APBRmetrics

mystic wrote:@motherwell Dallas does use WP48 and not adjusted +- as much anymore.

Hey, thanks for quoting me, but you know, maybe provide some context. Specifically:

When they signed Marcin Gortat to a contract – a good contract for a guy that never played many minutes – the only metric that “rated” him was WP48.

The context makes the quote make sense, no? And was there a way Dallas COULD trade Dirk for Humphries? Did their salaries ever match?

That is not surprising at all, because when a team commits a turnover there will score exactly 0 points in that possession. Each shot attempt has a higher probability to create points than a turnover. And that is a huge reason why shot creation can't be ignored (like WP48 is basically doing it).

That's an interesting question, but couldn't you work out the number of 24 second shot clock violations and attribute a value to the untaken shots? http://82games.com/random30.htm shows that there was an average in 06-07 of 1.14 per game TOTAL, e.g. .55 per NBA team per game. I can't find more recent stats sorry

Anyway, it seems like it isn't a large problem to me, but if the claim is that passing up shots HELPS a player, and taking them can HURT, I'm not sure that gels with the other data, e.g. that players get paid well for scoring (no matter efficiency). Seems a rational NBA player would willingly take those extra shots, even ones they shouldn''t take.

In any case, I wonder if criticisms like this are "orders of magnitude" problems, or if they are minor issues. Seems like all criticisms are laid out as "orders of magnitude" problems, but surely that isn't always the case?

I also find the specific player complaint arguments weird. Tolliver, Anthony is currently ranked 5th in 2 Year Adj. +/-. Tolliver for Lebron anyone?

motherwell wrote: I also find the specific player complaint arguments weird. Tolliver, Anthony is currently ranked 5th in 2 Year Adj. +/-. Tolliver for Lebron anyone?

I think the important difference here is that anyone who uses +/- stats would agree with you that it would make no sense to trade Tolliver for LeBron. OTOH, I am fairly confident in saying that Berri and his followers would have no problem recommending that transaction if Tolliver's WP48 was higher than LeBron's.

And honestly, I think that's all anyone wants Berri to acknowledge at some point, that WP is not a *perfect* metric, and that you can't simply use it in EVERY SINGLE SITUATION that comes up to claim that Player A is better than Player B simply because his WP48 is higher. But that's all you ever read on WoW. I've really never heard him admit a single problem with it (except 5 years down the road or something).

motherwell wrote: Hey, thanks for quoting me, but you know, maybe provide some context.

Are you Alique Williams? Because that is who I quoted!

motherwell wrote: When they signed Marcin Gortat to a contract – a good contract for a guy that never played many minutes – the only metric that “rated” him was WP48.

The context makes the quote make sense, no? And was there a way Dallas COULD trade Dirk for Humphries? Did their salaries ever match?

What? Gortat was listed with 17.1 PER and 0.200 WS/48. Both times clearly above average. He was listed after the 2009 playoffs with a +2.92 APM. My SPM had him with +1.2. There is NO indication that no other metric had Gortat "rated". He was above average and was supposed to be able to keep that with average minutes. His +/- numbers suggested better than average. That the Mavericks offered him a full MLE contract isn't something so out of the ordinary. Signing a guy supposed to be above average for the average salary is hardly something you can base on WP48.

Kris Humphries PLAYED for the Dallas Mavericks! He was TRADED away by the Mavericks in January 2010. At that time WP48 ranked him above Dirk Nowitzki. The Mavericks could have easily kept Humphries and his better WP48, but instead they decided to trade him to save some salary and take back Eduardo Najera, a guy basically having ZERO value according to WP48.

It makes no sense at all to assume that the Mavericks used WP48 in order to make decisions. It is quite absurd to think that given their history of moves.

motherwell wrote: That's an interesting question, but couldn't you work out the number of 24 second shot clock violations and attribute a value to the untaken shots? http://82games.com/random30.htm shows that there was an average in 06-07 of 1.14 per game TOTAL, e.g. .55 per NBA team per game. I can't find more recent stats sorry

Lol, what kind of argumentation is that? You just showed that the players, coaches, etc. in the NBA are not dumb. They rather shoot than commiting a turnover. That is pretty smart, because even after a missed shot the probability to score in that particular possession is HIGHER than after a turnover.

motherwell wrote:Anyway, it seems like it isn't a large problem to me, but if the claim is that passing up shots HELPS a player, and taking them can HURT, I'm not sure that gels with the other data, e.g. that players get paid well for scoring (no matter efficiency). Seems a rational NBA player would willingly take those extra shots, even ones they shouldn''t take.

