[ESPN] John Hollinger reveals new statistics: VA and EWA

[durvasa]

Originally Posted by nolimitnp Durvusa will love this

Who's that guy? I thought it was ok, but it seems like he's just recycling old material. I could have sworn he's written about value above replacement before.

It's just a common sense way of looking at the boxscore. It summarizes the info in the boxscore in a consistent, easy to understand way.

However, I would say I've become much more skeptical over how useful PER and its derivatives like VA are over the past couple years.

There was a nice commentary on PER on TrueHoop today:
http://myespn.go.com/blogs/truehoop/...-on-Trial.html

[jedicro]

Originally Posted by BigBenito Agreed, numbers don't lie... there can't be anything false with pure data. (At least, that I know of? side tangent, I need to ask my math major buddy if data can lie.) But, I'd argue that numbers can mislead us just as easily as our eyes. Both give data that can be misinterpreted/extrapolated beyond what we actually 'know'.

2+2=4


Most of the time at least..

[meh]

Originally Posted by jedicro 2+2=4 Most of the time at least..

Well, even something as simple as 2+2=4 require context. In this case an assumption that it's in base 5 or higher. So numbers can always lie.

My personal beef with those who automatically throw out "stats are misleading" all the time is that THEY personally use misleading stats all the time.

People who criticize newer stats cite points/rebounds/assists numbers like they're from the bible or something.

[Steve_Francis_rules]

Originally Posted by michecon LOL, it's basically PER, numerically subtracting some averages and give out another number. What this really is saying is: you are more valuable as a SF and SG than if you are a PF etc, given the same PER. And in terms of Lebron/Kobe/wade comparison, it's saying whoever played more minutes matters more.

It's not just saying that whoever played more minutes is more valuable. If that were the case, Lebron wouldn't still be #1 in this category because he has played fewer minutes. It's trying to correct for the number of minutes played so you can get an idea of how valuable a player is.

[Steve_Francis_rules]

Originally Posted by iconoclastic Hmm, so the average power forward adds more points to a team than any other position?

This stood out to me as well. It kind of makes sense to me as well, although I may be thinking about it wrong. Among players who don't play that many minutes, wing players have the lowest average PER. I was thinking that perhaps that's because these days, when a team takes a risk on a young player that's not really that good yet, it's usually the really athletic, but not quite a basketball player type.

[Northside Storm]

Originally Posted by Steve_Francis_rules This stood out to me as well. It kind of makes sense to me as well, although I may be thinking about it wrong. Among players who don't play that many minutes, wing players have the lowest average PER. I was thinking that perhaps that's because these days, when a team takes a risk on a young player that's not really that good yet, it's usually the really athletic, but not quite a basketball player type.

I think a quick way of looking at it is seeing the young, athletic PFs who contribute very well in limited minutes, which every contending team seems to have around...

Yours truly, Carl Landry
Millsap
Powe
Brandon Bass
J.J Hickson
Josh Powell
Anthony Randolph
Channing Frye
Darrell Arthur
Ryan Anderson
Luc Richard Mbah a Moute

etc. etc.

[Christopher]

Stats are only as good as the people taking them down.

Anyway you'll never have stats for intangible things. Give me a stat for the guy throwing himself at a lose ball before it heads of out bounds or a stat for a lock down defender who just refuses to allow an easy pass and changes the way an offense is going and then I'll start listening.

Till then this stat revolution is good information, just just one part of how a player should be reviewed.

[michecon]

Originally Posted by Steve_Francis_rules It's not just saying that whoever played more minutes is more valuable.

Where did I say that's the case?

Since Lebron et al all play SG SF position, the only difference the formula makes in addition to PER is minutes played. It's called monotone transformation.

[Steve_Francis_rules]

Originally Posted by michecon Where did I say that's the case? Since Lebron et al all play SG SF position, the only difference the formula makes in addition to PER is minutes played. It's called monotone transformation.

That's what I took this statement to mean:

And in terms of Lebron/Kobe/wade comparison, it's saying whoever played more minutes matters more.

