Legit rankings without preseason ranking consideration

[Great Dayne]

NCAA College Basketball RPI Rankings - ESPN

Finally, a ranking system that doesn't use preseason ranking predictions on how good a team will potentially be and just based on SOS and other stats.

RK TEAM RPI D1 W-L SOS NCRP NCSS CFRP CFSS 1-25 26-50 51-100 L12 LRPI OFFQ DEFQ ASM
1 Wisconsin .7190 12-0 0.66 13 16 43 243 0-0 3-0 6-0 12-0 5 7.7 13.0 20.7
2 Massachusetts .7100 9-0 0.66 15 18 57 215 2-0 2-0 1-0 9-0 14 14.6 0.9 15.6
3 Syracuse .6976 9-0 0.63 18 28 42 289 1-0 0-0 3-0 9-0 1 9.5 11.7 21.2
4 Baylor .6971 6-1 0.73 226 2 124 225 1-1 1-0 3-0 6-1 160 -- -- --
5 Kansas .6922 7-3 0.78 65 1 69 137 1-2 0-1 2-0 7-3 16 9.2 9.9 19.1
6 Colorado .6834 10-1 0.65 25 23 71 345 1-1 2-0 1-0 10-1 33 7.6 12.1 19.7
7 Villanova .6830 9-0 0.59 16 59 35 303 2-0 0-0 1-0 9-0 6 13.8 10.7 24.5
8 Oregon .6824 9-0 0.61 17 48 54 108 0-0 3-0 1-0 9-0 2 18.0 3.7 21.7
9 Oklahoma State .6804 9-1 0.62 28 36 68 294 0-0 2-1 2-0 9-1 4 18.4 10.4 28.9
10 Arizona .6739 11-0 0.59 19 58 56 334 1-0 1-0 2-0 11-0 3 7.0 17.3 24.3

 

[TrollyMcTroller]

NCAA College Basketball RPI Rankings - ESPN

Finally, a ranking system that doesn't use preseason ranking predictions on how good a team will potentially be and just based on SOS and other stats.

Don't get me wrong... I'm sure you have the most noble of intentions, but the RPI is not the rating you are looking for.

The RPI has it's place and is good at what it does, but it doesn't rank the best teams. Instead it evaluates resumes which is why the NCAA selection committee relies on it so heavily. The tournament selection process (contrary to popular belief) is less about how good you are and more about what you've actually done.

Joe Lunardi and his team of Bracketologists have replicated the Ratings Percentage Index (RPI), one of the tools used by the NCAA Selection Committee to evaluate teams. Lunardi & Co. update it daily, and have included additional ratings, providing readers with original data for every Division I team, available only on ESPN.com.

Kudos by the way, to Lunardi and his crack team of math wizards. a 4th grader can do the math for the RPI. Don't believe me? Here it is...

The RPI is made up of 3 components:
Your winning percentage makes up 25% of the RPI and is weighted by location
Your opponent's winning percentage (ignoring the game(s) against you) makes up 50%
Your opponent's opponent's winning percentage (ignoring games against your opponents) makes up the remaining 25%

The weighting of game location works as follows: Neutral court wins and losses aren't weighted. A home win, is multiplied by 0.6, a home loss by 1.4, an away loss by 0.6, and an away win by 1.4 (that means a win on the road is worth more than twice a win at home. (vice versa for losses)

So with all of that being said, using the RPI in place of the polls isn't inherently an awful idea, but you need to understand that they aren't really trying to measure the same things, and that the RPI is usually a trainwreck until a couple of weeks into conference play. The reason being is that the level of competition is usually pretty lopsided, and winning percentages simply don't tell the whole story.

Ohio State is a good example to illustrate my overall point. Most people would agree that Ohio State is a pretty damn good team this season. Most would also agree that they've played a shit schedule to this point. That is why they aren't very high in the RPI but they are in the polls (and many other rating systems)

The two main go-to ratings that people will typically point you to are Sagarin and Pomeroy. Neither rely on preseason rankings* and both are used by the gambling community which should probably tell ya that they have their stuff together pretty well. Sagarin is directly attempting to measure team strength based on game outcomes. He has different formulas based on score, and pure win/loss and his overall rating is a composite of those. Pomeroy takes a different approach and ignores wins and losses and instead is looking at offensive and defensive efficiency (points scored per possession and points allowed per possession) which as it turns out also ends up being a pretty good indicator of overall team strength.

*IIRC Sagarin does use the preseason rankings in some manner for his initial ratings but after a few weeks, and enough teams have played each other the data becomes "connected" enough that the preseason rankings are completely ignored.

 

[TrollyMcTroller]

Here's the current stats from those two site's if you're interested:

Pomeroy
1   Louisville
2   Ohio St.
3   Oklahoma St.
4   Villanova
5   Arizona
6   Kansas
7   Syracuse
8   Wisconsin
9   Pittsburgh
10  North Carolina

Sagarn
1   Oklahoma State
2   Arizona
3   Ohio State
4   Villanova
5   Iowa State
6   Iowa
7   Wisconsin
8   Louisville
9   Kansas
10  Wichita State

 

[jonvi]

It's waaaaaay to early for any poll to be worth a damn.

 

[Great Dayne]

Don't get me wrong... I'm sure you have the most noble of intentions, but the RPI is not the rating you are looking for.

The RPI has it's place and is good at what it does, but it doesn't rank the best teams. Instead it evaluates resumes which is why the NCAA selection committee relies on it so heavily. The tournament selection process (contrary to popular belief) is less about how good you are and more about what you've actually done.



Kudos by the way, to Lunardi and his crack team of math wizards. a 4th grader can do the math for the RPI. Don't believe me? Here it is...

The RPI is made up of 3 components:
Your winning percentage makes up 25% of the RPI and is weighted by location
Your opponent's winning percentage (ignoring the game(s) against you) makes up 50%
Your opponent's opponent's winning percentage (ignoring games against your opponents) makes up the remaining 25%

The weighting of game location works as follows: Neutral court wins and losses aren't weighted. A home win, is multiplied by 0.6, a home loss by 1.4, an away loss by 0.6, and an away win by 1.4 (that means a win on the road is worth more than twice a win at home. (vice versa for losses)

So with all of that being said, using the RPI in place of the polls isn't inherently an awful idea, but you need to understand that they aren't really trying to measure the same things, and that the RPI is usually a trainwreck until a couple of weeks into conference play. The reason being is that the level of competition is usually pretty lopsided, and winning percentages simply don't tell the whole story.

Ohio State is a good example to illustrate my overall point. Most people would agree that Ohio State is a pretty damn good team this season. Most would also agree that they've played a shit schedule to this point. That is why they aren't very high in the RPI but they are in the polls (and many other rating systems)

The two main go-to ratings that people will typically point you to are Sagarin and Pomeroy. Neither rely on preseason rankings* and both are used by the gambling community which should probably tell ya that they have their stuff together pretty well. Sagarin is directly attempting to measure team strength based on game outcomes. He has different formulas based on score, and pure win/loss and his overall rating is a composite of those. Pomeroy takes a different approach and ignores wins and losses and instead is looking at offensive and defensive efficiency (points scored per possession and points allowed per possession) which as it turns out also ends up being a pretty good indicator of overall team strength.

*IIRC Sagarin does use the preseason rankings in some manner for his initial ratings but after a few weeks, and enough teams have played each other the data becomes "connected" enough that the preseason rankings are completely ignored.

After reviewing all the systems and how they rank I would still consider RPI more legit as it takes the main factors. Win margin is flawed due to outcomes of games final scores can be skewed with FT's in the end, playing second stringers, etc. There's always games where the final score doesn't truly indicate how close the game was actually played. Also, not taking wins and loses? You may as well not call that a ranking system at all. That's why I see significant outliers in his ranking with two teams having 3 loses in the top 11. A system that includes all of these factors SOS, location of game, W%, total Wins, efficiency, margin of victory, etc might be what the doctor ordered.

 

[Great Dayne]

It's waaaaaay to early for any poll to be worth a damn.

Agreed, the most legit polls will be in week 3 or 4 in conference play.

 

[Jim Rome is Flaming]

NCAA College Basketball RPI Rankings - ESPN

Finally, a ranking system that doesn't use preseason ranking predictions on how good a team will potentially be and just based on SOS and other stats.

RK TEAM RPI D1 W-L SOS NCRP NCSS CFRP CFSS 1-25 26-50 51-100 L12 LRPI OFFQ DEFQ ASM
1 Wisconsin .7190 12-0 0.66 13 16 43 243 0-0 3-0 6-0 12-0 5 7.7 13.0 20.7
2 Massachusetts .7100 9-0 0.66 15 18 57 215 2-0 2-0 1-0 9-0 14 14.6 0.9 15.6
3 Syracuse .6976 9-0 0.63 18 28 42 289 1-0 0-0 3-0 9-0 1 9.5 11.7 21.2
4 Baylor .6971 6-1 0.73 226 2 124 225 1-1 1-0 3-0 6-1 160 -- -- --
5 Kansas .6922 7-3 0.78 65 1 69 137 1-2 0-1 2-0 7-3 16 9.2 9.9 19.1
6 Colorado .6834 10-1 0.65 25 23 71 345 1-1 2-0 1-0 10-1 33 7.6 12.1 19.7
7 Villanova .6830 9-0 0.59 16 59 35 303 2-0 0-0 1-0 9-0 6 13.8 10.7 24.5
8 Oregon .6824 9-0 0.61 17 48 54 108 0-0 3-0 1-0 9-0 2 18.0 3.7 21.7
9 Oklahoma State .6804 9-1 0.62 28 36 68 294 0-0 2-1 2-0 9-1 4 18.4 10.4 28.9
10 Arizona .6739 11-0 0.59 19 58 56 334 1-0 1-0 2-0 11-0 3 7.0 17.3 24.3

I personally hate the RPI, it always seems to have certain teams 15-20 spots over and underranked. Here's how I'd have it.

1: Arizona
2: Syracuse
3: Wisconsin
4: Villanova
5: Ohio State
6: Oregon
7: Iowa State
8: Mizzouri
9: Oklahoma State
10: Massachusetts

 

[TrollyMcTroller]

After reviewing all the systems and how they rank I would still consider RPI more legit as it takes the main factors. Win margin is flawed due to outcomes of games final scores can be skewed with FT's in the end, playing second stringers, etc. There's always games where the final score doesn't truly indicate how close the game was actually played. Also, not taking wins and loses? You may as well not call that a ranking system at all. That's why I see significant outliers in his ranking with two teams having 3 loses in the top 11. A system that includes all of these factors SOS, location of game, W%, total Wins, efficiency, margin of victory, etc might be what the doctor ordered.

Sagarin is the closest thing out there. His formula isn't made public, but there is a lot known about his methodology. He doesn't use efficiency stats, but he takes wins and losses into the equation. SOS is a little bit tricky. (regardless of what system you're using) Generally speaking, the SOS is a result of the ratings, not a direct input. That's not to say the ratings don't factor in who you've played, but it doesn't make much sense to rate your overall competition until you actually know how good they are. Many of them use recursion to accomplish that... Everyone starts off with an equal rating... everyone is evaluated based on their performance... now with a new baseline... teams are reevaluated based on the new weighting of their competition... repeat for x number of iterations, or until the variation between iterations drops below a certain threshold.

Realisticly though, you're never going to find a computer rating that you agree with 100%. There is just no replacement for the eyeball test. And flawed as the Polls may be (mainly do to a lack of clear voting guidelines, and an unhealthy dose of confirmation bias) the results seem more agreeable than any of the computer ratings I've seen.

Another rating you might like that I didn't mention earlier is the BPI on ESPN. But they don't have stats available until later in the season.

