How to determine rebound rate for pre-1969 players.
Posted: Thu Nov 5, 2009 5:47 am
To determine RebRate (or Reb %), you need to know the following five things
1.The amount of rebounds the players has grabbed
2.The amount of rebounds the team has grabbed
3.The amount of rebounds the opposing team has grabbed
4.The number of minutes the player played
5.The number of minutes the entire team played
We don't have listed rebound rates for the pre-1969 era, because don't have correct totals for #2, or listed totals for #3 or, often #5. You can try and do a pace adjustment, but that's not working out Rebound Rate (more on that later). But, actually, it's not that hard to get the exact amount for #2 and #5 most of the time, and a good estimate for #3. With that in mind, here's how to estimate rebound rates for pre-1969 players.
#1. Understand the problem with using listed rebounds total for a team.
Because of terminology, this will be tricky, but here goes. Prior to 1969, the NBA did not separately keep track of "team rebounds." These are rebounds that go out of bounds, and are awarded to a team. On average, these account for about 12.2% of all rebounds prior to 1969. It varies a bit, but not a whole lot...few teams are over 15 or 16%, or below 10%. For entire seasons, the variance is even less...between about 10.8% and 14.5%.
So that means we can make a guess for #3...of how many rebounds the opposing teams has grabbed.
We start by multiplying the league average listed rebounds for a pre-1968 season by .878. This is the average number of rebounds a team grabbed. Substitute it in for #3, and we have half of our missing numbers.
This is also why rebound totals suddenly dropped so radically in the one year of 1969...statisticians effectively took out 12.5% of the rebounds that had been counted before. According to league stats, the average team in 1968 grabbed 5431 rebounds. In 1969, it was 4666...a drop of 14%. Was it really that huge of a drop? Nope...about 12.2% was due to the team rebounds no longer being counted.
#2. How accurate is this going to be?
Welllll....close enough to make a very good analysis some of the time, but not all of the time. In 1965, the average team grabbed is listed at getting 5381 rebounds. If the average without team rebounds is between 10.5% and 14.5% lower, the number will be somewhere between 4601 and 4816. Multiplying by .878 gives 4724. In other words, you're not going to be more than 100 rebounds...about 2% off from the highest and lowest variances. Usually, you'll be 1% off or under. Close. But we can make it better.
#3. How close will that be to how many actual rebounds an opposing team grabbed?
Well, it will get you in the right direction. It's pretty rare to have opposing team rebound totals vary by more than 25%. Look at last year. Golden State surrendered 3860 rebounds; Portland gave up 2976. 26 out of 30 teams were between 3617 and 3187...within about 13%. Still more than a couple of teams would be pretty far out of line if we used "average" league rebounds to replace opposing team rebounds. Four teams would be off by 9.8% or more. How can we get it closer?
#4. Look at how often the team shot the ball relative to the league, and multiply that by the average.
Let's try that with last year's teams as an example. Only one team out of 30 is off by more than 5.88%. Standard deviation confirms this...it's about 127 rebounds, or about 3.7%. 95% of all teams should be within 7.4% of this recalculated amount.
I've actually run it through more teams, and last year was pretty typical...the standard deviation is around 3.6%. There's no wild fluctuations in deviation as we go farther back...it will goes as high as 5% or as low as 2.5%.
What this means, though, is that 95% of the time, we'll be within about 7.2% of the league average. And since this is only ~50% of the team rebound totals we need, we could cut that deviation/error rate in half if we only knew how many rebounds a pre-1969 team had.
#5. We know how many rebounds a pre-1969 team actually had. And how many minutes the team played.
Yup. It's true. And it'd kind of scandalously easy. All you do is add up all the individual player rebounds and minutes. Sometimes pre-1964 team don't have stats for all players, but most do.
1.The amount of rebounds the players has grabbed
2.The amount of rebounds the team has grabbed
3.The amount of rebounds the opposing team has grabbed
4.The number of minutes the player played
5.The number of minutes the entire team played
We don't have listed rebound rates for the pre-1969 era, because don't have correct totals for #2, or listed totals for #3 or, often #5. You can try and do a pace adjustment, but that's not working out Rebound Rate (more on that later). But, actually, it's not that hard to get the exact amount for #2 and #5 most of the time, and a good estimate for #3. With that in mind, here's how to estimate rebound rates for pre-1969 players.
#1. Understand the problem with using listed rebounds total for a team.
Because of terminology, this will be tricky, but here goes. Prior to 1969, the NBA did not separately keep track of "team rebounds." These are rebounds that go out of bounds, and are awarded to a team. On average, these account for about 12.2% of all rebounds prior to 1969. It varies a bit, but not a whole lot...few teams are over 15 or 16%, or below 10%. For entire seasons, the variance is even less...between about 10.8% and 14.5%.
So that means we can make a guess for #3...of how many rebounds the opposing teams has grabbed.
We start by multiplying the league average listed rebounds for a pre-1968 season by .878. This is the average number of rebounds a team grabbed. Substitute it in for #3, and we have half of our missing numbers.
This is also why rebound totals suddenly dropped so radically in the one year of 1969...statisticians effectively took out 12.5% of the rebounds that had been counted before. According to league stats, the average team in 1968 grabbed 5431 rebounds. In 1969, it was 4666...a drop of 14%. Was it really that huge of a drop? Nope...about 12.2% was due to the team rebounds no longer being counted.
#2. How accurate is this going to be?
Welllll....close enough to make a very good analysis some of the time, but not all of the time. In 1965, the average team grabbed is listed at getting 5381 rebounds. If the average without team rebounds is between 10.5% and 14.5% lower, the number will be somewhere between 4601 and 4816. Multiplying by .878 gives 4724. In other words, you're not going to be more than 100 rebounds...about 2% off from the highest and lowest variances. Usually, you'll be 1% off or under. Close. But we can make it better.
#3. How close will that be to how many actual rebounds an opposing team grabbed?
Well, it will get you in the right direction. It's pretty rare to have opposing team rebound totals vary by more than 25%. Look at last year. Golden State surrendered 3860 rebounds; Portland gave up 2976. 26 out of 30 teams were between 3617 and 3187...within about 13%. Still more than a couple of teams would be pretty far out of line if we used "average" league rebounds to replace opposing team rebounds. Four teams would be off by 9.8% or more. How can we get it closer?
#4. Look at how often the team shot the ball relative to the league, and multiply that by the average.
Let's try that with last year's teams as an example. Only one team out of 30 is off by more than 5.88%. Standard deviation confirms this...it's about 127 rebounds, or about 3.7%. 95% of all teams should be within 7.4% of this recalculated amount.
I've actually run it through more teams, and last year was pretty typical...the standard deviation is around 3.6%. There's no wild fluctuations in deviation as we go farther back...it will goes as high as 5% or as low as 2.5%.
What this means, though, is that 95% of the time, we'll be within about 7.2% of the league average. And since this is only ~50% of the team rebound totals we need, we could cut that deviation/error rate in half if we only knew how many rebounds a pre-1969 team had.
#5. We know how many rebounds a pre-1969 team actually had. And how many minutes the team played.
Yup. It's true. And it'd kind of scandalously easy. All you do is add up all the individual player rebounds and minutes. Sometimes pre-1964 team don't have stats for all players, but most do.