I don't even know where to start. In any case WP48 would award players for passing up shots and instead letting the shot clock expire unless it is a 3pt shot or free throws. Why? Because a player has to make 51.5% of his 2pt field goals in order to break even. That is above league average. So, in average a 2pt shot is costing a team wins according to WP48.

motherwell wrote: In any case, I wonder if criticisms like this are "orders of magnitude" problems, or if they are minor issues. Seems like all criticisms are laid out as "orders of magnitude" problems, but surely that isn't always the case?

Well, to me it rather seems like that you don't understand the issue at all.

motherwell wrote: I also find the specific player complaint arguments weird. Tolliver, Anthony is currently ranked 5th in 2 Year Adj. +/-. Tolliver for Lebron anyone?

Why? If we take the minutes into account James adds more value to a team than Tolliver. I brought up examples in which the inferior player brought more value according to WP to a team. A big difference here. Let alone that Tolliver plays a different role than James, while Humphries plays, according to WP, the same role as Nowitzki, PF.

mystic wrote:

motherwell wrote:In any case, I wonder if criticisms like this are "orders of magnitude" problems, or if they are minor issues. Seems like all criticisms are laid out as "orders of magnitude" problems, but surely that isn't always the case?

Well, to me it rather seems like that you don't understand the issue at all.

Clearly. So please, enlighten me. I'm happy to listen because...

mystic wrote:

motherwell wrote:That's an interesting question, but couldn't you work out the number of 24 second shot clock violations and attribute a value to the untaken shots? http://82games.com/random30.htm shows that there was an average in 06-07 of 1.14 per game TOTAL, e.g. .55 per NBA team per game. I can't find more recent stats sorry

Lol, what kind of argumentation is that? You just showed that the players, coaches, etc. in the NBA are not dumb. They rather shoot than commiting a turnover. That is pretty smart, because even after a missed shot the probability to score in that particular possession is HIGHER than after a turnover. ;)p

Hold on, I make a really sincere attempt to understand your hypothesis - a very good hypothesis, one I'd never seen written before, and your reply is "lol"? Seriously? You raise an issue, that potentially has a lot of merit, but one which you don't see through, and rather leave without proof or exploring the validity of, and then when I try to understand the context of your statement, you come back with lol?

This is what bugs me about these discussions - the complete lack of follow through and evidence.

You start well: "a turnover is the worst outcome, much worse than a bad shot." I can agree 100% with that. But you need to finish that thought like this: "Here are some stats that show players avoiding shots to save their averages, and here are players that take shots that, although bad shots in theory, are actually positive plays. This is not distributed evenly across players, which shows the flaw in efficiency over usage". Instead, we have a statement that not shooting helps WP48, but no exploration of whether in practice that happens, if the flaw is disproportionately attributed, or what competing influences will affect it.

Again, as pay is highly correlated with scoring, but not efficiency, and as 24 second shot violations are very rare, a flaw in a metric that allows not shooting and causing a 24 second shot violation to be a positive, a flaw that is probably entirely accurate, probably doesn't manifest as any significant statistical variance in the metric across players (but i can't prove it doesn't - clearly). In other words, this flaw likely advantages, and disadvantages, all players equally.

Is that not a fair summary of what your position is, and a well worked through understanding of all the implications and likely influences? Because you do raise a good point, but you stop half way, and don't actually show that this flaw matters. A flaw that is unlikely to be disproportionately attributed isn't really a flaw that has significant implications.

But we've now got a falsifiable hypothesis. If you can show that some players pass up shots, or that others take a disproportionate number of bad but positive late shot clock shots, I'd be happy to look at that.

motherwell wrote: Hold on, I make a really sincere attempt to understand your hypothesis - a very good hypothesis, one I'd never seen written before, and your reply is "lol"? Seriously? You raise an issue, that potentially has a lot of merit, but one which you don't see through, and rather leave without proof or exploring the validity of, and then when I try to understand the context of your statement, you come back with lol?

Yes, because your answer makes no sense. And I suspect that you were not really interested in understanding it, but rather tried to dismiss it. That goes hand in hand with your other points. You could have known the things about the Gortat or Humphries situation, but you rather choose to make stuff up before checking out the reality. The same is seen here.

