On Thursday I had the privilege of giving a seminar at BYU. The subject was the next book and it was fun talking with the students about some of the stories we will tell.
While I was having fun, though, the usual Thursday post in this forum was skipped.
To make up for this, here are a few items of interest (hopefully):
Superman vs. Shaq
Adrian Wojnarowski has written an interesting column detailing Shaquille O’Neal’s behavior towards Dwight Howard (and Kareem Abdul-Jabbar). Apparently Shaq believes there is a substantial gap between Shaq and Superman.
In an effort measure the gap, here is a ranking of every player who has ever played for the Orlando Magic.
Table One: Ranking the Orlando Magic (1989-90 to 2008-09)
As one can see, Superman tops the list. Fans of Shaq would note that Howard produced his wins in five seasons while Shaq only played four years in Orlando. In Shaq’s fifth season, though, he only produced 13.3 wins and this would not be enough to close the gap. Shaq did post a higher WP48 [Wins Produced per 48 minutes] in Orlando, but that’s primarily due to the fact Howard started playing at 19. If we look at what each player did from the age of 20 to 23, the WP48 of each player is essentially the same (not that a 0.329 vs. 0.306 is really that different in the first place). So it doesn’t look like Shaq can claim he was ever that much better than Howard.
Here are a few more observations from Table One.
- Tracy McGrady is ranked 4th in the history of the Magic. As one can see, once upon a time McGrady was a very productive NBA player. That is no longer the case today. With the Magic he posted a WP48 [Wins Produced per 48 minute] above 0.200 each season. He has not done this for the Rockets since his first season in Houston. And now that he is 30 years of age, we might suspect that McGrady is not likely to reach the 0.200 mark again.
- Nick Anderson is currently second on the list. The next players on the list who are still active with the Magic are Hedo Turkoglu and Jameer Nelson. Turkoglu would have to produced 58 more wins to catch Anderson, and given Turkoglu’s level of production and age, that seems unlikely. Nelson could catch Anderson, but he is going to have maintain his current productivity and stay healthy. If that happens (and those are big ifs as Nelson ages), Nelson will catch Anderson in five seasons.
- Scott Skiles is currently ranked 10th. Nelson should pass him next season, but I am not sure Skiles is remembered for being an average point guard.
Readers Explain Randomness
My last post on the random nature of the playoffs resulted in a number of comments that suggested the point of the post was being missed. While I was getting ready to post a reply, though, readers jumped in with comments explaining the role randomness plays in the playoffs.So rather re-write my argument, I thought I would just post some of these comments.
from Zach:
I thought it would be useful to post the wiki on Fooled by Randomness:“Taleb sets forth the idea that modern humans are often unaware of the very existence of randomness. They tend to explain random outcomes as non-random.
Human beings: overestimate causality, e.g., we see Mosques in the clouds instead of understanding that there are just random clouds that appear to our eyes as Mosques (or something else); tend to view the world as more explainable than it really is, i.e., we look for explanations even when there are none.
Other randomness misperceptions discussed:
Survivorship bias. We see the winners and “learn” from them, while forgetting the huge unseen cemetery of losers.”
I think it’s instructive because this comment thread certainly supports that the argument that many people are very resistant to the idea that randomness is a plausible explanation for unexpected events.
I also happen to think this is an important argument in case Cleveland’s management decides to over-interpret their loss and try to match up better (e.g. ditch the productive Varejao for a multi-talented but unproductive player like Al Harrington).
from Jim Glass:
I don’t see what all the fuss is about here. A lot of people seem too grossly over-estimate by how much Cleveland was supposed to be better, and so think some sort of special explanation for the upset is necessary.
Cleveland won only 7 more games than Orlando out of 82, in conference only 3 more out of 52. That means they won 1 more game per 13 played, or per 17 — yet there was only a max of 7 games in the series. With such an objectively very small difference between the teams each game between them was basically a coin flip with a coin just slightly weighed against Orlando.
Cleveland’s expected advantages were only one-half game or less out of seven.
Specifically, the full season W-L pcts give Cleveland an expected probability of winning a seven-game series of about 67% (about 63% by the confenence w-l %) which means Orlando had a good one-in-three chance of winning the series on the plain face of things — and one-out of-three ain’t any kind of historic upset. (There are many more sophisticated ways of projecting expected w-l but they give very similar results.)
So “Orlando won with matchups” … duh, of course they did! They were a 59-win team, of course they had some matchup advantages. That doesn’t change the chance element of the series at all. Cleveland had advantages too, which go ignored when they lose. Fans see who won, then look backward and come up with your explanation of why they won. Orlando’s matchup advantages paid off! (Except in the two games they lost.)
But if the only thing different in the entire series had been that Cleveland won all the close games determined by luck, so Cleveland had won in six, nobody would be talking about Orlando’s matchup advantages — even though they would have been totally unchanged. They’d be talking about how the great LeBron dominated, etc.
And have no doubt that those close games are determined by luck. Sports fans are loath to think that games — especially crunch-time playoff games, and thus championships! — are determined by luck and chance, so they come up with beliefs like “great teams win close games, with character, guts, top coaching …”, but nothing could be further from the truth.
Vince Lombardi’s record with the Packers in games decided by 7 points or less was exactly .500.
Bill Walsh’s with all his 49er championship teams was 43%. All baseball sabremetricians know one-run games have chance outcomes. As a Knicks fan back from the Red Holzman championship era of smart and great basketball, Isiah pretty much put me off the NBA so I haven’t followed it closely recently. But a few years back when I last checked the NBA for this, Detroit with the best record in the NBA was just a tad over .500 in close (5-pt or less) games, and Portland with the worst record in the league was very close to .500 too.