But Lebron comes out on top even though he's played fewer minutes, so this statement is not true.

[Easy]

Originally Posted by Christopher Give me a stat for the guy throwing himself at a lose ball before it heads of out bounds or a stat for a lock down defender who just refuses to allow an easy pass and changes the way an offense is going and then I'll start listening.

I wouldn't be surprised that Morey and his minions are tracking these things. Did you read the article where the reporter asked him about knowing players making a defensive stop at crucial time? Morey answered, "Why do you think we have Chuck Hayes?"

When people think of stats, they think of boxscore stats, or better, the Hollinger numbers or things from 82games. But there are a lot of other numbers they look at that we outsiders don't know. Intangibles are tangibles. You just need to know what to look.

Actually, the hard part is not getting the numbers. It is the analysis. How much value do that kind of hustle play add to winning games as opposed to say, getting an offensive rebound, or setting a teammate up for an easy basket?

It is easy to say, this guy hustles, or is tough, or has leadership. But how much does that translate to winning?

[DaronMalakian]

Originally Posted by Artesticles http://sports.espn.go.com/nba/column...PERDiem-090325 Here's the story explaining the stats, and why LeBron should be MVP. http://insider.espn.go.com/nba/holli...r%2fstatistics New stats already in use. In short, VA = Value Added is the estimated number of points a player adds to a team’s season total above what a "replacement player" (for instance, the 12th man on the roster) would produce. Value Added = ([Minutes * (PER - PRL)] / 67). PRL (Position Replacement Level) = 11.5 for power forwards, 11.0 for point guards, 10.6 for centers, 10.5 for shooting guards and small forwards. EWA = Estimated Wins Added is Value Added divided by 30, giving the estimated number of wins a player adds to a team’s season total above what a "replacement player" would produce. I'm not really sure how effective these new sabermetric stats will work out in a much more team oriented sport in basketball, compared to baseball. If you really believe the MVP should go to the most VALUABLE player, then you agree with Hollinger's assessment. But I don't think these stats prove who's the best player. Lakers I believe have a better supporting cast than the Cavs, so Kobe wouldn't be as valuable than LeBron's value on the Cavs. If Kobe and LeBron switched teams, Kobe's value would probably go up, and LeBron's would decrease.

That's it. Kobe maybe is better player, but the most valuable for his team? ...The team that acknowledge more his absence. That's pure concept of value, isn't?

[ImmortalD24]

John "Chimpy" Hollinger's statistics are Exposed:

