Sabermetrics (for nerds only)

[njm8845]

I'm interested in sabermetrics, and wanted to apply Oliver's equations to Purdue's basketball team last year. After slogging through them I arrived at an Offensive Rating and Defensive Rating for each player.

It should be noted that he defines them as follows:

"Just as Oliver's Offensive Rating represents points produced by the player per 100 possessions consumed, his Defensive Rating estimates how many points the player allowed per 100 possessions he individually faced while on the court."

Below are the results.

Player Ortg
Octeus, Jon 121.3
Davis, Rapheal 112.7
Edwards, Vince 112.4
Stephens, Kendall 107.6
Hammons, A.J. 106.8
Smotherman, Basil 104.1
Haas, Isaac 103.7
Mathias, Dakota 103.6
Scott, Bryson 97.3
Thompson, PJ 97.0

PLAYER Drtg
Hammons, A.J. 92.4
Scott, Bryson 94.2
Haas, Isaac 97.2
Octeus, Jon 99.1
Smotherman, Basil 99.6
Thompson, PJ 101.3
Davis, Rapheal 101.3
Mathias, Dakota 101.4
Edwards, Vince 102.3
Stephens, Kendall 103.0

Talking Points:
I left out Taylor because of the small sample size but his defensive rating was actually the best on the team.

This underscores how important Octeus was to our offense last year. I think most people on the boards realize he's a huge loss. But the margin by which he was the most important offensive player surprised me.

I was hoping this would capture how valuable Davis is on defense. Unfortunately, it doesn't. He has the second most "stops" on the team (to AJ) but he played so many minutes that this is diluted by the time we arrive at Drtg.

Does anything stand out to anybody else?

I have the excel version available if there are people who'd like to see the underlying data.

 

[GodFamilyCountryPurdue]

I'm interested in sabermetrics, and wanted to apply Oliver's equations to Purdue's basketball team last year. After slogging through them I arrived at an Offensive Rating and Defensive Rating for each player.

It should be noted that he defines them as follows:

"Just as Oliver's Offensive Rating represents points produced by the player per 100 possessions consumed, his Defensive Rating estimates how many points the player allowed per 100 possessions he individually faced while on the court."

Below are the results.

Player Ortg
Octeus, Jon 121.3
Davis, Rapheal 112.7
Edwards, Vince 112.4
Stephens, Kendall 107.6
Hammons, A.J. 106.8
Smotherman, Basil 104.1
Haas, Isaac 103.7
Mathias, Dakota 103.6
Scott, Bryson 97.3
Thompson, PJ 97.0

PLAYER Drtg
Hammons, A.J. 103.8
Scott, Bryson 106.1
Haas, Isaac 109.7
Octeus, Jon 112.1
Smotherman, Basil 112.7
Thompson, PJ 114.8
Davis, Rapheal 114.8
Mathias, Dakota 115.0
Edwards, Vince 116.1
Stephens, Kendall 116.9

Talking Points:
I left out Taylor because of the small sample size but his defensive rating was actually the best on the team.

This underscores how important Octeus was to our offense last year. I think most people on the boards realize he's a huge loss. But the margin by which he was the most important offensive player surprised me.

I was hoping this would capture how valuable Davis is on defense. Unfortunately, it doesn't. He has the second most "stops" on the team (to AJ) but he played so many minutes that this is diluted by the time we arrive at Drtg.

Does anything stand out to anybody else?

I have the excel version available if there are people who'd like to see the underlying data.

Not that surprising. Davis, Edwards, and Hammons were our best offensive players last year, but Jon had the ball in his hands the most. He was a great addition, but the numbers don't prove he was the most valuable offensively. Defensively, Davis was always defending the other team's best offensive threat, so it is given that he will give up more points. I'd like to see a stat that shows how many points Player A gave up to Player B compared to Player B's average.

 

[BoilerSmac]

I'm interested in sabermetrics, and wanted to apply Oliver's equations to Purdue's basketball team last year. After slogging through them I arrived at an Offensive Rating and Defensive Rating for each player.

It should be noted that he defines them as follows:

"Just as Oliver's Offensive Rating represents points produced by the player per 100 possessions consumed, his Defensive Rating estimates how many points the player allowed per 100 possessions he individually faced while on the court."

Below are the results.

Player Ortg
Octeus, Jon 121.3
Davis, Rapheal 112.7
Edwards, Vince 112.4
Stephens, Kendall 107.6
Hammons, A.J. 106.8
Smotherman, Basil 104.1
Haas, Isaac 103.7
Mathias, Dakota 103.6
Scott, Bryson 97.3
Thompson, PJ 97.0

PLAYER Drtg
Hammons, A.J. 92.4
Scott, Bryson 94.2
Haas, Isaac 97.2
Octeus, Jon 99.1
Smotherman, Basil 99.6
Thompson, PJ 101.3
Davis, Rapheal 101.3
Mathias, Dakota 101.4
Edwards, Vince 102.3
Stephens, Kendall 103.0

Talking Points:
I left out Taylor because of the small sample size but his defensive rating was actually the best on the team.

This underscores how important Octeus was to our offense last year. I think most people on the boards realize he's a huge loss. But the margin by which he was the most important offensive player surprised me.

I was hoping this would capture how valuable Davis is on defense. Unfortunately, it doesn't. He has the second most "stops" on the team (to AJ) but he played so many minutes that this is diluted by the time we arrive at Drtg.

Does anything stand out to anybody else?

I have the excel version available if there are people who'd like to see the underlying data.

