played 7 road/neutral games in the non-con, the SOS entirely ignores it.

Thus, from my perspective, all the SOS does is provide a convenient out to discount road/neutral games.

[MadaboutWu]

Excellent post. Hadn't thought of that angle before.

[Jamar Howard 4 President]

One team on your schedule might have played a bunch of home games. Another team on your schedule might have played a bunch of road games. In the end, it usually evens out. Unless coaches are secretly filling their schedule with teams who are home-bodies (the Syracuses of the world), then I doubt the lack of home and road factor in the SOS really makes much of a difference. It's built into the 25% (your own team's record) that really matters.

I'd love to be proved wrong though. Somebody go find me an example of a team that has stocked up on opponents who played mostly home games and I'll analyze just how much of a benefit they gained from it.

[ShockTalk]

The real issue with the SOS is that it completely ignores road/home. The road/home variant is only calculated in the 25% that is your own personal record. The SOS is purely 66% your opponents record and 33% your opponents opponents record. Thus, if 2 teams had the exact same schedule and one played all games on the road and one played all games at home, the SOS would still be the same even though the RPI's would be drastically different. However, logically, there is no doubt the team that had all road games would actually have had a much more difficult schedule.

In my opinion, the SOS is purely purposed to help inflate big schools who refuse to play road games. If for instance, you look at Southern Miss, they have an SOS of 154 which is absolutely awful for anyone considered bubble or better; however, they have an RPI of 39. Their RPI which is solid, takes into account that they played 7 road/neutral games in the non-con, the SOS entirely ignores it.

Thus, from my perspective, all the SOS does is provide a convenient out to discount road/neutral games.

Why does the SOS need to factor in H/N/A? SOS is simply about who you played and their opponents. Where you play them is factored in with the AWP part of the RPI. If you added some sort of factor to the SOS, they'd just reduce the factor in the AWP portion or take it away all together. If they have factors of 1.4 and 0.6 in the AWP, they'd just change them to 1.2 and 0.8 (or something like that for example) if used in the SOS part of the RPI calculation.

If you believe the H/A is not a big enough part of the calculation, that's fine. However, having the H/A be a bigger factor in the RPI can also be achieved by increasing the effect in the AWP without any need to make a change in the SOS.

My opinion is only based on my limited knowledge of how the RPI works, so forgive me if I'm missing something.

[ciaomichael]

Also, a thing to remember about the RPI is that it's not meant to be a predictor of future performance. It's simply a statistic that summarizes a team's resume at a given point in time. Most statistical predictors take into account factors like margin of victory, points scored, points scored against, etc. Since the NCAA doesn't want to be in the business of encouraging teams to run up scores (possibly inviting influence by gambling interests), it avoids the use of statistical predictors in things like tournament selection.

[MadaboutWu]

One team on your schedule might have played a bunch of home games. Another team on your schedule might have played a bunch of road games. In the end, it usually evens out. Unless coaches are secretly filling their schedule with teams who are home-bodies (the Syracuses of the world), then I doubt the lack of home and road factor in the SOS really makes much of a difference. It's built into the 25% (your own team's record) that really matters.

I'd love to be proved wrong though. Somebody go find me an example of a team that has stocked up on opponents who played mostly home games and I'll analyze just how much of a benefit they gained from it.

I think you misunderstood Heinro's point. It isn't that a teams' opponents play more or fewer games at home or not; it is that a team does skewing their SOS. Since the odds of winning at home are dramatically higher you have scheduled easier the more home games you play.

[MadaboutWu]

Why does the SOS need to factor in H/N/A? SOS is simply about who you played and their opponents. Where you play them is factored in with the AWP part of the RPI. If you added some sort of factor to the SOS, they'd just reduce the factor in the AWP portion or take it away all together. If they have factors of 1.4 and 0.6 in the AWP, they'd just change them to 1.2 and 0.8 (or something like that for example) if used in the SOS part of the RPI calculation.

If you believe the H/A is not a big enough part of the calculation, that's fine. However, having the H/A be a bigger factor in the RPI can also be achieved by increasing the effect in the AWP without any need to make a change in the SOS.

My opinion is only based on my limited knowledge of how the RPI works, so forgive me if I'm missing something.

The problem is that SOS is used as a factor independent of RPI as well as a component of rpi. Teams that play more than average games at home are given credit for scheduling tougher than they actually have when SOS is broken out. RPI has accounted for the bias but SOS used independently has not.

[Jamar Howard 4 President]

The problem is that SOS is used as a factor independent of RPI as well as a component of rpi. Teams that play more than average games at home are given credit for scheduling tougher than they actually have when SOS is broken out. RPI has accounted for the bias but SOS used independently has not.

Bingo. Good point. Thanks for clarifying.