You can't escape when you have AB written all over you, with or without anybody's goggles. Ariza is fine. his biggest sin is still the same: he can't shoot. But NO is doing quite well this year. Besides injury, the biggest change is Ariza plays D for them. And he's logging heavy minutes because of that.
You just "granted" the counter-argument. He got those stats playing against backups and going "all-out". That style had to be toned down when he played with the starters. So his free throw attempts and rebounds are down, which is expected given the different role and him playing next to Scola/Hayes and shooting more 3s. I will say this about Lowry. He went from being one of the best backups in the league last year to being a middle-of-the-pack (lower middle half) starting PG this year. That's what the boxscore statistics, if we want to put much stock in that, would suggest. As a starter, Lowry has averaged 12.5 pts, 4.2 reb, 7.2 assists, 1.8 steals, and 2.2 turnovers per 36 minutes played. Using Hollinger's boxscore rating system (GmScr), that's 12.7. Brooks last year averaged 19.8 pts, 2.7 reb, 5.4 assists, 0.8 steals, and 2.9 turnovers per 36 minutes. His GmScr rating came to 13.0. 12.7 for Lowry this year (who's played poorly according to you), and 13.0 for AB last year (who was a top 5 PG according to you). But that's somewhat superficial, I'm sure you'd agree. Boxscore stats tell you some things, but like you yourself said you have to put it in context and recognize that there are things these players do that simply aren't captured well by the boxscore (like defense or lack thereof and creating good looks for teammates). That's why I think its also helpful to pay attention to how other things in the game change when various players are on the floor that aren't directly credited to them in the boxscore. That's where +/- and related stats come in.
Gracias (and to the other poster). The only thing that concerns me about what's going on with Brooks is that he's still a very young player. He's had quite a bit of growth over the last two years and it seems like the whole mess this season has become is causing the team to give up on him too soon. I think we are all aware that when you take the high impact guys - the guys who have the ball in their hands a lot, who take a higher volume of shots - and start fumbling around with their PT, they suffer. When you do that you are essentially calling it quits on those type of guys. I am not sure we should be doing that so quickly with Brooks.
So you're saying you want the ups and downs to be shown in the data? I would have thought you'd want the ups-and-downs to be smoothed out instead, as a particular player (like Jordan Hill perhaps) may have been just going through a cold spell (in this case a dumb spell). If you looked at a 5 game sample a few weeks ago though, you'd have thought he was on his way to becoming a future allstar. See this is my problem, you'd have to still know all the the mitigating factors why the +/- stat works. I guess it might be a difference of opinion, as I want my stats to be easily understood and clear cut (i.e. more TS%= more efficient scorer). In this case you need to be aware of other factors that might unbalance the +/-, for example Mo Williams would have great +/- when LBJ was there, even if all he did was shoot 3 pointers and pretty much nothing else. IMHO stats are there to make life easier for you, not give you additional work to do. I guess if you were looking at the team as a whole, finding out why a solid two-way player has horrible +/- would be a worthwile endeavor, but if you were say another GM and you were evaluating just Courtney Lee to see if you were gonna trade for him, you'd have to do more investigating only to find out that the horrible +/- is due to him not playing with Hayes and Lee having a cold spell the last 5 games. That has like 0% impact on your decision/evaluation of Courtney Lee, yet because Lee's +/- is so horrible you need to look into it. Like I said, unlike other stats +/- is extremely ambigous and can be misleading at times, since a lot of other factors other the player itself influences the stat which is why personally I'm not a big fan of +/-. I can see why you (and a ton of other people) find it useful though.
+/- is a trend statistic... the more game data you collect, the more information you can gain from it. It's also something that can easily be tracked during the game. If you will notice the traditional box score is filled with the easiest numbers to find in the game... points, shots, fouls, rebounds, blocks and now blocks against... are all things fairly easily identified during play. That's the reason plus/minus was added, it is very easy to track, with just the play by play, and with a large base of games you gain some interpretation of what happens when a player is on the court.
In the original post, I gave: (1) average +/- for entire season, (2) +/- for every game, and (3) moving average +/- for prior 5 games. If you want the ups and downs to be hidden away, you can ignore (2) and (3). I don't need every statistic I look at to be clear-cut in the conclusions that can be drawn. So, when ranking greatest players of all time, one thing I think it makes sense to look at is winning. Did the player win a lot of games? Did they win a lot of championships? To me, that's relevant information, and I can't ignore it. But at the same time I can't just say "Robert Horry won 7 championships. He's greater than Charles Barkley." Sometimes you have to be more nuanced in how you take in information and interpret it. That's kind of how I see +/- statistics.