The theory says that shooting is better than letting the shot clock expire. The reality tells us that the players are going by the theory and trying to avoid a shot clock violation. You presented the evidence.
Your conclusion was: It is a minor issue. The issue is that WP48 suggests to let the shot clock expire, because that will improve the individual WP48 for that said player. Why? A shot clock violation is not listed as individual turnover. But the chances to convert a forced shot at the end of the shot clock is lower than average, which means a player collects negative WP48 by doing so. That is completely thought through and I bet I'm not the first one to ever mention such thing.

For you that is a minor issue, because the players are actually doing what the theory tells us and not WP48. And you want to use that to defend WP48. That is an absurd argumentation in which cause and effect are completely confused. The low amount of shot clock violation is the result here, not the cause.

motherwell wrote: You start well: "a turnover is the worst outcome, much worse than a bad shot." I can agree 100% with that. But you need to finish that thought like this: "Here are some stats that show players avoiding shots to save their averages, and here are players that take shots that, although bad shots in theory, are actually positive plays. This is not distributed evenly across players, which shows the flaw in efficiency over usage". Instead, we have a statement that not shooting helps WP48, but no exploration of whether in practice that happens, if the flaw is disproportionately attributed, or what competing influences will affect it.

And that is a perfect example of what I mean. You have no clue what you are talking about, that is the problem. If you would even be close to have knowledge about it, you wouldn't even ask question like that. Check out 82games.com, there is the shot clock usage listed for each player. You will see that this widely differs from player to player. Some players are taken more shots late in the shot clock than others. And for sure that compromises their shooting numbers. Did you ever watch a game of basketball? Each team has that player, their go-to-scorer, who is asked to take the tough shot when the play is broken down. And we can even see that in the amount of turnovers when that said player is on the court and when not. Even Allen Iverson, who had a really low scoring efficiency, helped his team tremendously with avoiding turnovers. Last season Nowitzki took 47% of his shots in the so called crunch time, Kevin Love 25%. That is a big difference.
There is even a thread here about the shot selection in which the shot clock usage is discussed.

And you are really surprised that I answered with "lol"? Honestly, I can't even comprehend how someone tries to make an argument against all logic and all common knowledge we have. Do you honestly think that the shot distribution is equal for all players? How do you come up with such an idea?

Hey, folks. Kyrie Irving isn't rookie of the year. He isn't even in the discussion. His WP48 is too darn low. Jon Leuer or Kawhi Leonard would be more appropriate selections. Kanter should also be considered, even though he's playing 15 minutes per game. Or so WoW says.


http://wagesofwins.com/2012/02/01/kyrie ... -the-year/

I think I figured out the logic of not wanting to improve your metric. There's a perverse incentive to being wrong more than being right. If you're wrong about a player, people will find it surprising. You then just have to convince those people that your metric is not wrong, and that everyone else has it wrong. Conversely, if your metric confirms what most people already know, they tend to find it uninteresting. Human nature.

WoW is just a psychology experiment, demonstrating what people are capable of believing.

Mike G wrote:WoW is just a psychology experiment, demonstrating what people are capable of believing.

Yes, but unlike most psychology experiments, the people conducting the experiment are also the subjects.

Mike G wrote:Landry Fields is better than Nowitzki, Kobe, Duncan, etc ... Is that because someone has said he is a guard? And as a guard, he's the best rebounder in the league, by a big margin.
But if he's a forward (a la b-r.com), he's an average rebounder. Does his rating change, as though he's not as good?

mystic wrote: And that's how a player like Michael Smith can end up being the 7th best player in the league according to WP48 ...

Are you saying Dennis Rodman wasn't the best player in the world, right up to the moment he was unemployable?

IMO, Landry Fields is a good case study of the problem.

Virtually every box score system I have seen measures Cs and PFs as somewhat more productive than SGs at the other extreme. Part of that seems to be due to scoring efficiency and part of it is due to rebounding.

Some people aren't comfortable with the idea that some positions may actually be more productive than others. So they create positional adjustments or make up new values for each stat to equalize them. The problem with the former is that not all players play like the stereotypical player of that position. The problem with the latter should be obvious.

Landry Fields job isn't to shoot a lot of jumpers and help space the floor like the typical SG. Nor is he really a SG. He's actually a SF playing SG. He scores a lot on drives, back doors, off offensive rebounds, in transition and doesn't shoot many contested jumpers. So you have a guy who is going to rebound more and score more efficiently than the average SG. When you compare him to the average SG using a positional adjustment he's looks way better than he actually is.