Great teams stomp on other teams, and split their close games when they don’t. Rotten teams get stomped on by other teams, and split their close games.
In a Cleveland-Orlando series, where the W-L differential between the teams is 0.085 — only 8.5 games per 100 — *every game* is fundamentally close. That is all the explanation needed for any outcome. Playoff series, as Billy Beane says about baseball, are basically weighted crap shoots. Then the fans and sportswriters afterward come up with all the dramatic rationalizations they want for what happened.
more from Jim Glass
in response to this statement “The nature of sports is that the ‘better team’ is the team who wins in any given competition (single regular season game or playoff series).”
This is exactly wrong. Ring Lardner knew “the race is not always to the swift … though that’s the way to bet”.
Upsets happen all the time, in the playoffs as well as the regular season. In baseball, the Cards won the World Series a few years back after an 82-win season. Were they the best team in baseball?
The nature of sports is that the goal is to win The Championship. You build as good a team as you can because that *increases your chances* of doing so. The best team has the best chance. But it does not guarantee winning.
The best team/competitor often doesn’t win, at Wimbledon, the Super Bowl, wherever. It is reality-denying to ignore the role chance plays in these competitions, and so deny that “the best” may lose.
And it is no slight to The Champion to admit it is not the best team/competitor — because the whole goal of building a good team is to win the championship … and if you win the championship with the 2nd best or 3rd best team, no matter … you have **achieved your goal**!!
Believe me, if you are at tennis player who goes into Wimbledon as the 16th seed and win it all, everyone will hail you as The Champion, nobody will think you are the best player in the world, that won’t bother you a bit, you’ll be proud as a peacock about it the rest of your life, and have every right to be.
It’s the same for the upset Champion in every other sport.
and even more from Jim Glass
in response to this comment: this comment thread certainly supports that the argument that many people are very resistant to the idea that randomness is a plausible explanation for unexpected events.
Yes, and this behavior although intellectually irrational is easily explainable both in evolutionary terms for people generally and for coaches and players in the sporting world in particular.
Say success is determined by a combination of things you can control, plus random events you can’t control (well, as it actually is). To maximize success of course you want to focus on the things you can control and forget the things you can’t, put the latter out of your mind. This is true even if the ratio of impact they have on your life is 10% what you do and 90% random chance. You still want to totally maximize what you get out of that 10%.
You don’t want to wander around distracted by all the random things that can do you in at any moment … how your success or (failure) is unfairly due to chance … how your rival’s success is due to dumb luck … etc. Wasting mental resources on such makes people more prone to failure — and thus more likely to be removed from the gene pool.
So natural selection has culled people to be largely blind to the randomness of life, and instead to see causation everywhere. Even, very often, where it’s not.
And pro football coaches never, never say to their teams, “You know guys, more than 50% of all NFL game outcomes are determined by random chance”, even though it is true.
and in response to this comment: I also happen to think this is an important argument in case Cleveland’s management decides to overinterpret their loss and try to match up better (e.g. ditch the productive Varejao…
Yes, while you want the players’ efforts focused entirely on what they can control, coaches and GMs had better know the difference between what they can control and what results from dumb luck.
If you are a coach you had better not punish/reward players for chance events, or you will be heading for problems.
I’m old enough to remember Vince Lombardi, and one of the curious things about him was that while he was so tough he would often tear his team a new one even after they won a one-sided game, after close loses he was supportive of his team. “Sometimes the clock runs out on you when you happen to be behind” was one of his sayings. If outcomes of close games indicated team character then Lombardi of all people would’ve raged after close loses — but it was about the only time he was philosophical. I mentioned he had a .500 career record in one-score games.
Similarly, if you are a GM and your team puts together a good record solely by winning more one-score games than anyone else in the league, you’d better not think “We’re really good, good teams win close games, this proves it, we can stand pat”, or you’ll be heading for a fall. And if your team puts together a top record in the league by stomping other teams all year, then gets eliminated from the playoffs in a tough, one-score game, against another good team, you’d better not think, “Damn, we have everything but character and leadership under pressure, so I’m going to fire the coach and shake up the line-up”, or you could be taking a knife to your own throat.
from Brian Tung:
One need only read the comments here to appreciate how difficult it is to understand randomness and statistical conclusions. Hell, I do it for a living and I don’t understand it nearly as well as I’d like.
I’m not even sure where people claim Berri is supposed to have “dismissed” matchups. All I see is him claiming that Orlando’s series win did not demonstrate that Orlando matches up better with Cleveland than vice versa. He specifically did not claim that Orlando does NOT match up better.
It’s a subtle point, but one I think must be understood to read this post properly: Just because something doesn’t happen to show that something is true doesn’t mean it isn’t; it just means it hasn’t been shown. I get the feeling that many people are inferring (incorrectly) that the result must be either all superiority, or all randomness, and that Berri is saying it is the latter. On the contrary, it is always a mixture of the two (in any interesting contest), and it is the latter that makes it impossible to reliably determine the former–at least in any timely fashion.
As an aside, I’m pretty sure that when folks like Barkley say that “the better team always wins a seven-game series,” they are not saying there’s something magical about seven-games series, as much as they’re defining “better” as “wins a seven-game series.” At least, that’s the way I’ve always understood them.
By the way, these were not the only good comments. Hopefully these comments help people understand the role randomness plays in the playoffs and the tendency people have to invent explanations after the fact. If not, perhaps Zach, Jim, and Brian (and others) can step in again and help clear up the confusion.
