Earlier this week I blogged about how consistent NBA players were from the 2012/13 regular season to the 2013/14 regular season. What I found is that compared to players in other leagues, NBA players are much more consistent. We also report this for a number of seasons in chapter 9 of our book, The Wages of Wins. I got to thinking about whether this result holds over a longer period of time, and asked Dave Berri if he would send me the data, which he graciously did. Dave's NBA player data starts with the 1973/74 regular season and goes up to the 2013/14 regular season, which for those keeping track is over forty years of player data.
So, just as I did earlier this week, I ranked all the players (see link above for the details) and look at players who played in consecutive regular seasons playing at least 10% of the total regular season minutes (having adjusted for the 1998/99 and 2011/12 shortened seasons). What I find is that 42.1% of all the players who played at least 10% of total minutes in consecutive regular seasons are consistent (i.e. are in the same quantile), while 39.2% either move up or down one quantile from one season to the next; meaning that only 18.7% of NBA players are very inconsistent over consecutive seasons.
Showing posts with label Player Consistency. Show all posts
Showing posts with label Player Consistency. Show all posts
Friday, May 23, 2014
Tuesday, May 20, 2014
NBA Player Consistency
Last week I blogged about NHL goalie regular season production consistency. I only included goalies who started a minimum of 10% games played in consecutive regular seasons. Now I am turning to NBA players using the regular season data from those NBA player stat gurus at boxscoregeeks.com. Here I am looking at players who have played a minimum of 10% of regular season minutes (approximately 394 minutes) in consecutive regular seasons from the 2012-13 and 2013-14 NBA regular seasons.
How is consistency measured? Previously I stated for MLB batters that, "...consistency is measured by where they rank in terms of runs created in a given season. For simplicity in The Wages of Wins, we took a player who was in the top 20% of the league in runs created as having a rank equal to 1, the next highest 20% with a rank of 2, all the way to the lowest 20% with a rank of 5. If a player maintains their rank from one season to another they are considered to be consistent and if not, they are considered to be inconsistent."
When I run the NBA player consistency analysis (step-by-step guide here), I find that there are 287 NBA players that have played at least 394 minutes in both seasons and of those 287 NBA players, 45.6% of those players have keep their same "grade" in each season. Additionally, another 34.5% of NBA players move up or down only one "grade" from season to season, leaving about 19.9% of NBA players moving two or more "grades" from the 2012-13 season to the 2013-14 NBA regular season. This is similar to what we found in The Wages of Wins (see the end of chapter 9).
There were 131 NBA players that were consistent in consecutive seasons. Of those, 34.35% were in the top 20% of the distribution (or "A grade" player production), while 19.08% were "B grade" players, 13.74% were "C grade" players, 16.79% were "D grade" players and 16.03% were in the bottom top 20% of the distribution in consecutive seasons.
For those who read Dave Berri's piece on John Wall, he is a consistent "B grade" player.
How is consistency measured? Previously I stated for MLB batters that, "...consistency is measured by where they rank in terms of runs created in a given season. For simplicity in The Wages of Wins, we took a player who was in the top 20% of the league in runs created as having a rank equal to 1, the next highest 20% with a rank of 2, all the way to the lowest 20% with a rank of 5. If a player maintains their rank from one season to another they are considered to be consistent and if not, they are considered to be inconsistent."
When I run the NBA player consistency analysis (step-by-step guide here), I find that there are 287 NBA players that have played at least 394 minutes in both seasons and of those 287 NBA players, 45.6% of those players have keep their same "grade" in each season. Additionally, another 34.5% of NBA players move up or down only one "grade" from season to season, leaving about 19.9% of NBA players moving two or more "grades" from the 2012-13 season to the 2013-14 NBA regular season. This is similar to what we found in The Wages of Wins (see the end of chapter 9).
There were 131 NBA players that were consistent in consecutive seasons. Of those, 34.35% were in the top 20% of the distribution (or "A grade" player production), while 19.08% were "B grade" players, 13.74% were "C grade" players, 16.79% were "D grade" players and 16.03% were in the bottom top 20% of the distribution in consecutive seasons.
For those who read Dave Berri's piece on John Wall, he is a consistent "B grade" player.
