Post by greeme on Jan 4, 2011 7:43:31 GMT -5
The Celtics gm asked a great question in the pf/c ratings thread "Randolph has been playing SF for two years now. But really, how does a guy doing 32/10 get 10th? Actually, I know the answer, blocks. But he shoots 51% FG and 83%FT and has obscenely low TO's. I don't think you take into account things like this. He is usually 1 or 2 in the league in FTA too."
Why does Randolph come out 10th right behind Darko who has little scoring ability?
The short answer is because they are my rankings and I decide what goes into them. Not that I tweak the results but the formula has my own imprint and therefore my own bias.
Mark Twain (among others) popularized a phrase "There are three kinds of lies: lies, damned lies, and statistics."
Statistics per se have no value judgments attached to them. For example, points per game. You can't argue with it because it does exactly what it says, it shows you who scores the most points per game. You can argue how important the statistic is but not the correctness of the stat. In my opinion ppg is one of the most overrated statistics around. It is connected, among other things, to how many shots someone takes and to how many minutes somebody plays. Maybe points per minute is a better gauge of a shooter's effectiveness? Or maybe a statistic showing how many points a player averages per shot?
On the other hand if you believe in your statistical formula sometimes you find that a player you intuitively rank high is actually ranked lower. Should you tweak the formula in order to rate the player higher? Not if you believe in the statistical formula. If most players are in an order which seems logical and only a few seem out of order then it would make sense to reevaluate your opinion of the player and not reevaluate the formula. If a lot of players seem out of place the problem is probably in the formula. I think that the list more or less is in sync with how I see players and so the question is just if you agree with the formula or not. The formula I use is based on my own value judgments of how I value player performance. For instance I think ppg is a vastly overrated stat and I am more interested in the effectiveness of a player's shooting than how many points he scored. If you disagree with that then the formula is of less use to you.
The statistical formula that I use is
((assists*1.067336)-(turnovers*0.932664))*mpg+ (3pt fgs*2.067336+(fgs-3pt fgs)*1.067336 + fts/2*1.067336)-(fga-fg)*0.932664)-(fta/2-ft/2)*0.932664) + (((rpm+spm+bpm*0.75)*mpg)*0.932664)
0.932664 = worth of an average possession over the season (possessions were calculated by adding fga, fta/2 and turnovers. The worth of a possession was calculated by dividing possessions by points)
1.067336 = the difference between the worth of a possession and a fg. If a player scores on a 2 pt fg he actualizes the potential of the possession and helps his team the difference between the basket and the original worth of the possession.
For any formula to be useful we must be aware of its blind spots/biases/weaknesses. The main weakness in this formula is that it beyond steals and blocks it can't compute defense. A great man to man defender who does not come up with steals or blocks will be shortchanged. Also players who play pg, pf and c usually have better ratings because they get more assists and blocks/rebounds, respectively.
To go back to the original question, who had a better statistical year Randolph or Darko?
Darko
Games MPG FG FGA FG % FT FTA FT % 3P 3PA 3P % RPG APG SPG BPG TOPG DQ PPG
84 34.2 263 637 .413 204 261 .782 0 0 .000 13.1 1.9 0.9 3.7 1.2 9 8.7
Randolph
Games MPG FG FGA FG % FT FTA FT % 3P 3PA 3P % RPG APG SPG BPG TOPG DQ PPG
78 39.1 921 1796 .513 610 724 .843 34 101 .337 9.7 1.9 1.0 1.0 1.4 2 31.9
Randolph is clearly a vastly superior offensive player. Even ignoring the fact that he outscores Darko by 23 ppg it is very clear that Randolph has better %s and is just an all around better offensive player. Steals, tos and assists are pretty much even. Darko has 2.7 more blocks per game and 3.4 more rebounds. Still should that be enough to offset Randolph's clear advantage at the offensive end? Let's say that Randolph takes 28 fgs a game and 9 fts a game. He will make about 14 fgs and 7.5 fts. How would this compare to an average player? An average player shoots approximately 45.6% from the field and 78% from the line. So an average player would make about 12.5 fgs and 7 fts. In total, the difference between an average player and Randolph, given the same number of shots, is only about 4 ppg. Randolph's rebounds and blocks are less than avg for a pf so that lowers his rating.
Darko shoots fts at an avg rate but hurts his team with a low fg%. That deficiency is not so damaging because of the low number of shots he takes. He only take 9 shots a game so with his 41% he would hit about 3.5 and an average player would hit about 4. So Darko hurts his team by 1 point per game on the offensive end.
If Randolph helps his team by 4 points and Darko hurts his team by one point then the difference is 5 points. On the other hand Darko gets his team 3.4 more rebounds a game giving his team 3.4 more possession. In addition he blocks 2.7 shots per game more than Randolph. Since it is impossible to know how many changes of possession are caused by blocks I use a coefficient of 0.75. Therefore the additional blocks give Darko's team an additional 2+ possessions a game. In total Darko gives his team approximately 5.5 additional possessions per game.
Each possession is worth 0.932664 points so 5.5 are worth approximately 5.1 points per game. Since 5.1>5 Darko is more valuable according to the formula than Randolph. Obviously another formula based on a different rating system could yield different results.
I would appreciate any comments regarding the formula itself because it is still a work in progress but comments such as this player is too high/low are pretty useless.
Why does Randolph come out 10th right behind Darko who has little scoring ability?
The short answer is because they are my rankings and I decide what goes into them. Not that I tweak the results but the formula has my own imprint and therefore my own bias.
Mark Twain (among others) popularized a phrase "There are three kinds of lies: lies, damned lies, and statistics."
Statistics per se have no value judgments attached to them. For example, points per game. You can't argue with it because it does exactly what it says, it shows you who scores the most points per game. You can argue how important the statistic is but not the correctness of the stat. In my opinion ppg is one of the most overrated statistics around. It is connected, among other things, to how many shots someone takes and to how many minutes somebody plays. Maybe points per minute is a better gauge of a shooter's effectiveness? Or maybe a statistic showing how many points a player averages per shot?
On the other hand if you believe in your statistical formula sometimes you find that a player you intuitively rank high is actually ranked lower. Should you tweak the formula in order to rate the player higher? Not if you believe in the statistical formula. If most players are in an order which seems logical and only a few seem out of order then it would make sense to reevaluate your opinion of the player and not reevaluate the formula. If a lot of players seem out of place the problem is probably in the formula. I think that the list more or less is in sync with how I see players and so the question is just if you agree with the formula or not. The formula I use is based on my own value judgments of how I value player performance. For instance I think ppg is a vastly overrated stat and I am more interested in the effectiveness of a player's shooting than how many points he scored. If you disagree with that then the formula is of less use to you.
The statistical formula that I use is
((assists*1.067336)-(turnovers*0.932664))*mpg+ (3pt fgs*2.067336+(fgs-3pt fgs)*1.067336 + fts/2*1.067336)-(fga-fg)*0.932664)-(fta/2-ft/2)*0.932664) + (((rpm+spm+bpm*0.75)*mpg)*0.932664)
0.932664 = worth of an average possession over the season (possessions were calculated by adding fga, fta/2 and turnovers. The worth of a possession was calculated by dividing possessions by points)
1.067336 = the difference between the worth of a possession and a fg. If a player scores on a 2 pt fg he actualizes the potential of the possession and helps his team the difference between the basket and the original worth of the possession.
For any formula to be useful we must be aware of its blind spots/biases/weaknesses. The main weakness in this formula is that it beyond steals and blocks it can't compute defense. A great man to man defender who does not come up with steals or blocks will be shortchanged. Also players who play pg, pf and c usually have better ratings because they get more assists and blocks/rebounds, respectively.
To go back to the original question, who had a better statistical year Randolph or Darko?
Darko
Games MPG FG FGA FG % FT FTA FT % 3P 3PA 3P % RPG APG SPG BPG TOPG DQ PPG
84 34.2 263 637 .413 204 261 .782 0 0 .000 13.1 1.9 0.9 3.7 1.2 9 8.7
Randolph
Games MPG FG FGA FG % FT FTA FT % 3P 3PA 3P % RPG APG SPG BPG TOPG DQ PPG
78 39.1 921 1796 .513 610 724 .843 34 101 .337 9.7 1.9 1.0 1.0 1.4 2 31.9
Randolph is clearly a vastly superior offensive player. Even ignoring the fact that he outscores Darko by 23 ppg it is very clear that Randolph has better %s and is just an all around better offensive player. Steals, tos and assists are pretty much even. Darko has 2.7 more blocks per game and 3.4 more rebounds. Still should that be enough to offset Randolph's clear advantage at the offensive end? Let's say that Randolph takes 28 fgs a game and 9 fts a game. He will make about 14 fgs and 7.5 fts. How would this compare to an average player? An average player shoots approximately 45.6% from the field and 78% from the line. So an average player would make about 12.5 fgs and 7 fts. In total, the difference between an average player and Randolph, given the same number of shots, is only about 4 ppg. Randolph's rebounds and blocks are less than avg for a pf so that lowers his rating.
Darko shoots fts at an avg rate but hurts his team with a low fg%. That deficiency is not so damaging because of the low number of shots he takes. He only take 9 shots a game so with his 41% he would hit about 3.5 and an average player would hit about 4. So Darko hurts his team by 1 point per game on the offensive end.
If Randolph helps his team by 4 points and Darko hurts his team by one point then the difference is 5 points. On the other hand Darko gets his team 3.4 more rebounds a game giving his team 3.4 more possession. In addition he blocks 2.7 shots per game more than Randolph. Since it is impossible to know how many changes of possession are caused by blocks I use a coefficient of 0.75. Therefore the additional blocks give Darko's team an additional 2+ possessions a game. In total Darko gives his team approximately 5.5 additional possessions per game.
Each possession is worth 0.932664 points so 5.5 are worth approximately 5.1 points per game. Since 5.1>5 Darko is more valuable according to the formula than Randolph. Obviously another formula based on a different rating system could yield different results.
I would appreciate any comments regarding the formula itself because it is still a work in progress but comments such as this player is too high/low are pretty useless.