- pretty dang good I would say. Maybe those folks that came up with all this offensive and defensive per-possesion ratings know what they are talking about .
However interesting these stats are, they are useless to a coach because he can't tell his guys make sure the points per possession is low. So it is as effective of stat as saying average points for minus average points against. Coaches need stats(that directly contribute to points per possession) that they can translate into a game plan such as the things that Van Gundy put up on the wall of the practice court. Gang rebound, force opponents in taking one contested jumpshot per possession, low turnovers, limit the fouls, play inside out.
Sure you need to know possessions per game in order to calculate points per possession. However, nobody can determine possessions per game, given points per possession. FG% and PPG are not necessarily correlated with each other in all the cases, but they are definitely not antagonistic to each other (an example of a pair of antagonistic stats is defensive rebounding and offensive rebounds allowed). More often than not, a team ranked highly in one category is expected to do well in the other. I could've just said points per game allowed alone is enough. By including FG%, I intended to stress that this stats is perhaps the one of the most, if not the most, important factors contributing to the overall defensive strength. I could've also said points allowed per game along with defensive FG% and defensive rebounding, in all of which the Rockests are ranked No. 1, paint a more complete picture of the Rockets defensive prowess. The question is, why would the Spurs allow their opponents to have 2 more possessions to score on them? The current debate appears heading towards the same stalemate I had with durvasa some time ago in the Yao POM thread, although it's much more relevant to this thread. There is nothing intrinsically wrong to derive secondary stats such as the per possession stats. The problem lies in how one views and applies these stats. The game as we know is played on 48-minute basis, not by a fixed number of possessions. A team geared towards fast-paced game is inherently neither superior nor inferior to a slower team. The point differential is all that matters. Although the Spurs allowed 3 few points per game than the Grizzlies, the former couldn't/didn't limit their opponents to a lower number of possessions such that these two teams were no different in terms of the overall defensive achievements measured by the ultimate indicator -- points allowed per game. Per possession-wise, the Spurs' defense may be more "stiff," but it didn't really matter since San Antonio couldn't put a stranglehold on opponents who would score on more opportunities, as opposed to the Grizzlies. Sure it does factor in pace, but by doing so, it artifically "deflates" the stats of those teams that play on faster paces. It could mean SAS are better 3-point defender than MEM, but that's a different issue. True, FG% alone is not enough. But points per possession data is insufficient, either. Well, the perception that Spurs is a better team overall couldn't be mistaken.
Certainly. I was more making an observation on the consistency of winning . If you're building a team and you had the choice between a great defense and a great offense, which is going to be more consistent in producing results? The offensive ppp had a higher correlation with winning over the last 5 years, but also more deviation in results. Obviously, the best teams are going to be great on both ends of the floor, but if you had to pick one side to favor, which would it be?
Gotta tell you that what Phoenix has accomplished the last two seasons is better than we've done with JVG in that same time. Yes, we've been injured during both of those years, but still...Pheonix's high octane offense should not be dismissed as ineffective because it hasn't yielded a championship.
Actually one thing points/poss leaves out is the teams ability to defensive rebound. A team can stop you on multiple possessions but if they dont box out and the other team scores you can have a low points/poss, low FG% allowed, but a high PPG allowed.
I think we've established ppp as the ultimate metric for measuring offensive and defensive effectiveness. Of course, it can't measure intangibles like have Kobe Bryant or T-mac on your team. Thank goodness or basketball would become a boring statistical sport. That said, a lot of people say the notion that defense wins championships is not true or that era is over.... Let's not forget the Heat beat the offensively superior Mavs And the Pistons didn't do too shabby either before them. Nor the Spurs. Fact is, you have to go back nearly two decades to find an NBA champion that didn't have the top or near top defense. The late 80's lakers..... Good luck Phoenix. But Steve Nash ain't going to the finals.
LOL ... Tango, nice try. But I have my version of analogy (which I think is better ): Let's say we have two cones of exactly the same shape and height. There is a race between two miniature cars, each of which is to travel from the bottom of a cone to its apex. By design, car A is faster in near-horizontal linear velocity, but its upward mobility is less than that of car B. To reach the apex, each car has to travel spirally upward on the surface of the cone. Apparently, car A would have to travel more distance (both angular and linear) than car B to reach the same height on the cone, but keep in mind car A is a faster car. The race is judged by which car reaches the cone apex faster. In the case of 2005-6 SAS and MEM, the former is car A whereas the latter is car B. The result is both cars reached the apexes at about the same time.
