Guy986 wrote:Zubby wrote:I wonder where Harden is on this list. lol
I think most would assume Lin/Asik would be high, but I had no idea Stephen Curry is a top 5 defensive guard... 1 among point guards.
What does this all mean?
Why are we suck a bad defensive team if all of our players are average to excellent defenders, aside from Harden who is pretty poor defensively.
Thats ****ing bullsh*t. Curry is the worst defender at PG i've seen this year. His man defense is atrocious.
It means you take these rankings with a grain of salt. Morey himself said there is no advanced stats available that measures perimeter defense accurately. steal=/= good defense.
Took me a bit. Here's something I came up with so far:
By the phrasing of the statements, this is what I think how he's pushing out these numbers:
Here's an RAPM primer:
http://www.swishappeal.com/2011/10/11/2 ... s-a-primerHere are the numbers for Defensive RAPM:
http://stats-for-the-nba.appspot.com/ratings/2012.htmlSo what they did was take a vector of every player's defensive regularized adjusted +/-. So per player, they have a score attached that is generated by a +/-, that is adjusted so that their +/- is adjusted based on the competition they have faced so far, as well as the quality of the team mates they have been playing with. The point of this is you want everyone to be compared like they play with 4 other average players, against average competition.
Given the RAPM of player a, it is multipled by the average possessions per game of the player, and then finding a value for an average player's +/- who has the same average number of possesions per game, the score probably is (player_a(defenensive_reguarlized_plus_minus) - average_player(defensive_regularlied_plus_minus)) * similar_possessions_per_game.
I can see the reasoning behind some of this. By adjusting to defensive regularized adjusted +/-, you are only ranking folks, up and down the list as if they are playing the same level of competition. If this can be done, it is very useful.
GS's starters are David Lee, Klay Thompson, Curry, Barnes and Festus Ezeli. Before Bogut went down, he started. The thing is: If the team is winning, as GS is, then their starter's on court +/- must invariably be positive. If Golden State is winning by a large margin, then they are obviously going to have more +/- to go around. Here is a few ways that Curry's new super stat can be spiked.
a) The issue with comparing with players who share similar possessions per game is that players who play as many possessions as Steph Curry does generally hold back a bit on defense to generate offense, especially in the regular season. So he's probably being compared to Russell Westbrook (winner) as well as Kyrie Irving (loser). The Kyrie Irving (or Kemba Walker) comparasion is going to make his adjusted +/- look really really good. He's winning, and a lot of high usage point guards in the league are not. Although they don't play as many possessions as we do, they are also a high possessions per game ball club.
b) As one normalizes against team mates, i.e. making sure you are normalizing your team strength against all teams, you are trying to extract one player's contribution on defense and adjusting it to the defensive capabilities of other team mates. Again, they are a high possessions per game ball club, ranked 24th in points allowed. Since team wise, their defense sucks anyways, Steph Curry isn't going to get docked as much. Everyone is defensively poor on his team, so if he were playing with defensively average team mates, his +/- would look better...
I get the importance of trying to be fair across the league, but a lot of the losing teams are playing in a way that encourage players to figure stuff out, rather than win games... Further more, a rotation that is Adelman like is really going to make players look a lot better if they are winning games, vs someone who is on an average team in Hubie Brown like rotation is going to get smothered by this stat. It's not a useless stat, but one needs to take into account for outliers such as Steph Curry.
My beef with the rhetoric right here ISNT that the stat is useless, I think it's okay. My deal is that the writer spit out a bunch of numbers from an R script and proceeded to compose a rhetoric around it. I don't mind folks do data analysis work and then try and explain what is going on, but once you start from numbers to rhetoric, rhetoric needs to pose a hypothesis and have another means of testing said hypothesis, just to see if the method is correct.