– DJ
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Ray
June 7, 2009
As a basketball fan, I’m apalled that Pat Garrity is next to last on this list. I remember during the Magic’s amazing run in the playoffs in 2003 that resulted in a crushing first round exit, Garrity was their rock all season. McGrady garnered the star status; Garrity carried the team on his back. He was a threat to pull up from the ‘W’ in Amway and splash in a 3. He ripped down boards with the furiosity of a tazmanian devil. He’d make a no-look pass to wide open teammates because defenses feared his long range scoring ability. You can take your Larry Birds and your Magic Johnsons. I’ll take Pat Garrity.
PJ
June 7, 2009
Does randomness play a role in the playoffs? Of course. Could Cleveland beat Orlando in a seven-game series? Almost certainly.
However, the post strongly implied that randomness was the only reason that Cleveland had lost — that, if these two teams were to play each other 23 times or 57 times or however many times, Cleveland would eventually win more of those games.
In fact, there is evidence in their head to head contests in both the regular season and the playoffs that suggests this is not the case — that suggests, as (yes) the conventional wisdom has it, that Orlando presents serious match-up problems for Cleveland and is likely to beat them head-to-head more often that not. Simply pointing to randomness and saying “there you go” seems lazy and not very interesting.
DR
June 7, 2009
Is Grant Hill overrated, or did he just have bad seasons drag down good years?
Will there be end of season full team lists like you did for mid-season?
brgulker
June 7, 2009
in response to this statement “The nature of sports is that the ‘better team’ is the team who wins in any given competition (single regular season game or playoff series).”
That was my original comment. I don’t think I communicated well what I was trying to say very well.
I wasn’t trying to deny randomness. I don’t know statistics well enough to do so.
But I was questioning how the word “better” was being defined…. or at least think about the word differently.
I understand that Cleveland was better all season against the rest of the NBA than Orlando was (although not that much better against the East). My point was simply to say that it’s the nature of sport that the ‘best’ team — like Cleveland, the team that’s the best with respect to the league — doesn’t always win.
The team who wins on any given night was better than the other team on that given night. Orlando was better than Cleveland during the regular season when those two teams played, and that happened in 4 of 6 games in the playoffs.
So my conclusion is simply that when Orlando and Cleveland play, Orlando has been the better of the two teams.
On any given night, the team that plays better will win. In any given series, the team that plays better will win (most of the time).
Obviously, there are exceptions. Rashard Lewis hits a corner 3. Lebron hits a fade-away three from the top. Lucky shots go in. But in the ECF, Orlando played better, and as I understand the nature of sports and the word ‘better,’ that makes them a better team than Cleveland when the two teams have played each other.
I think that’s a different claim that saying that randomness doesn’t exist or doesn’t affect sport, which is how I think Jim Glass and Dr. Berri understood my comment.
mrparker
June 7, 2009
David,
Can you make a post on coaches being idiots. Your work clearly states that JJ redick is a bum. Why on earth did SVG give him 27 minutes tonight? Is he trying to get too cute with this whole matchup thing?
Ray
June 7, 2009
mrparker, do you like how Redick blew the game for the Magic? Blowing a wide-open 3 and then turning the ball over the next time down. SVG had been good about screwing the natural order and playing his best guys, but he missed the mark tonight. And it sucks so badly that Lee missed that shot. It kills me to think Kobe’s ridiculous antics are going to be rewarded with a title. Screw Courtney Lee.
anon
June 7, 2009
it’s really weird that taleb’s book is being brought up in this context. taleb wrote a polemic against statistical analysis that flourished in wall street in the last decade. He’s not a behavorial economist, who is trying to show how people act irrationally and underestimate risk. He’s trying to show how statistical models fail to fully account for randomness and low probability catastrophic risks.
He’s not talking about the randomness berri is discussing, which is randomness that doesn’t refute the underlying results of his statistical model. He’s talking about randomness that shows the ultimate vulnerability of statistical models.
John Giagnorio
June 7, 2009
mrparker,
It was pretty fun to watch Redick self-destruct on the court, though. During the game I started thinking about NBA coaches. If they’re so important to a team’s success, why are the coaches paid so little with respect to the players? Are their salaries included in the cap? If not, one would think a true “impact coach” like Phil Jackson, someone with the “basketball knowledge” to play black holes like Derek Fisher and Luke Walton, would make money similar to what Kobe Bryant makes unless there was a huge market failure. More support for DBerri’s research on NBA coaches?
John Giagnorio
June 7, 2009
anon,
Berri’s post showed a basically perfect understanding of what Taleb was writing – people coming up with (possibly) nonsense explanations for events after the fact. Please explain exactly where he misquoted.
John Giagnorio
June 7, 2009
This is also the only sports statistics-related website I know of that has ever discussed power laws (the 80/20 rule for Wins Produced), another focus of Taleb’s writing.
Ray
June 7, 2009
John, Coaches aren’t in the cap. Only players.
Also, it really bothers me that the 1 time a game that Derek Fisher goes to the free throw, the announcer, as contractually obligated, says that Fisher is one of best free throw shooters in the game. Ah yes, as indicated by his 81% career average. Not too long ago, in 2004 (the year before Golden State got him) he shot 79.7% from the line. Fisher is a good free throw shooter, but when I think best in the league, I think Ray Allen, Dirk, Nash. You know, guys who shoot in the 90s, and go to the line more than once. There’s a slew of players who go out and shoot 84-85% just like Fisher. Don’t say one of the best.
John Giagnorio
June 7, 2009
Ah but Ray, who else can “handle the moment” quite like Fisher? ;-) That’s worth another 15% from the line.
Ray
June 7, 2009
More Derek Fisher ranting…
He’s always regarded one of the class acts in the league. Also, not really that true. He got way too much credit for visiting his daughter when she had eye cancer, during the playoffs. His daughter had EYE CANCER! What decent human being wouldn’t visit their daughter with eye cancer. Later that night, he played back-up point guard for the Jazz.
Fisher is aways arguing calls, giving “the look” to refs, and you remember the dropped shoulder he put on Scola. Suspended. So, he’s basically the normal NBA player. He doesn’t go out and form everlasting peace like a great ambassador. Derek Fisher, I cannot stand. The most inaccurately described player, maybe in the history of sports.
Ray
June 7, 2009
John, it’s easy to handle the moment when you’re the 10th best player on your team and you don’t have nothing to do. If I knew that Kobe, Pau, Odom, and Ariza were going to handle absolutely everything, would I even break a sweat? Nope, I’d just smile and act like the wiley vetern. Wiley vetern is code for a sucky player that makes 1 play that he probably wouldn’t make in 10 more tries.
Ray
June 7, 2009
Don’t have anything to do, sorry. I knew I’d get bashed if I didn’t correct that.
simulator
June 7, 2009
There’s no doubt that statistical models are very useful, but there’s a huge difference between a flip of a coin and complex statistical models. Games like basketball are not simple binomial distributions.
The criticism of the majority of people has been that the basis of Berri’s assertion is just way too simplistic. And of course, real world events are not that simple.
I’d be more inclined to believe that baseball is heavily affected by chances (we’ve seen quite a few different teams winning the world series in recent years). But not in basketball, we’ve seen a lot of champions repeat quite a few times in the last 20 years.
mrparker
June 7, 2009
well, Orlando beat the Lakers twice this season…where are all the Matchup problems that Lakers must have?
Ray
June 7, 2009
They won those two games by a combined total of 9 points. Looking at these players, the Lakers can match the Magic’s length, and attack them where they’re weak, which is making Lewis defend a post player. Nice job trying to get all cute, mrparker, but I’m not buying it/
Mike G
June 8, 2009
I’m with PJ.
It’s true that many readers misunderstand randomness.
However….
Is Professor Berri arguing that there’s no plausible scenario where a team which is empirically better overall (using his model or any other) is worse against a specific team over 100 or 1000 games?
Basically, instead of arguing that the Cleveland result is likely the result of chance OR possibly the result of match-ups, you’re EXCLUDING the latter.
JohnG
June 8, 2009
Mike,
No, Professor Berri is arguing that the result *could* be chance, he is not necessarily excluding any other factors. He (and others) are simply pointing out that based on efficiency differential alone, Orlando had at least a 1/3 chance of winning the series based on chance alone. As such, it’s hard to say with any confidence whether or not the match ups matter. The match up argument is very tough to quantify, and Professor Berri generally refrains from talking about the game in anecdotes, so he has not talked about it much which I think is causing the confusion.
mrparker
June 8, 2009
Ray,
thats 4.5 pts a game. Thats not a small amount. I guess we’ll have to hear about the matchup problems if Orlando wins more than 2 at home.
Daniel
June 8, 2009
Wow. I forgot how good McGrady was. And Pat Garrity was terrible. Don’t anybody forget it and start thinking he “carried the team on his back.” He was a power forward who averaged 15 points, 2 assists, 6 rebounds, and a combined 1.5 steals and blocks with 42% shooting per 48 minutes in his two big minute seasons. In Duncan’s worst season, he was about 25 pts-15 reb-5 ast-4 steals/blocks with 48% shooting. Power forwards are supposed to score efficiently, rebound well, and block shots. Garrity did NONE of those things even adequately.
“As a basketball fan”, did you notice that the Magic improved 4 wins just by letting Garrity retire and giving Gortat 6 of his minutes?
Jason J
June 8, 2009
If you look at Howard and Shaq’s WS per game, Shaq’s is considerably higher. 0.26 vs 0.19 (that amounts to 6 wins over an 82 game season).
Shaq was more productive per minute and played more minutes per game, and so to call him the better player doesn’t seem to be objectively wrong when using win score as your sole criterium.
However, Shaq’s health was a concern, so Howard has 112 more games played in just one more season, which accounts for his overall better WS total. And a player who can’t get on the court is not producing any wins.
On the other hand if you had to choose one or other to make a playoff run, wouldn’t you choose the more productive per minute / per game player and trust his health to hold up for 20 games (none back to back) or so?
TRad
June 8, 2009
Re:Garrity
You may throw any stats you like, but they won’t show Garrity’s biggest assets. His clutch performance, hot hand, leadership, grittiness, and you shouldn’t underestimate the fact that he was a great momentum changer.
Italian Stallion
June 8, 2009
As a gambler that makes money betting on horses, I agree with the idea that the best team (or horse) doesn’t always win.
In horse racing, the idea is to look at the relative merits of the horses, trainers, conditions of the race etc… an create an odds line that reflects their probability of winning.
Then you compare your odds line to the actual odds and look for overlays that are large enough for you to profit.
Horse racing is really no different than other sports. The best horses win more often, but there is an element of randomness in how a horse is going to perform on any given day and/or how a race is going to develop.
The results of any given race (or sporting event) rarely prove one way or the other who was actually better.
All that said, if you very arrogant and fall back on the “randomness” argument for every single loss or result that was not consistent with your pre game anlaysis, you are surely never going to improve yourself or learn a darn thing.
I’m no basketball handicapper, but going into that series, I thought Clevelend had a reasonably solid advantage over the Magic. The result in and of itself is not enough to convince me I was wrong. It was a darn close series that could have gone either way.
However, before that series I read and heard several people more knowledgeable about basketball suggest that the series was going to be a lot tougher than it looked on paper because of specific matchup issues that ultimately did prove to be a major factor.
When someone shows that kind of foresight about any sport (or race), I tend to take their insights seriously and try to learn and incorporate their thinking into my own analysis as soon as I am convinced of the merits of it.
So while randomness is absolutely factor in sports results, I am not so sure randomness was as much a factor in the outcome as some people might think.
Perhaps the true odds were really 50-50 because of the matchups and not really 67% – 33% (or whatever) based on other stats that were not comprehensive enough to consider the matchups.
John Giagnorio
June 8, 2009
When you’re ready to translate that post into some quantifiable claims/predictions, I’m more than ready to listen.
Hate Flowers
June 8, 2009
This is all fine and well, I can’t argue against random events influencing life. The thing that is peculiar to me is the attitude that even Prof Berri mocked by saying that if he were to win the Stat Geek Smackdown, well, forget what he said about randomness.
So I know he will agree when I say that it’s curious when others play up the “randomness” card only when reality does not match up to their forecasts. When reality does coincide with the predictions, then we tend to hear “well, that confirms the model”, nothing about randomness.
It takes further investigation and analysis to determine whether your forecasts and models are valid. It’s too easy to cite “randomness” when things go wrong, and to ignore it when things go right.
To be clear, I don’t believe Prof Berri is doing this, I’m attacking other’s misuses of the randomness concept.
The concept itself is solid, the application of the concept is widely abused.
Italian Stallion
June 8, 2009
John,
I’m not sure if you were referring to me, but I didn’t claim to be a good basketball handicapper or that matchups were going to be an issue in the Cavs/Magic series.
However, I do feel comfortable in saying that IMO the Cavs did not cope with the Magic’s style of play well and could not defend certain Magic players well.
Could that be random?
Certainly.
But, IMO it is way less likely to be random when some sophisticated basketball analysts identified those specific players and matchups as being key variables that would make the series tougher than it looked on paper beforehand.
Those analysts could have just been lucky, but perhaps not.
When I’m confronted with this kind of thing in horseracing (something I do have a higher degree of expertise in), I always examine the long term evidence for whether this “insight” has any real predictive value in the hands of an expert. I don’t start with the assumption of “randomness” in order to make myself feel better about being wrong.
My guess is that matchups do matter.
So when you look at seasonal point differentials and try to translate that into win probabilities for a specific series, you probably need to move things one way or the other to some degree to account for specifics like mismatches.
John Giagnorio
June 8, 2009
Italian Stallion,
I agree that examining long term data is important. However, you’re picking one data point and accusing everyone of being “arrogant” or “trying to feel better about being wrong.” I would appreciate it if you could provide your input on more than this one data point, along with additional claims or predictions that we can quantify. I’d also be interested in hearing what, in your opinion, makes a basketball analyst sophisticated. In my experience, most of the analysts who harp on things like match ups have little grounding in statistics and tend to be very results (as opposed to the underlying distribution of outcomes) oriented. In other words, the most likely to be fooled into assigning causation for a truly random event.
I am confident in saying that the fact that some individual analysts predicted this does *not* make it less random; you can always find a few dissenting voices if you look hard enough.
Michael
June 8, 2009
I love how on the one hand you have a table with numbers ranking precisely and to decimal points how many wins a player produces, and then on the other hand a long discussion over how randomness plays a major role in the outcome of sporting endeavours!
Italian Stallion
June 8, 2009
John,
You are suggesting I am doing things I’m not doing and then attacking me for it. That’s not appreciated.
I didn’t accuse anyone of being arrogant.
I suggested it would be arrogant to assume it had to be randomness because your own personal mental and statistical model for basketball does not include match ups.
That would be equally applicable to people that are assuming it had to be matchups without having the data to support that view considering it could be randomness.
Each person can decide for himself if he is being arrogant. :-)
Like I said, I am not a basketball handicapper. I am a horse racing handicapper. I don’t have the data or insights required to answer the questions you are asking.
However, as a fan, I have often seen situations where it seemed to me that matchups were being exploited by one team, but another seemingly equal team could not accomplish the same things because of different personnel and a different style of play.
All in all, it would not shock me if you could refine an analysis that was based on point differential to include some extreme matchup situations and improve your ability to handicap the probabilities of a basketball series.
Given my knowledge of horseracing and gambling in general, I would be shocked beyond belief if there aren’t professional sports bettors that haven’t already explored this area and are already doing that kind of thing extremely well.
That’s the way I approach horse racing.
I make observations (or hear observations being made by others) that seem somewhat logical or possible. Then I explore them and decide if they have any predictive value in my hands or the hands of other experts.
If I was trying to become a professional basketball gambler, I would certainly look at the matchup issue based on what I saw and heard prior to the Cavs/Magic series.
I would definitely do it if I was trying to become a boxing handicapper. In that sport it’s clear that if A > B and B > C that A is not always > C. Matchups and styles are definitely important. I don’t even need the data to know that.
John Giagnorio
June 8, 2009
Italian Stallion,
I thought your statement was very clear:
“All that said, if you very arrogant and fall back on the “randomness” argument for every single loss or result that was not consistent with your pre game anlaysis, you are surely never going to improve yourself or learn a darn thing.”
Maybe I did misinterpret what you meant. If so, my mistake.
Michael,
Randomness plays a major role in almost everything :) However, I don’t think Wins Produced is meant to be interpreted as “player x produced exactly 26.5642 wins.” At least that’s not quite how I look at it.
Ray
June 8, 2009
Daniel,
Did you notice how I said I’d take Pat Garrity over Magic Johnson and Larry Bird. Reread that. Your spirited defense of Garrity’s inefficiency wasn’t needed.
John G
June 8, 2009
The arrogance of western science astounds, at times.
John G
June 8, 2009
Not to discount the arrogance of western sportswriters, of course. Cause is not theirs to ascribe. Nor is it mine.
But the traditionalists are not the only ones fooled by randomness around here.
anon
June 8, 2009
taleb is absolutely being invoked inappropriately (and if he cared about things so trifling, I’m sure he’d agree).
There’s a difference between systemic errors caused by people’s attempts to taxonomize everything and attempt to rationalize random events, and saying that cavs/magic series is not random.
A general phenomenon is being invoked here for the narrow purpose of defending a statistical model. And a parade of sophists and faux statisticians are acting like any attempt to draw lessons from the series is foolish and demonstrates the folly of rational man.
It’s sort of absurd and that’s what is so irritating to the people who take issue with invoking randomness.
Instead of invoking randomness, a much more sensible post from Berri would have read: “I was wrong about the series. Although my model has a lot predictive value, it is not perfect. At best, it can predict how many wins a team will have over the course of a statistically significant sample size with a variety of opponents.”
Michael
June 8, 2009
John I know what you mean, I wasn’t having a go or anything I just thought the juxtaposition was pretty funny!
John Giagnorio
June 9, 2009
anon,
for all the pretentious word choices, your response still doesn’t make a lot of sense. what statistical model was professor berri defending? point differential? i’m not sure how berri was “invoking randomness,” it seemed to me like he was saying “point differential tells us the cavs had x probability of winning and the magic had y probability of winning, and the magic won.” what exactly is wrong with this?
John Giagnorio
June 9, 2009
michael,
i didn’t take it that way at all! i like discussing the uses/implications/interpretations of ideas and models so i guess i tried to move things in that direction :)
John Giagnorio
June 9, 2009
“My major hobby is teasing people who take themselves & the quality of their knowledge too seriously & those who don’t have the courage to sometimes say: I don’t know….” (You may not be able to change the world but can at least get some entertainment & make a living out of the epistemic arrogance of the human race).
– Nassim Taleb, from the front page of fooledbyrandomness.com
“Of course, despite such agreement, I am currently in the lead. This must mean I know a little more than everyone else.
Although I like that story, it really is just a story. In other words, my current lead is probably just luck. A key component of this contest is the requirement that we pick the number of games in each series, and although I think the data helps somewhat with that question, I am not sure it helps that much.”
– David Berri, from “Picking the Conference Finals and Playoff Science”
As an aside, Berri’s sentiments in the “controversial” post are essentially the same as the “non controversial” post from right before the conference finals.
anon
June 9, 2009
“what statistical model was professor berri defending?”
you’re joking, right?
randomness has to be relative to something. the series is only random if can’t be meaningfully captured. the post would not have been written if the results were consistent with berri’s prediction. It was only written because orlando beat the winscore favorite.
anon
June 9, 2009
I don’t think I’m displaying epistemic arrogance. I’m not saying matchups matter or the series isn’t random. I think epistemic arrogance is attempting to describe the real world through a statistical model and saying that anything that doesn’t conform to that model is random. That is an applied epistemology!
dustin
June 9, 2009
Anon, what people are saying is the Cavs losing does conform to the model, as they simply had a greater chance of winning (according to point differential), not 100% chance of winning.
John Giagnorio
June 9, 2009
Randomness: The property of all possible outcomes being equally likely; A type of circumstance or event that is described by a probability distribution; A measure of the lack of purpose, logic, or objectivity of an event
en.wiktionary.org/wiki/randomness
Everyone is referring to the “a type of circumstance…” definition as dustin implies.
Michael
June 9, 2009
“the post would not have been written if the results were consistent with berri’s prediction. It was only written because orlando beat the winscore favorite”
Anon, Professor Berri makes these predictions based on efficiency differentials, not winscore. This isn’t that important really is it so stop getting your knickers in a twist. Even if he did write the whole randomness stuff to justify the discrepancy (I really doubt that he did), wouldn’t that in itself validate the central premise which is that human beings will attempt to rationalise, and explain random and unexpected outcomes? So isn’t this really a non-starter!
simon
June 9, 2009
anon, you’re misunderstanding dberri’s methodology on choosing the winners. Team efficiency differential actually is used pretty much by everybody who’s doing statistical analysis, it’s not just a WoW thing. dberri’s model was developed to assess impact of individual players.
I suspect you just took this site’s claim and decided to dislike its principle without actually delving into it. I apologise if you did spend a considerable amount of time studying it, but a notion such as “winscore favorite” is simply wrong in this context because it’s not true.
And I find it intriguing how many people are flocking to comment on the latest entries after the Orlando win. When he wrote an entry with the exactly same idea prior(https://dberri.wordpress.com/2009/05/18/picking-the-conference-finals-and-playoff-science/) it only attracted 9 comments, hmm. So the commentators didn’t have a problem with the concept back then but all of sudden after the Orlando won the series they get a heated dislike to the long-existed idea of randomness in sports?
Kobe Bryant's Robotic Cousin
June 9, 2009
How dare you censor me. I am no longer one of your well wishers (people who would like to throw you down a well).
Kobe Bryant's Robotic Cousin
June 9, 2009
Odd, my last post never went through. That’s irritating because it included several links. Oh well. KBRC out.
Go Magic.
Joe
June 9, 2009
KBRC,
If you include 2-3+ links in a post on most blogs your comment won’t go through. It is to filter out spam.
Ray
June 9, 2009
Can somebody do WP and winscore for Leroy Smith’s 2 on 2 Hall of Fame Challenage? Those players can really slam it down!
Harold Almonte
June 9, 2009
I think it can be right for an external observer or a bettor to say there’s randomness in basketball outcome (game or series), because he doesn’t have control of that, just to make a prediction based on statisticals of preference.
But players do have control of their play, or their opponents otherwise, unless we say shooting, or clutch plays, are a matter of luck because their probabilities are 50% and under. Then what I think it should be said is that basketball teams (good and bads) doesn’t have transitive relation, and even their eff. diff. records are set under asymmetrical conditions or squedules. And here there’s a chink for the matchup approach, that can cancel or turn statistical probabilities.
I think seven games (more games than what they play between themselves in reg. season, and there’s no slumps in basketball, until it is mentally), and having home advantage, is enough to demonstrate you’re the best (under that condition). Upsets mean something, but just not luck.
John Giagnorio
June 10, 2009
It seems like different people using “luck” and “randomness” in different ways is causing confusion. Maybe it would help to use more precise terms like expected value? To use Harold’s shooting example, a player who makes 50% of his 2pt shots has an expected value of somewhere around 1pt per 2pt shot taken, but more often than not the play results in either 0 or 2 points. Extensions of this idea are at the root of this entire debate, since a team can only record 1 or 0 wins in a game. Something is missing when we say the player made or missed the shot and end our analysis there. Same goes for only saying which team won and which team lost. The challenge is how to talk clearly about what we’re missing.
brgulker
June 11, 2009
more from Jim Glass
in response to this statement “The nature of sports is that the ‘better team’ is the team who wins in any given competition (single regular season game or playoff series).”
This is exactly wrong. Ring Lardner knew “the race is not always to the swift … though that’s the way to bet”.
Looks like Jim was responding to me.
I was not arguing that Cleveland was not ‘better’ in the regular season than Orlando; obviously, they were.
I was suggesting that the way we were talking about ‘best’ was a bit confusing to me.
Let me put it this way. Anyone who’s played sports, watched sports, been invested in sports in any way knows that the ‘best’ team doesn’t always win. And how many times have you and I (and everyone here) heard, “Well, we were better than them, and we should have beat them.”
When I read Dr. Berri’s post about randomness, that’s how I would sum it up. It sounded like the excuse of a losing team. Cleveland was ‘better,’ yet they didn’t beat Orlando (regular season or playoffs). Appealing to randomness seems to lend credibility to the excuse of the loser, and I think that’s very a very strange thing for Dr. Berri to say.
A couple more points. What I would have expected here is some type of brief statistical analysis of the Orl/Cle series. Perhaps Dr. Berri could have analyzed the regular season meetings as well as the playoffs to determine why Orlando is the better of the two teams when Orlando and Cleveland play each other (or perhaps better, played each other this season).
Maybe we could have seen per-minute production for each player that demonstrated how amazing Orlando’s offensive efficiency was and how LBJ’s teammates didn’t contribute enough to his stellar performance in order to get the job done.
But instead, we got a lesson in randomness, which to me does two very strange things. First, it seems to lend credibility to excuses, as I said above. But second, it seems to downplay the absolutely incredible performance of Orlando. It was amazing, and appealing to randomness just seems to diminish that to me.
Finally, back to my point about being ‘better.’ Perhaps the sample size is too small to come to any firm conclusions. Orlando and Cleveland played 9 games (10?) this season, and when the two teams played, Orlando won more than it lost. Further, my understanding of the nature of sports is that the team who plays better than the opposition on any given night will be the winner of the competition; consequently, I don’t think it’s unfair to call the winner ‘better,’ especially when that statement is qualified by admitting that Cleveland is ‘better’ with respect to the rest of the NBA.
So, when I say that I think Orlando is ‘better,’ I simply mean that when Orlando and Cleveland have played each other this season, Orlando has played better than Cleveland when the two teams have met and thus have won the majority of the contests.
And I’d be much more interested to dig into the statistics that demonstrate this to be the case in hopes of making some guesses — even if they are just guesses — as to why. To me, teams are not coins, and randomness just doesn’t cut it.
Harold Almonte
June 12, 2009
Would you say is a matter of randomness that Van Gundy had not called the foul and gived up the 3p to tie the game? Yes, he played the randomness. Does that mean Orlando being better than the Lakers is a matter of a lucky decission in the clutch? No, the Lakers have performed better all the series, not only covered the opp. home advantage, they won in the opponent home. They have a better bench (a team is not only starters), and the use of this bench by coaching is also decisive on how team reg. season stats are built, and how expectancy at playoffs can change when you need to give more or less game to your bench.
Jim Glass
June 13, 2009
**in response to this statement “The nature of sports is that the ‘better team’ is the team who wins in any given competition (single regular season game or playoff series).”**
“This is exactly wrong. Ring Lardner knew ‘the race is not always to the swift … though that’s the way to bet’”.
Looks like Jim was responding to me…
Let me put it this way. Anyone who’s played sports, watched sports, been invested in sports in any way knows that the ‘best’ team doesn’t always win. And how many times have you and I (and everyone here) heard, “Well, we were better than them, and we should have beat them.”
When I read Dr. Berri’s post about randomness, that’s how I would sum it up. It sounded like the excuse of a losing team…
Look, it is very simple. Either you believe random events play some role in the outcomes of games — and self-evidently more so in close games — or you believe they play no role. One or the other.
What are random events? A round ball bouncing this way or that. A ref making a call or not. A coach calling a play that will work *if* the other coach calls a given play but not otherwise, and having to make an educated guess about it in a hurry, and being right or wrong. A player’s foot slipping to mess up a play. Multiply by all the times such things happen in a game.
How may points do you think such things can be worth to a team in game? 1, 2, 3, 5? None at all?
If your answer is >0 then you have to admit that random events can turn ball games, and in playoff series between teams that are closely matched, they can turn a series.
Appealing to randomness seems to lend credibility to the excuse of the loser…
Only if you assume random events account for zero points a game, so that losing the close game must show some character fault or other personal failure on the part of the loser, which requires an excuse.
But then you have factual conumdrums to explain. Why do great Championship teams (and last place teams) in all sports converge on .500 records in close games? Does their character suddenly fail them (or rise from the pits)?
Why did Vince Lombardi have a .500 record in one-score games, and Bill Walsh a .430 record? Because they and their multi-time Super Bowl winner teams choked under pressure, didn’t know what to do?
What’s more plausible? That the dominant champion coaches and teams of their generations were chokers in close games, or that luck is worth a few points a game and so determines close games?
second, it seems to downplay the absolutely incredible performance of Orlando. It was amazing, and appealing to randomness just seems to diminish that to me.
Not at all! One of the greatest sports series I ever saw was the Red Sox coming from down 0-3 to beat the Yankees 4-3 with Curt Schilling pitching the key game with blood pouring out his shoe, but not stopping. He was a ‘effing Hero! The whole team was “never say die!” But do you think the Sox had no luck at all coming from down 0-3 to win 4 straight? C’mon. Do you really believe that after going up 3-0 Torre forgot how to manage and Jeter and A-Rod and all the veteran World Series champions on the Yankees suddenly had an attack of character failure?
When two top closely matched teams play each other the result is close games, and if random events affect close games, well, … but that doesn’t mean you don’t get to watch both teams fighting like tigers! The fact that they both fight like tigers is what makes the great game and great series close!
Further, my understanding of the nature of sports is that the team who plays better than the opposition on any given night will be the winner of the competition…
You are making your own definition there. Nobody can argue with you when you just decide that.
My understanding, and that of Bill James and other sports stat analysts who’ve followed him, is that teams that play better (worse) than their opponents win (lose) one-sided games, games not close at the end so chance doesn’t determine them, while teams that play equal to their opponents play close games determined largely by luck. The difference is that this understanding is backed up by the ability to accurately predict the percentage of times underdogs will win. (Do you know how many times the team with the best record in baseball has won the Series in the last 20 years?)
Look, say one team with an 0.08 “better” win record plays another, that’s half a win per seven games. Is your prediction: “The best team will always win 4-0. This time too!”? Probably not.
Probably you’ll predict the best team to win in 6 or 7. But why not 4-0?
Will your rationale be: “The best team will be best in 4 games but probably not be best in 2 or 3 because it will choke or forget how to play well or maybe the other team will rise above its natural ability, but only temporarily…”
Or will your rationale be: “The best team will win, but these teams are very closely matched, the underdog is good too, and I think it probably will get enough breaks to win 2 or 3”.
That’s what I’d say.
But the bottom line is this: Either you believe random events are worth a few points per game, or you think they account for zero points, always. All else follows from that.
To me, teams are not coins, and randomness just doesn’t cut it.
The thing is, when predicting future W-L performance, sports teams behave exactly like weighted coins. Believe it or not. Ask Billy Beane!
Brian Tung
June 13, 2009
I think Berri is regretting his use of the coin analogy. :)
People are suggesting that because teams and games are not coins (which they aren’t, it’s true) that the analogy is not applicable. If anything, it is MORE applicable, because it was never meant to be applied to the teams or the games, but to anyone attempting to prognosticate the teams and games.
Everyone recognizes that the coin situation is ludicrous, because there is obviously no skill involved in predicting coin flips; it’s clearly random. But basketball games are NOT coins, surely. There’s a rhyme and reason to games and there must, the reasoning goes, be some way to use that rhyme and reason to improve predictive ability.
And sure enough, someone will indeed perform better than average, significantly better, in predicting the results of the games. The real question is, why? Is it really that they have a model that is better than anyone else’s? Or is that there are SO MANY people predicting game results that there is bound to be someone who performs so much better? The truth of the matter is that in many cases, you can’t tell for sure. THAT is the point of the coin analogy.
There’s another anecdote, which might be apocryphal, but is on point in any event. There was, it seems, a study that a parapsychologist did on ESP–a large study. Something like 1,000 volunteers were tested on their ability to divine the symbols on the back of ESP cards (clairvoyance, I believe is what this is called). The researcher found one volunteer that did much better than statistically expected. Three sigma above the mean, in fact. Now, if you know anything at all about statistics, you know that three sigma is a lot. So it sure seems as though this volunteer might be a true clairvoyant.
Except…the reason that three sigma is typically considered statistically significant is that it is very unlikely in the context of the null hypothesis (that the deviation is due to random variation). About 1 in 1,000. But there are 1,000 volunteers! You would EXPECT there to be one volunteer three sigma above the mean. (And one three sigma below the mean, for that matter, but we don’t hear about that.) Whether or not you believe in ESP, if you are statistically rigorous, you’d have to conclude that ESP was specifically not demonstrated in this case, despite the presence of what appears in isolation to be a statistical outlier.
The lesson is that events that seem like statistical anomalies in isolation are often perfectly ordinary when considered in the larger context. They are illustrations of why anecdotal evidence blows, and blows hard: Look long and hard enough, and you can always find SOME evidence to support practically ANY cockamamie hypothesis you like.
Harold Almonte
June 13, 2009
I agree with Jim, but not with his use of the concept of “luck”. To choke in the clutch is not bad luck, is to play bad (below expected) under stress situation. I think what some people mean to say is that being 0.08″better” or so, on stats built under regular season squedule, doesn’t mean a tangible advantage, and one would need to re-analyze again.
There’re lots of randomness situations in sports, due to decission makings, but to loose in seven games, starting with the home advantage? a little more tangible advantage?. I think that taking advantage and being “matador” is quality of being best, the same that struggling, being trailed, and to come from behind and win. The best team is wich orders or puts randomness on its favor, not in a casual event, but a series of them.
Harold Almonte
June 13, 2009
Note: based on stats only, I would have predicted Clev. in seven.
brgulker
June 15, 2009
Look, it is very simple. