AAC and new Big East - power conference or mid-major

[Icebear]

The definition of an expert is a drip under pressure, or anyone 100 miles from home.

 

[Fightin Choke]

When I look at rankings like Sagarin, I see that they use a) record against 0-25 ranked teams and 0-50 ranked teams. This is a fine measure for men where the 25th best team can and has won a NC but in wcbb you can't equate a top 3 win to a win against a team ranked 23-25. It's not unusual to see a top 3 team lose to a 20th ranked team in mcbb. This is the same as Louisville (5 seed) beating Baylor last year. And Sagarin would score a win against Louisville as equal to beating Baylor. I don't see it. And I think that not having a team in the top 16 precludes a conference from being considered a "power conference".
.
My problem with RPI is that 3/4 of the rank is determined not by who you beat. It's about who are the teams you played and who those teams played. Wins & losses and the margins of those games are hardly considered.

Sagarin does not employ 'record vs. top 25' and 'record vs. top 50' when he computes his rankings. He first computes his rankings and then he determines each team's record against the top 25 teams, and the top 50 teams (just for you to see). ELO CHESS only considers which teams you have played, where those games were played, and whether or not you won the game. PREDICTOR uses all the information that ELO CHESS uses, as well as the margin of victory. His actual ranking is the average of ELO CHESS and PREDICTOR. A win against UConn helps you far more than a win over Louisville.

 

[Alydar]

Sagarin does not employ 'record vs. top 25' and 'record vs. top 50' when he computes his rankings. He first computes his rankings and then he determines each team's record against the top 25 teams, and the top 50 teams (just for you to see). ELO CHESS only considers which teams you have played, where those games were played, and whether or not you won the game. PREDICTOR uses all the information that ELO CHESS uses, as well as the margin of victory. His actual ranking is the average of ELO CHESS and PREDICTOR. A win against UConn helps you far more than a win over Louisville.

.
Thanks for the correction.

 

[RegisteredUconn]

A problem with using the mode is that it can ignore too much data. Below is hypothetical data for two conferences giving the number of teams sent to the NCAAs over a ten year period. In each case, the numbers in the series represent how many teams were invited.

Conference A: 1, 2, 1, 1, 1, 1, 1, 2, 1, 1

Conference B: 2, 5, 6, 3, 1, 7, 4, 1, 9, 8

Conference A has a mode of 2 and Conference B a mode of 1 but no rational analyst would try to make a case that A was the stronger conference.

 

[Fightin Choke]

A problem with using the mode is that it can ignore too much data. Below is hypothetical data for two conferences giving the number of teams sent to the NCAAs over a ten year period. In each case, the numbers in the series represent how many teams were invited.

Conference A: 1, 2, 1, 1, 1, 1, 1, 2, 1, 1

Conference B: 2, 5, 6, 3, 1, 7, 4, 1, 9, 8

Conference A has a mode of 2 and Conference B a mode of 1 but no rational analyst would try to make a case that A was the stronger conference.

There are definitely problems with using the mode for this analysis. In fact, I have no idea why this measure of central tendency was chosen by the author. If he was worried about outliers affecting his analysis, he could have used the median instead. But in the example you provided, the mode (defined as the most common value in the data set) of both conference A and conference B is 1. (Conference A has 8 ones and 2 twos --> mode of 1, while Conference B has 2 ones and one of everything else --> mode of 1.)

The median would have better revealed the differences between the conferences. The middle value of Conference A was 1, while the middle value of Conference B was 4.5, as (4 +5)/2=4.5

 

[KnightBridgeAZ]

And in terms of looking at teams individually, record against the top 25 and top 50 are of interest. The quantity of games is indicative at a quick glance of how many tough games a team played, and - if you played a lot and lost them all, meh.

Sure it is tougher to have beaten UConn than Louisville or whoever, but that doesn't mean that beating Louisville, Baylor and Tennessee while losing to UConn - with a 3-1 record against the top 25 is a bad thing.

 

[Big UConn]

There are definitely problems with using the mode for this analysis. In fact, I have no idea why this measure of central tendency was chosen by the author. If he was worried about outliers affecting his analysis, he could have used the median instead. But in the example you provided, the mode (defined as the most common value in the data set) of both conference A and conference B is 1. (Conference A has 8 ones and 2 twos --> mode of 1, while Conference B has 2 ones and one of everything else --> mode of 1.)

The median would have better revealed the differences between the conferences. The middle value of Conference A was 1, while the middle value of Conference B was 4.5, as (4 +5)/2=4.5

Good example. A weighted mean would be even better, since it would take into consideration how many teams are in each conference.

 

[RegisteredUconn]

There are definitely problems with using the mode for this analysis. In fact, I have no idea why this measure of central tendency was chosen by the author. If he was worried about outliers affecting his analysis, he could have used the median instead. But in the example you provided, the mode (defined as the most common value in the data set) of both conference A and conference B is 1. (Conference A has 8 ones and 2 twos --> mode of 1, while Conference B has 2 ones and one of everything else --> mode of 1.)

The median would have better revealed the differences between the conferences. The middle value of Conference A was 1, while the middle value of Conference B was 4.5, as (4 +5)/2=4.5

You are absolutely correct, Choke. Each distribution has a mode of 1. In my mind, as I constructed my post, the twos were where the ones were and vice versa.