Forgive me if this data is available somewhere, but how are the "points allowed" tabulated? Is it just as arbitrary as RD started the 3 in Game A for 32 minutes, and the other team started Joe Shmo at the 3, who also played 32 minutes, and he had 11 points, so RD gave up 11 points in 32 minutes?

It would seem extraordinarily difficult, to accurately tabulate this statistic, unless there is actually someone watching the games and keeping track of actual defensive assignments during the course of the game.

 

[mathboy]

I'm interested in sabermetrics, and wanted to apply Oliver's equations to Purdue's basketball team last year. After slogging through them I arrived at an Offensive Rating and Defensive Rating for each player.

It should be noted that he defines them as follows:

"Just as Oliver's Offensive Rating represents points produced by the player per 100 possessions consumed, his Defensive Rating estimates how many points the player allowed per 100 possessions he individually faced while on the court."

I have the excel version available if there are people who'd like to see the underlying data.

I would think this was terribly tedious to assemble. You would need to track when each player was on the floor and what the score differential was during their time playing. I assume that points scored against us while the player was on the floor went into his defensive numbers regardless of who he was guarding, right? In the same manner, if we scored while player X was on the court, his numbers would get the increase.

If you try to organize the numbers around the points scored by each player, on a per minute basis, I think you will get data that is less useful. Maybe you can tell us the calculation method - for us nerds that like that stuff.

 

[BoilerDaddy]

I'm interested in sabermetrics, and wanted to apply Oliver's equations to Purdue's basketball team last year. After slogging through them I arrived at an Offensive Rating and Defensive Rating for each player.

It should be noted that he defines them as follows:

"Just as Oliver's Offensive Rating represents points produced by the player per 100 possessions consumed, his Defensive Rating estimates how many points the player allowed per 100 possessions he individually faced while on the court."

Below are the results.

Player Ortg
Octeus, Jon 121.3
Davis, Rapheal 112.7
Edwards, Vince 112.4
Stephens, Kendall 107.6
Hammons, A.J. 106.8
Smotherman, Basil 104.1
Haas, Isaac 103.7
Mathias, Dakota 103.6
Scott, Bryson 97.3
Thompson, PJ 97.0

PLAYER Drtg
Hammons, A.J. 92.4
Scott, Bryson 94.2
Haas, Isaac 97.2
Octeus, Jon 99.1
Smotherman, Basil 99.6
Thompson, PJ 101.3
Davis, Rapheal 101.3
Mathias, Dakota 101.4
Edwards, Vince 102.3
Stephens, Kendall 103.0

Talking Points:
I left out Taylor because of the small sample size but his defensive rating was actually the best on the team.

This underscores how important Octeus was to our offense last year. I think most people on the boards realize he's a huge loss. But the margin by which he was the most important offensive player surprised me.

I was hoping this would capture how valuable Davis is on defense. Unfortunately, it doesn't. He has the second most "stops" on the team (to AJ) but he played so many minutes that this is diluted by the time we arrive at Drtg.

Does anything stand out to anybody else?

I have the excel version available if there are people who'd like to see the underlying data.

Did you look at individual stats that go into offensive rating (which is hard to do for defensive rating) or team data (i.e. points per possession) for possessions while the player was in the game? Your offensive rating numbers are pretty close to those on statsheet.com, but there are a few significant differences (such as RDavis).

 

[icewind7]

Off topics.. "nerd" this word is outdated in today's world =)

 

[septimusharding]

Wow, good work. Your Ortg numbers are somewhat different than those on KenPom, but the ordering of the players is pretty much the same.

The one thing that I would add is that we have to consider the players' usage ratings as well. Ortg numbers can't be looked at in a vacuum. These numbers show that Octeus was very efficient offensively, but they don't show that he was our most important player on offense. Here are the usage rates for our top players during last season:

Haas 30.3
Hammons 26.6
Scott 25.1
Stephens 20.7
Davis 19.7
Edwards 18.4
Octeus 18.0

This shows that Octeus was more of an opportunistic scorer, rather than a go-to-guy ala Hammons. That doesn't mean he wasn't important, but in general, it's harder to replace a guy with high usage and decent efficiency (like Hammons) than it is to replace a guy with medium usage and great efficiency (like Octeus).

I still think the loss of Octeus is important, both from a leadership perspective and a positional perspective. But I don't think it's a disaster. And if either Thompson or Hill (or some other pseudo point guard that some people have been suggesting) steps up, our offense can be just fine.

 

[njm8845]

Really good point about the difference between efficiency and usage. The problem is that efficiency numbers are significantly skewed by small samples.

For anyone who doesn't know, the usage rate is the percentage of time a particular player looks to score. So according to the data posted above, Scott looked to score on 25% of the possessions when he was in the game. Octeus' was 18%. While Octeus had a lower usage rate, he was still able to produce about 24 more points per 100 possessions than Scott. This is obtained by the difference in their offensive rating and is chiefly attributed to Octeus' assists and high A/TO ratio.

Below are the equations I used to calculate both Ortg and Drtg. I don't have much faith in the Drtg - as others have said, it's just too difficult to quantify defense by using traditional metrics such as steals, points, etc. I wish the wealth of data that is becoming popular in the NBA was also available in the college game.

http://www.basketball-reference.com/about/ratings.html

 

[njm8845]

Off topics.. "nerd" this word is outdated in today's world =)

If I've used the term "nerd" to describe myself, and that term is outdated, does that make me even more of a nerd?