Wednesday, May 14, 2014
2013-2014 NHL Regular Season Goalie Consistency
At the end of the regular season I ran the numbers for NHL goalies and found that Tuukka Rask, Semyon Varlamov and Carey Price were the top three NHL goalies for this season using Wins Above Average (WAA). WAA is the marginal value of a goal against divided by
two (since each win is worth two standings points) times the number of
shots on goal that a goalie faces times the difference in the save
percentage of the goalie and the average save percentage of all goalies
for that season in absolute value (since a goal against has a negative
effect on team wins). Here is a step-by-step guide to calculate NHL goalie performance.
Now what I want to do is figure out how consistent NHL goalies are from one season to another, so I am comparing the regular season performance from last season (which was shortened due to the lockout) and this season. To do that, I am going to only look at goalies that started at least 10% of the regular season games played for each season, since the number of games played is different between the two regular seasons.
Thus to find how consistent NHL goalies are in consecutive regular seasons, I am looking at the percentage of NHL goalies who are in the same WAA quantile in consecutive regular seasons and those who have moved up or down one quantile from one season to the next. Here is a step-by-step guide on measuring player consistency.
What I find is that of the NHL goalies who started at least 10% of the total number of regular season games, 30.8% were consistent (in the same quantile in each season), 32.7% were near consistent (either moved up or down one quantile from one regular season to another) and 36.5% were inconsistent (moved two or more quantiles from one regular season to another).
Of those that were consistent, there were only three NHL goalies who were in the top quantile (i.e. top 20% of NHL goalies in consecutive regular seasons). They are Tuukka Rask (Boston Bruins), Sergei Bobrovsky (Columbus Blue Jackets) and Henrik Lundqvist (New York Rangers). Given that Tuukka Rask was the most productive goalie (but just barely) and that he is also only one of three goalies to be in the top 20% from one season to the next, I would say that he is the best candidate (statistically) for the Vezina Trophy.
Now what I want to do is figure out how consistent NHL goalies are from one season to another, so I am comparing the regular season performance from last season (which was shortened due to the lockout) and this season. To do that, I am going to only look at goalies that started at least 10% of the regular season games played for each season, since the number of games played is different between the two regular seasons.
Thus to find how consistent NHL goalies are in consecutive regular seasons, I am looking at the percentage of NHL goalies who are in the same WAA quantile in consecutive regular seasons and those who have moved up or down one quantile from one season to the next. Here is a step-by-step guide on measuring player consistency.
What I find is that of the NHL goalies who started at least 10% of the total number of regular season games, 30.8% were consistent (in the same quantile in each season), 32.7% were near consistent (either moved up or down one quantile from one regular season to another) and 36.5% were inconsistent (moved two or more quantiles from one regular season to another).
Of those that were consistent, there were only three NHL goalies who were in the top quantile (i.e. top 20% of NHL goalies in consecutive regular seasons). They are Tuukka Rask (Boston Bruins), Sergei Bobrovsky (Columbus Blue Jackets) and Henrik Lundqvist (New York Rangers). Given that Tuukka Rask was the most productive goalie (but just barely) and that he is also only one of three goalies to be in the top 20% from one season to the next, I would say that he is the best candidate (statistically) for the Vezina Trophy.
Friday, March 7, 2014
Guide to Calculating Player Consistency
In the previous blog I talked about how consistent MLB batters were in back-to-back seasons. Here I want to present the step-by-step process to do this type of analysis.
Step 1: You have to have an algorithm to evaluate team performance that allows you to take the impact of various team statistics such that you can evaluate player performance. For MLB batters I use Runs Created and for NHL Goalies I use the marginal value of a goal against.
Step 2: You have to have an algorithm to evaluate player performance, whether it is Runs Created, Wins Above Average, WP48, QB Score or another metric. I will assume that you have a method to evaluate player performance.
(I assume that you have the team data from step 1 to calculate the marginal value of the team statistics and that you have used those marginal team values to calculate the player performance measure in step 2. Once this is done then we can get to step 3. If not - see the links above to replicate the analysis for either MLB batters or NHL goalies).
Step 3: I use a cut-off point for the players (but this is not required). If you use a cut-off point such as minimum AB's for MLB batters, sort the data and eliminate the data that does not meet your minimum requirements. Then re-sort the data by season and by your player performance measure from highest to lowest.
Step 4a: Calculate the order from 1 to n (n being the lowest player performance number for that season). If the order is in column A, then in cell A2 type in a 1, and in cell A3 =A2+1, and copy the formula and paste down for each player in that season. Repeat for each season.