Why Kobe is the MVP Part 1 March 25th, 2009 John Hollinger wrote an article on ESPN arguing that Lebron James should be the MVP. As usual, he came up with some arbitrary mathematical equations to back up his argument. Now if you know anything about me, you know that I think that Hollinger’s PER system is absolutely absurd. The unfortunate thing is that he still continues to use it, and so do his fans. I do have to say though, I’m pretty pleased with how many people are noticing its stupidity. Without getting into too much detail, I’ll explain the problem with his PER system. It’s a mathematical formula, and as such, if the results are off, people should abandon it. No one would be using the famous Pythagorean Theorem today if it only worked SOME OF THE TIME. Yet that’s what Hollinger’s system does. It only works some of the time. The problem with arguing against him in the past however has been the fact that the PER formula is not something that you can actually refute by bringing up facts. I mean don’t get me wrong, Hollinger’s system has a ton of flaws in it. » It doesn’t take into account intangibles » It sets the standard PER for every season at 15, a completely arbitrary number that Hollinger picked out » It doesn’t effectively take into account coaching styles, differences in eras or fatigue » It weighs some things more than other things and assigns arbitrary values to certain stats. The problem though, is that you can’t measure it against something empirical. It’s an individual ranking and you can’t very well measure individual PER’s against real life team success. There could very well be a correlation, but there’s no proof of a cause and effect relationship. That in a nutshell has been the genius of the PER system, you can’t refute it based on clear cut evidence, you can argue it based on simple logic, but can’t really argue against it using empirical evidence. (Of course you can’t prove it using empirical evidence either because it’s far too flawed, but his followers seem to ignore this). It’s a math formula, so to their worshipping minds, it’s got to be infallible. That’s where this new article comes into play, John Hollinger has effectively shot himself in the foot today. He did this because he took the individual PER number that was so hard to argue against (because basketball is a team sport) and he used it to create two new metrics: VA (Value Added) and EWA (Estimated Wins Added). He essentially took the PER and used it to come up with numbers that translate into something we can actually compare to, WINS. Therein, he shot himself in the foot. Now before I explain and show you why he basically made his entire formula null and void, and thus his entire Lebron James for MVP article indefensibly stupid, we have to agree on something. Math is not an opinion. Let’s all agree that if a Math formula doesn’t work, we drop it. We don’t continue to use it. Even in statistics, once it’s shown that a formula only-kind-of-sometimes-sorta-works people drop the formula. If you can agree with me that a mathematical formula should work all of the time than continue to read on. If you think that a statistical formula only has to work most of the time and the other times it has to be near the correct answer (hence a spread) then read on. But if you think that Hollinger is right, Lebron’s the MVP, and that all proofs against his formula don’t mean diddly squat, than leave. Still here? Ok let’s continue. Part one of his argument for Lebron centered around the idea that Lebron James may have the best individual season ever by eclipsing Michael Jordan’s record setting PER mark. Since Hollinger chose to introduce MJ, I’m going to use Michael Jordan himself to disprove Hollinger’s entire VA and EWE theory. According to Hollinger’s calculations for this season, these are the values given to the top 5 MVP candidates: What he’s essentially saying is that Lebron James leads the league in “PER” (his made up statistic that equates to Player Efficiency Rating). Lebron James also leads the league in “VA” (another made up statistic that equates to Value Added). Finally, Lebron James also leads the league in “EWA” (Another made up statistic that equates to Estimated Wins Added). Value Added and PER are difficult to measure in real life because they are individual characteristics, but EWA is very easy to measure. Just look at the teams record. If the team misses a key player for the majority of, or the entirety of a season, see how much worse they actually do without this key player. This is another reason why I’m using Michael Jordan. He missed the majority of the season in 1985-1986 and the entirety of the season in 1993 - 1994. Before I show you the breakdown of Michael Jordan’s numbers I want to explain briefly how, once finding out the player’s PER, Hollinger finds Value Added and Estimated Wins Added. To find VA (Value Added) Hollinger uses the following formula: Total Minutes Played in a Season * ([PER - PRL]/67) PER in this equations stands for “Player Efficiency Rating” PRL stands for “Player Replacement Level” (which Hollinger puts at 10.5 for both Lebron James and Michael Jordan) So for an example, lets take a look at the 1990-1991 season. In this season Michael Jordan: » Played 3,034 total minutes in the regular season » Had a PER of 31.6 So to calculate Michael Jordan’s 1990-1991 VA (Value Added) you would: » First: subtract 10.5 (PRL) from 31.6 (Jordan’s PER). The result would be 21.1 » Second: divide 21.1 by 67. The result would be 0.31 » Third: Multiply that number (0.31) by total minutes played (3.034). The result would be 955.48 According to this formula, Michael Jordan’s Value Added in the 1990-1991 season would be 955.48 Following me so far? If not, review that again. If you are, we need to do one last step, find the EWA. To do this, Hollinger