It seems to me the only way around this is to simply accept the fact that Cs on average are more productive than SGs or do positional adjustments and deal with the occasional Landry Fields. Every attempt I have seen to adjust the box score values looks more like an attempt to get the stats to say what people already believe instead of the other way around.

Italian Stallion wrote: IMO, Landry Fields is a good case study of the problem.

Virtually every box score system I have seen measures Cs and PFs as somewhat more productive than SGs at the other extreme. Part of that seems to be due to scoring efficiency and part of it is due to rebounding.

With ASPM, every position is almost equally balanced.

DSMok1 wrote:

Italian Stallion wrote: IMO, Landry Fields is a good case study of the problem.

Virtually every box score system I have seen measures Cs and PFs as somewhat more productive than SGs at the other extreme. Part of that seems to be due to scoring efficiency and part of it is due to rebounding.

With ASPM, every position is almost equally balanced.

IMO none of the adjusted plus minus systems have been perfected. You need large samples. But once the samples become large enough, you more or less start to lose the improvement/decline of young and old players, impact of injuries etc...

It's not that I don't think they are a great addition to the picture (I think they are terrific), but in the end I have more faith in my own approach.

I have more or less come to the conclusion that basketball is sufficiently complex that the goal should just be to eliminate the major errors (like Fields in Wins Produced or low efficiency volume scorers in PER) and calculate a reasonable probable range of ability for each player. Then do a subjective analysis of his role on the team, teammates etc.. to weight those numbers up or down with simple pluses or minuses based on the conditions the player is playing in. That doesn't sound very accurate (and it's not), but I'd rather be approximately right than precisely wrong.

Of course many of you guys may have a few extra IQ points to work with than I do, but I've been down this road before with horse racing. The simpler methods seem to work better, even for the rocket scientists.

Italian Stallion wrote:
Of course many of you guys may have a few extra IQ points to work with than I do, but I've been down this road before with horse racing. The simpler methods seem to work better, even for the rocket scientists.

Your method doesn't seem "simpler" to me.

EvanZ wrote:

Italian Stallion wrote:
Of course many of you guys may have a few extra IQ points to work with than I do, but I've been down this road before with horse racing. The simpler methods seem to work better, even for the rocket scientists.

Your method doesn't seem "simpler" to me.

It is.

I basically use Berri's Win Score values, no positional adjustments, a break even point of 40% instead of 50% for scoring efficiency (I have a reason I chose that), assists at .67 instead of .50 (I have a reason for that too), DREB = .5.

I don't adjust individual players for defense, but I do look at pace adjusted team defense and accumulate insights on individual defense from various other stats and watching games.

My mathematical values produce fairly logical results and get pushed up and down based on my more subjective observations. I can't tell you precisely if player "A" is better than player "B" when they are reasonably close, but I never produce ridiculous results like Fields (Wins Produced) or high PERs for volume scorers. I think that's about as well as I can do with the intellectual ability and statistical knowledge I am blessed with other than some slight tinkering over time. But I sincerely think I get good results this way.

EvanZ wrote:Hey, folks. Kyrie Irving isn't rookie of the year. He isn't even in the discussion. His WP48 is too darn low. Jon Leuer or Kawhi Leonard would be more appropriate selections. Kanter should also be considered, even though he's playing 15 minutes per game. Or so WoW says.


http://wagesofwins.com/2012/02/01/kyrie ... -the-year/

I think I figured out the logic of not wanting to improve your metric. There's a perverse incentive to being wrong more than being right. If you're wrong about a player, people will find it surprising. You then just have to convince those people that your metric is not wrong, and that everyone else has it wrong. Conversely, if your metric confirms what most people already know, they tend to find it uninteresting. Human nature.

Both non-prior informed RAPM and new WP like Rubio over Irving. They both like Leuer. They differ on Leonard and Kanter.

Italian Stallion wrote:

EvanZ wrote:

Italian Stallion wrote:

I basically use Berri's Win Score values, no positional adjustments, a break even point of 40% instead of 50% for scoring efficiency (I have a reason I chose that), assists at .67 instead of .50 (I have a reason for that too), DREB = .5.

I don't adjust individual players for defense.

WinScore with different weights on scoring, rebounding, assists and a different approach to defense...it is another metric that I'd be curious to see results from but it might be better called your own as not that much major is the same as WinScore after these changes.