Step 4b: Calculate the rank for each player Calculate each quintile (range of 20%) for each season. In Excel I use a formula like this: =IF(A2/$A$438<=0.2,1,IF(A2/$A$438<=0.4,2,IF(A2/$A$438<=0.6,3,IF(A2/$A$438<=0.8,4,5)))), where A2 is the highest ranked player by number (i.e. 1) and A438 would be the last player for that season.
Step 5: Sort the data by player name and player year (lowest to highest).
Step 6: Create a formula such that only the same player will be listed if they met the player performance criteria in consecutive seasons. Thus for each variable you are interested in, I use the following formula in Excel in row 3: =IF($D4-$D3=1,IF($C3=$C4,B4,""),""), where column D is the column that has the season played and column C has the players name and column B has the players rank. Since you are looking at row 4, that should be the next season. If it is not, then the formula above reports a blank cell. Copy and paste for all the players.
Step 7: Calculate consistency (and near consistency) as follows: =IF(I3="","",IF(I3=B3,1,0)) which says that if the column with the next years rank (column I in this case) is blank, to leave it blank, otherwise to put a one if they are the equal and a zero if they are not equal. For near consistency I use the following formula: =IF(I3="","",IF(I3+1=B3,1,0)) if I want to know how many have increased their rank by one and =IF(I3="","",IF(I3-1=B3,1,0)) if I want to know how many have decreased their rank by one. Copy and paste for all the players.
Step 8: Finally I calculate the percentage of players who are consistent and near consistent and then you can find out how many are not even nearly consistent from one season to another.
Step 1: You have to have an algorithm to evaluate team performance that allows you to take the impact of various team statistics such that you can evaluate player performance. For MLB batters I use Runs Created and for NHL Goalies I use the marginal value of a goal against.
Step 2: You have to have an algorithm to evaluate player performance, whether it is Runs Created, Wins Above Average, WP48, QB Score or another metric. I will assume that you have a method to evaluate player performance.
(I assume that you have the team data from step 1 to calculate the marginal value of the team statistics and that you have used those marginal team values to calculate the player performance measure in step 2. Once this is done then we can get to step 3. If not - see the links above to replicate the analysis for either MLB batters or NHL goalies).
Step 3: I use a cut-off point for the players (but this is not required). If you use a cut-off point such as minimum AB's for MLB batters, sort the data and eliminate the data that does not meet your minimum requirements. Then re-sort the data by season and by your player performance measure from highest to lowest.
Step 4a: Calculate the order from 1 to n (n being the lowest player performance number for that season). If the order is in column A, then in cell A2 type in a 1, and in cell A3 =A2+1, and copy the formula and paste down for each player in that season. Repeat for each season.
Step 4b: Calculate the rank for each player Calculate each quintile (range of 20%) for each season. In Excel I use a formula like this: =IF(A2/$A$438<=0.2,1,IF(A2/$A$438<=0.4,2,IF(A2/$A$438<=0.6,3,IF(A2/$A$438<=0.8,4,5)))), where A2 is the highest ranked player by number (i.e. 1) and A438 would be the last player for that season.
Step 5: Sort the data by player name and player year (lowest to highest).
Step 6: Create a formula such that only the same player will be listed if they met the player performance criteria in consecutive seasons. Thus for each variable you are interested in, I use the following formula in Excel in row 3: =IF($D4-$D3=1,IF($C3=$C4,B4,""),""), where column D is the column that has the season played and column C has the players name and column B has the players rank. Since you are looking at row 4, that should be the next season. If it is not, then the formula above reports a blank cell. Copy and paste for all the players.
Step 7: Calculate consistency (and near consistency) as follows: =IF(I3="","",IF(I3=B3,1,0)) which says that if the column with the next years rank (column I in this case) is blank, to leave it blank, otherwise to put a one if they are the equal and a zero if they are not equal. For near consistency I use the following formula: =IF(I3="","",IF(I3+1=B3,1,0)) if I want to know how many have increased their rank by one and =IF(I3="","",IF(I3-1=B3,1,0)) if I want to know how many have decreased their rank by one. Copy and paste for all the players.
Step 8: Finally I calculate the percentage of players who are consistent and near consistent and then you can find out how many are not even nearly consistent from one season to another.
Subscribe to:
Posts (Atom)