explains that you simply need to take the Value Added figure and divide that by 30. So Michael Jordan’s EWA in the 1990-1991 season would be (955.48 / 30) = 31.85 Using this formula, Hollinger concludes that Michael Jordan was worth roughly 32 wins to his team in 1991. I followed this formula in order to find the results of Michael Jordan’s PER, VA, and EWA results for two sets of seasons. The two sets of seasons revolve around two events in Michael Jordan’s Career. Note: As you will see I left out 94-95 season because of all the injuries the Bulls sustained that year, and because Jordan came back to play part of that season. Since Jordan was not in peak physical condition and because the Bulls sustained a ton of injuries, I left that season out. The First Set includes 4 seasons from 1983-1987 This set starts with the Bulls’ season right before MJ was drafted. I do this in order to see how many wins the Bulls had directly before drafting Michael Jordan, and how many wins Michael Jordan gave them the season he was drafted. The second important comparison here is to see how many wins Michael Jordan gave the Bulls in his third season after missing the majority of his second season due to injury. As is very noticeable, there is a HUGE discrepency between how many wins Hollinger THINKS Jordan is worth, and how many wins Jordan was actually worth. If we take Jordan’s average EWA over his three seasons (including the dismal 4.33 which is a result of playing only a handful of games that season) we reach 20.96. Yet the average amount of games that the Bulls won with Jordan is 36. Before Jordan was even drafted the Bulls had won 27 games. That’s a difference of 9 games. Hollinger’s formula predicted 21 games. This means that Hollinger’s formula was 43% correct. In other words, Hollinger’s EWA formula should say “plus or minus 57%” at the end of that. That’s a pretty bad spread. Ok lets take it on an individual season. In fact, let’s take it on the BEST individual season. Let’s compare the Bulls before they got Jordan (because that was the year with the lowest win total) with the 1986-1987 season (because that was the year with the highest win total). Ok so, In the 1986-1987 season Jordan had his highest EWA ever, a shocking 34.61. That means that the Bulls should have won 34-35 more games than they did before they even drafted Jordan. In reality however, The bulls sans-Jordan got to 27 wins, and the Bulls in that season got to 40 wins. That’s a difference of 13 wins. So that means that Hollinger’s system was 38% right. That’s a spread of 62%. Would you EVER higher this guy for your statistics predictions? His estimate missed the mark by 62%. The Second Set includes the 4 seasons from 1990-1994. This set starts with the Bulls first championship season and goes on to the Bulls first season without Michael Jordan. Of particular importance in this second set is seeing the difference between the amount of games that the Bulls won during their championship run with MJ, and how many they won during the season that he missed entirely due to his retirement. In all probability the second set should produce more Hollinger-friendly results. The Bulls were winning championships with Jordan in the lineup and his EWA’s weren’t as high as they were in 1986-1987…The EWA accuracy actually gets worse though! As is very easily noticeable, there is a huge discrepancy between what Hollinger says will happen and what actually happens, again. Hollinger claims that Michael Jordan has an Estimated Win Addition of 29.30 during the 1992-1993 season. If you look at the following season without Michael Jordan, the Bulls only lose 2 fewer games than the previous one with a healthy Michael Jordan. That means that Hollinger was 7% right. His spread for this perticular year was 93%. If you look at the average Michael Jordan EWA value during his first championship three-peat run of 29.23 games, you don’t see ANY season during which Michael Jordan adds 29 wins to the 55 win record the Bulls had in 1993-1994 without him.Basically, no season from 1990-1991 until Jordan retires in 1993 do the Bulls win 29 more games then they did during the Jordan-less season. Yet that’s what Hollinger’s system predicts. Even if you take the 72-10 season, the season with the most wins in NBA history…the math still doesn’t add up. MJ’s EWA using Hollinger’s formula for the 95-96 season is 29.6. The difference between 72 wins and the 55 wins that the Bulls had without MJ is only 17. 17 more wins, when the Bulls should have had 29 more wins. That’s barely 60% of Hollinger’s quota, AND THAT’S THE GREATEST SINGLE SEASON WIN RECORD OF ALL TIME. So in Hollinger’s most accurate comparison, his formula was only 60% correct. I’m not a math guy, but plus or minus 40% is a pretty pitiful spread. If you do what every statistician is supposed to do, and take an average, it’s actually pretty pitiful… Jordan’s average EWA during the 3 seasons in the chart is 29.23. The average increase in wins that the Bulls achieved in the three seasons they had MJ versus the one season they didn’t have MJ is only 7. So, divide 7 (number of wins Jordan is ACTUALLY worth on average) with 29.23 (number of wins Hollinger thinks Jordan is worth on average) and you get a measily 24% Yup, Hollinger’s EWA system should have an asterisk after it that says “plus or minus 75%”. That right there, is perhaps the WORST SPREAD in statistics history. Wait, no it’s not, the 93% spread that Hollinger’s formula resulted in between the 92-94 seasons is the worst spread in history. Congrats Hollinger, you’ve managed to out-stupid yourself yet again. Now, notice that I didn’t even argue FOR Kobe, I just argued AGAINST Hollinger’s logic and his formula. By this point, if ANYONE accepts these stupid formulas of his, please go get your head checked.

[Christopher]

Originally Posted by Easy I wouldn't be surprised that Morey and his minions are tracking these things. Did you read the article where the reporter asked him about knowing players making a defensive stop at crucial time? Morey answered, "Why do you think we have Chuck Hayes?" When people think of stats, they think of boxscore stats, or better, the Hollinger numbers or things from 82games. But there are a lot of other numbers they look at that we outsiders don't know. Intangibles are tangibles. You just need to know what to look. Actually, the hard part is not getting the numbers. It is the analysis. How much value do that kind of hustle play add to winning games as opposed to say, getting an offensive rebound, or setting a teammate up for an easy basket? It is easy to say, this guy hustles, or is tough, or has leadership. But how much does that translate to winning?

The BS Report on ESPN has a good podcast with Morey himself on just this subject.


Sure they have stats on players that are ridiculously in depth, but I still think that there are a lot of things you simply can't put a number on.



Actually, you just got me thinking. I wonder if they have a flopping stat! :D

[durvasa]

Originally Posted by ImmortalD24 John "Chimpy" Hollinger's statistics are Exposed:

That's pretty good, and you're right. His EWA is very, very rough as an indicator of how many wins a player adds. Basketball is too complicated to be able to estimate something like that just based on individual numbers and minutes played. A suggestion I've seen mentioned elsewhere for improving his formula is to take calculate how many VAs the player have relative to all his teammates, and then award that proportion of the wins his team got. It's still essentially would be based on PER, but it gives the player extra or less credit depending on how many wins his team got. Based on that, LeBron would lead the league by an even wider margin, because his team has more wins than any one else.

That said, your article is mistitled. It's called "Why Kobe is the MVP, Part 1." I didn't see any argument put forward as to why Kobe is the MVP. If we all acknowledge that Hollinger's formulas are very noisy indicators of whatever he considers important in an MVP race, I'm interested to see the superior evidence on the other side supporting Kobe. After all, neither is the MVP by default. Whoever eventually wins, there needs to be a good reason why they are more deserving than the other.

[JeopardE]

Originally Posted by Nuggets4 Can someone honestly explain to me why Hollinger's statistics are met with skepticism and such hostility? I'm not saying anyone in this thread has (yet), but it seems like every time Hollinger's name comes up in a discussion it's met with "HOLLINGER'S JUST A NERD WHO CAN'T PLAY!" or something equally asinine. This happens on the Nugs board I'm on as well. I don't get it. He's not saying that his stats answer EVERYTHING, but they're additional tools to use in rating players. Why are people threatened by this stuff?

It's the "earth is flat" syndrome. It's the same mob mentality that led people want to throw Galileo in jail. Like someone said, a lot of people *are* stupid. Not all of Hollinger's critics are stupid, mind you -- there are a few who actually have cogent and substantial arguments against his work. But the majority of them are plainly idiots who find that their brain hurts too much when math is presented to them, so they find it easier to just shoot the messenger instead. On the Internet, especially, it is rather rare to find a critic of a Hollinger article who actually bothered to even read through the article he is criticizing.

[JeopardE]

Originally Posted by durvasa That's pretty good, and you're right. His EWA is very, very rough as an indicator of how many wins a player adds. Basketball is too complicated to be able to estimate something like that just based on individual numbers and minutes played. A suggestion I've seen mentioned elsewhere for improving his formula is to take calculate how many VAs the player have relative to all his teammates, and then award that proportion of the wins his team got. It's still essentially would be based on PER, but it gives the player extra or less credit depending on how many wins his team got. Based on that, LeBron would lead the league by an even wider margin, because his team has more wins than any one else..

That is interesting. I too, at first glance, was surprised at the coarse simplicity of the EWA formula. I'm not really attaching too much importance to it myself -- for all the weaknesses of win shares and other related approaches that have become commonplace among basketball stat geeks, none of them are this pedestrian, and none are wholly based upon an inherently flawed box score.

[Like A Breath]

I don't get it. So if we can't go by normal statistical means (where LeBron blows Kobe away), and we can't go by "advanced" statistics (where LeBron still leads comfortably), and we can't go by team record (where they're basically tied, with Kobe having Pau Gasol)...what argument is there for Kobe over LeBron other than "he seems better when he's playing."

Think about it -- if you had no stats, wouldn't you think that Kobe was like 15% at his 3's than LeBron? I would, but you can't deny the fact that their 3 point percentage is nearly identical with a gigantic sample size for the season.

I need 1 person to tell me why Kobe is more deserving than LeBron other than "he just is a better player, watch the games." I still haven't seen anything else that suggests that he's having a better year.

[JeopardE]

Originally Posted by Like A Breath I don't get it. So if we can't go by normal statistical means (where LeBron blows Kobe away), and we can't go by "advanced" statistics (where LeBron still leads comfortably), and we can't go by team record (where they're basically tied, with Kobe having Pau Gasol)...what argument is there for Kobe over LeBron other than "he seems better when he's playing." Think about it -- if you had no stats, wouldn't you think that Kobe was like 15% at his 3's than LeBron? I would, but you can't deny the fact that their 3 point percentage is nearly identical with a gigantic sample size for the season. I need 1 person to tell me why Kobe is more deserving than LeBron other than "he just is a better player, watch the games." I still haven't seen anything else that suggests that he's having a better year.

The one thing people have in the back of their minds is Kobe's supposedly superior clutchness, but even that is debatable: the numbers say that while LeBron doesn't score (or shoot) as much as Kobe does in clutch situations, he actually scores more efficiently (eFG of 60.3% compared to Kobe's 53.3%), and gets to the line more (17.2 FTM to Kobe's 15.1). Personally I don't think there is any evidence, visual or numerical, that truly validates a choice of Kobe over LeBron for the MVP award. What LeBron has done this season is truly amazing. It's not a slight to Kobe or Wade who, in any other year, would be worthy of the honors, but LeBron is just a cut above the rest of the field this year in my opinion.

[durvasa]

Originally Posted by Like A Breath I don't get it. So if we can't go by normal statistical means (where LeBron blows Kobe away), and we can't go by "advanced" statistics (where LeBron still leads comfortably), and we can't go by team record (where they're basically tied, with Kobe having Pau Gasol)...what argument is there for Kobe over LeBron other than "he seems better when he's playing." Think about it -- if you had no stats, wouldn't you think that Kobe was like 15% at his 3's than LeBron? I would, but you can't deny the fact that their 3 point percentage is nearly identical with a gigantic sample size for the season. I need 1 person to tell me why Kobe is more deserving than LeBron other than "he just is a better player, watch the games." I still haven't seen anything else that suggests that he's having a better year.

People think MVP is the same thing as best player. You could legitimately argue that Kobe is the "best player". So, if you could have one player on your side in the last quarter of a game 7, who do you think? I can understand why someone would take Kobe (I'm not sure I would, but I get it).

But I think MVP really is about who has done more for their team over the course of a season. By that criteria, you're right. It's hard to think of a reason why Kobe deserves it more.

[meh]

Originally Posted by Christopher Sure they have stats on players that are ridiculously in depth, but I still think that there are a lot of things you simply can't put a number on.

But you can put a number on "results". It's like when Morey scouted Battier. He didn't know exactly what Battier brings, but knew that Battier produces when he's on the court.

You don't have to be able to explain every phenomenon through numbers. But as long as you understand what kind of players win you basketball games, it's all good.