BPI -The College Basketball Power Index explained - ESPN

 

[Smart]

Meh...RPI heavily rewards teams for beating the best cupcakes. This year, we have been the team that scheduled good cupcakes (ie Eastern Kentucky, Milwaukee, Green Bay) and solid but unelite teams (SLU, Marquette, West Virginia, Virginia, etc.). We've really only beaten one very good team (Florida), and it was a close home win when some of their best players were suspended.

Ohio State, Villanova, Arizona, and Syracuse all have looked better than us to me. I'd have us #5.

 

[Great Dayne]

Meh...RPI heavily rewards teams for beating the best cupcakes. This year, we have been the team that scheduled good cupcakes (ie Eastern Kentucky, Milwaukee, Green Bay) and solid but unelite teams (SLU, Marquette, West Virginia, Virginia, etc.). We've really only beaten one very good team (Florida), and it was a close home win when some of their best players were suspended.

Ohio State, Villanova, Arizona, and Syracuse all have looked better than us to me. I'd have us #5.

OSU's schedule is two notches below the Badgers. AZ quality wins are vs a mid level Big Ten team with 4 L's and a struggling Duke with 3 L's. I can agree with Cuse and Nova though.

 

[Arizona_Sting]

Really going by RPI this early in the year...?

 

[Arizona_Sting]

OSU's schedule is two notches below the Badgers. AZ quality wins are vs a mid level Big Ten team with 4 L's and a struggling Duke with 3 L's. I can agree with Cuse and Nova though.

How convenient of you to leave out a road win vs. a ranked SDSU and btw Duke doesn't have three loses and they aren't "struggling"... ranked 8th right now and their only two losses are to Arizona and Kansas.

Drexel and NMSU are both teams projected to go to the NCAA's as well. At least be informed on what you are typing before making yourself sound foolish.

 

[Great Dayne]

How convenient of you to leave out a road win vs. a ranked SDSU and btw Duke doesn't have three loses and they aren't "struggling"... ranked 8th right now and their only two losses are to Arizona and Kansas.

Drexel and NMSU are both teams projected to go to the NCAA's as well. At least be informed on what you are typing before making yourself sound foolish.

I never said they were not projected to be in the tourney. However believing that 4 teams AZ has played to UW's 7 teams (projected NCAA tournament) is a superior schedule is the epitome of foolish. Duke lost to Kansas (3 L's) at HOME, they escaped vs Vermont at home by 1pt which should have been a loss, and have no current wins vs teams in the top 25 or even a team that's ranked in the top 35 according to the BPI. I'm not saying that's a terrible schedule by any stretch of the imagination, I just believe it's slightly below the Badgers.

 

[element1286]

Any rating system that doesn't use MOV isn't worth a damn, look at Kenpom, Massey, or Sagarin. RPI is useless.

 

[TrollyMcTroller]

I never said they were not projected to be in the tourney. However believing that 4 teams AZ has played to UW's 7 teams (projected NCAA tournament) is a superior schedule is the epitome of foolish. Duke lost to Kansas (3 L's) at HOME, they escaped vs Vermont at home by 1pt which should have been a loss, and have no current wins vs teams in the top 25 or even a team that's ranked in the top 35 according to the BPI. I'm not saying that's a terrible schedule by any stretch of the imagination, I just believe it's slightly below the Badgers.

Duke lost to Kansas in New York City, not at home.

Duke won at home vs Vermont. It may have been ugly, but it was a win. you can try to rationalize that it should have been a loss, but at the end of the day, it's a win on the resume, just like every other W.

 

[Doctor Krieger]

Did the op just recently find out about rpi? Lmao

 

[The Derski]

Any rating system that doesn't use MOV isn't worth a damn, look at Kenpom, Massey, or Sagarin. RPI is useless.

What is MOV?

 

[Monte Carlo Wildcat]

I love me some Massey ratings.

 

[TrollyMcTroller]

What is MOV?

Margin Of Victory

 

[geneh_33]

If you have North Carolina ranked you have a good poll. This is always the best indicator as to whether a poll is any good or not.