You got it backwards. By NOT factoring in pace, the stats of those teams are artificially inflated. PPG + FG% allowed does not tell you as much about a team's defensive prowess as points allowed per possession. A team can rank highly in those categories by slowing the game down sufficiently enough (by using up their shot clock and extending their offensive possessions with offensive rebounds) and forcing a lot of misses. But if they don't force turnovers, they don't defensive rebound, and they commit a ton of fouls, their defense is still going to suck. I mentioned in a previous post two criteria for good offensive and defensive metrics. (1) they should be independent of eachother (2) both should correlate well with winning I established that the per possession metrics (points scored per possession, points allowed per possesion) satisfy both of these criteria better than the per game metrics (points scored per game, points allowed per game). We can evaluate your proposal of looking at PPG + FG% in the same way. First, we need to somehow combine the two into a single number. I do this by normalizing each metric, and taking the average of the two. By "normalize", I mean that for a particular team in a given season, I look at the number of standard deviations from the mean for each stat (ppg and FG% for offense, PPG and FG% allowed for defense). For example, for the 2006 season: Code: Team off/g def/g FG% dFG% offmetric defmetric -------------------------------------------------------------------------- atl 0.052 -1.112 -0.016 -1.475 0.018 -1.293 bos 0.236 -0.558 1.055 -0.116 0.646 -0.337 cha -0.035 -0.863 -1.695 -1.484 -0.865 -1.174 chi 0.197 -0.036 -0.630 1.699 -0.217 0.831 cle 0.143 0.370 -0.015 -0.082 0.064 0.144 dal 0.528 0.881 0.704 0.630 0.616 0.756 den 0.821 -0.691 0.615 -0.013 0.718 -0.352 det -0.041 1.530 0.093 0.105 0.026 0.818 gsw 0.366 -0.634 -1.658 -0.175 -0.646 -0.404 hou -1.711 1.195 -1.660 1.485 -1.685 1.340 ind -0.770 1.121 -0.779 1.115 -0.775 1.118 lac 0.046 0.310 0.874 1.136 0.460 0.723 lal 0.601 0.016 -0.069 0.233 0.266 0.124 mem -1.195 1.910 -0.434 1.071 -0.815 1.490 mia 0.712 0.220 1.977 0.845 1.345 0.532 mil 0.197 -0.410 -0.037 -0.724 0.080 -0.567 min -1.304 0.761 0.146 0.796 -0.579 0.778 njn -0.795 1.025 -1.125 0.903 -0.960 0.964 nyk -0.340 -1.125 0.082 -0.783 -0.129 -0.954 noh -1.036 0.308 -1.072 -0.301 -1.054 0.003 orl -0.514 0.226 1.479 0.003 0.482 0.115 phi 0.579 -0.962 0.383 -0.551 0.481 -0.756 pho 2.806 -1.300 2.008 -0.029 2.407 -0.664 por -2.018 -0.287 -0.674 -0.898 -1.346 -0.593 sac 0.456 -0.069 0.061 -0.015 0.258 -0.042 sas -0.355 1.847 1.471 1.232 0.558 1.540 sea 1.375 -1.922 0.387 -1.926 0.881 -1.924 tor 1.001 -1.576 0.010 -2.244 0.506 -1.910 uta -1.150 0.452 -0.938 0.269 -1.044 0.360 was 1.149 -0.626 -0.544 -0.707 0.302 -0.667 The offmetric and defmetric columns are the offensive and defensive ratings strictly using these stats. Ok, here's the correlation table including these two new metrics: Code: [SIZE=2] off/poss def/poss off/g def/g offmetric defmetric pace win% ----------------------------------------------------------------------- off/poss 1.000 def/poss [B]-0.039[/B] 1.000 off/g 0.846 -0.158 1.000 def/g -0.173 0.831 [B]-0.522[/B] 1.000 offmetric 0.863 -0.045 0.909 -0.334 1.000 defmetric -0.118 0.906 -0.372 0.934 [B]-0.194[/B] 1.000 pace 0.224 -0.202 0.697 -0.701 0.507 -0.499 1.000 win% [B]0.684 0.641 0.490 0.441 0.593 0.537[/B] 0.005 1.000[/SIZE] As expected, they satisfy both the above criteria better than simply looking at points per game and points allowed per game (Off/g, Def/g above). But, still, points per possession and points allowed per possession are superior. They are more independent of eachother, and they each correlate with winning% better.