I didn't want to create a new thread on this, but I came across these adjusted +/- numbers recently (linked to from APBRmetrics board): http://stats-for-the-nba.appspot.com/ Adjusted +/-, for those that don't know, tries to estimate a player's +/- impact after "adjusting" for other factors (usually the other players on the floor). It does this by performing a regression over many (many) observed "stints" -- by which I mean a time interval during a game in which there are no player substitutions. Sometimes, the analysis isn't able to correctly distinguish who should get credit for positive or negative stints and so the final results are strange. Here's a plot for the current Rockets (including error bars for offense and defense) using the last 1 1/2 seasons worth of data: Kyle Lowry does well, which isn't totally surprising given results from last year or the data shown in the first post of this thread. On the other hand, Courtney Lee's strengths as a defensive player don't seem to show up very well in this data. Perhaps that can be attributed to him playing on such an awful New Jersey team last year which has had a coaching change this season -- the analysis might be unfairly penalizing Lee for the Nets getting better this year. Shane and Chuck are both considered positive contributors on offense by this analysis, another surprise. Though Adelman seems to be comfortable with what they do in his system, and maybe this analysis is picking up on some of the positive things they do which the coach likes.
There is some validity to this,but it many limitations. Chuck for instance is having a much bigger impact than Shane right now.
Yeah this adjusted +/- is a very accurate stat, the only thing you have to watch for is that each of these guys got enough minutes with enough different combinations.
Good graph. Jordan Hill is pretty lonely there in the 3rd quadrant. As for Battier and Hayes being plus guys on offense-- not really suprising. There are other analysis done by stats-oriented bloggers in the past who found, for example that Battier has a bigger positive impact on offense than he does on D. I think it says something about the importance of stuff that a low-usage player can do to positively contribute to offense. Battier and Hayes may not shoot a lot, but they generally have smart shot selection (i.e. taking shots when it's likely the best available shot you can get in a possession), make the right passes in the offense and does so without eating up the clock. Battier also spaces the floor while Hayes hits the offensive boards.
Here's my basic problem with what I regard as arithmatic gooble-dee-gook. From the x-y point of origin (0,0), the players in the upper right quadrant are theoretically the best combination of O&D. Those three players (according to the graph) would be named Kyle Lowery, Chuck Hayes and Shane Battier. There are two players just barely in the positive O&D quadrant...Scola and Brad Miller. The worst defensive player on the graph is Kevin Martin. Who just happens to be the 3rd highest O player. What I propose (from viewing...not graphing) is that players like Martin compensate for the lack O on the part of Hayes & Battier. The latter two getting a huge O lift into the upper right quadrant because they are on the floor with Martin (as well as Lowery). The other issue I have is this. If your goal is competing for the WC Finals, the associated +/- that should be used (secondarily to watching) should be stats only from the elite WC teams. I say this for this reason. While Chuck Hayes may frustrate a rookie like Blake Griffin, he is overmatched on both ends of the court by Dirk. A challenge to Dirk (or Chandler) is the only goal that matters. We could argue/discuss this until Hell freezes over. But IMHO give major minutes to players in the positive quadrant (Lowery, Battier, Hayes, Miller and Scola) without anyone named Martin and get ready for some serious Lottery balls. We already saw the protype for the comment. The tail end of last season w/o Martin but with Brooks having a MIP season.
While every stat will have its flaws and deficiencies, I believe when used in the proper context, adjusted +/- is one of the most powerful tools out there, as far as statistical analysis goes. Despite what some who are ignorant to it may think. The reason for this is that it measures the correct variable, or at least seeks to. The game is not about how many points you score or how many rebounds you grab, but simply to win. That is the end game, while everything else tries to get you there. What does every other statistical measure matter if they don't contribute to wins? This stat is at least trying to isolate it down to how/if you help your team win. The only thing that matters. With that said, it is no surprise that my favorite player on the team is Lowry and despise Jordan Hill with a passion.
Both your reasons are flawed. First your competition issue: we are comparing within the team. If Chck Hayes really does suck against good teams, then what does it say about the other guys on this team? If they're good against good teams then they must suck really bad against bad teams. It's simple math. Once again, we are comparing within the team. As for your lineup composition issue, you aren't reading the graph properly. Once you get from Lowry to the origin, you've reached mediocrity. At this point, anyone along the diagonal between Bud and Jeffries is equal in overall value. Anyway, I don't like splitting +/- into off and def anyway, it makes little sense.
So why is it that Hayes and Battier can't give Martin a lift on D? That's fine, I like the idea. Bad example for your reasoning though. 1. You discredit Hayes by saying he only frustrates a "rookie", while he does the same thing to elite big men like Amare, KG, Duncan. 2. Everyone looks overmatched against a HOFer like Dirk. 3. Chuck doesn't look overmatched against a HOFer like Dirk. In fact Dallas closed the game last night with Dirk on the bench and going small, because he was ineffective against the Chuck. Just say, we should only look at elite teams in the west because to win the championship we have to go through the elite teams in the west. Period. No need for pot shots.
There's nothing to argue there. You don't construct a team by only looking a offense/defensive +/-. Its one thing to look at to help gauge how players are performing within their roles. But you still need players with complementary skill-sets and roles.
And another thing I forgot to add. Nobody said the top 5 players = starting lineup, that's another failure to understand the statistic. Imagine if AB, Lowry, Martin were shown to be the 3 best in this stat, that doesn't mean you start all three...