Umm, I think that's a misconception. If your team doesn't grab the rebound off a missed shot, the opposing team still maintains the same possession. In other words, poor defensive rebounding doesn't result in more possessions for the opponents.
How can raw data be artifically inflated? You only inflate/deflate secondary data through "manipulation" of primary (raw) data. Among other things, defensive fouls committed are not factored into opponents' FG%.
How do you define "raw data"? There is nothing inherently more raw about looking at FG% (field goals made per field goals attempted) versus points per possession. Even PPG is a "manipulation" -- you take all the points a team scores in a season and divide it by the number of games they played. Are saying that a possession is somehow not connected to something real? That's obviously not true. Just because it isn't in the official stat sheet, that doesn't mean it's just a mathematical construction. Anyways, I added additional stuff to my earlier post to analyze the PPG+FG% metric in more detail.
He doesn't have any idea about what he is talking about. Raw data is what the score keeper tabulates. Free throw, foul, turnover, etc. Everything else is a metric off of the Raw data. FG% is a stat. So is Avg pts per game. So is Points/Possesion (ppp). All stats. The data is sufficient to calculate all of those stats. The meaning is how the data is interpreted. The significance is based on sample size, variability, and the differences in results. Points per possesion takes everything into account. Turnovers, steals, blocks, rebounding, fouls, scoring, free throws - everything. That's why it's the best overall metric. It takes pace out - if you want to see the effect of pace just look at ppg. This isn't Rocket science guys!
But it's a misleading stat. If AI scores 32 points against the Rockets on 30 shots, and Yao scores 30 on 22 shots, who was the better scorer? Please don't choose AI.
Then what does that say for the Suns, whose defensive accomplisments have always been given the short stick when pace isn't attributed for, though they're actually a top 10 defensive team? Seriously, even without these stats to help me, or even League Pass, I could tell you that FG defensive % can be mis-leading. Just watching the hometown Bulls from the last two years, who never forced turnovers, who sent teams to the line more than any other team in their conference, and who had to deal with big guards getting hot from behind the arc over their small backcourt -- all these things added to points and more points against the Chicago defense without having one iota difference on the FG % defense. Even when teams shot 40 percent from behind the arc, the Bulls would lose, but the FG % defense would go down. There is NOTHING more important than POINTS.
Your point makes sense, but it is impossible to tell whot he better scorrer is. How many of the next 10 shots Yao takes would he make? Could he even get 10 more shots up? The point is that there are so many variables out there, that no one variable encompasses everything. You have to look at them all, objectively look at what you're trying to measure, and go from there.
Technically NewYorker is correct, that is, none of the "per" (including % stats, which is a form of normalization) metrics should be considered raw data. Although both PPP and PPG are obtained by the same kind of arithmetic operations and both denominators in these operations are themselves raw data, their meanings have different bearings on an NBA season. Teams' records are ultimately tied to points scored/allowed in games played, not to the "rates" at which teams score and lose points. Therefore, per-game stats are intuitively more fundamental than per-possession stats. I took some time looking at your correlation results. They are surely interesting but I do have a few questions. 1) Are the two criteria preconceived or merely an after-thought? 2) Could you please define or clarify the meaning of win% in your correlation tables? Is 42-40 considered a winning record in your methodology? 3) Why poss is the same for both def and off? Doesn't each have its own number? 4) The correlation results seem to suggest as long as a team is doing fine in one aspect of the game, it will fare pretty well in a regular season? In other words, good off/poss or def/poss is like a sufficient condition for winning (note that off/poss is not correlated with def/poss)? 5) Since off/game (def/game) is basically the product of off/poss (def/poss) and pace, by looking at the correlation table of "Total (last five seasons)" in post #138, how do you reconcile the phenomena that off/game (def/game) is weakly correlated with win% and pace has virtually nothing to do with win%?
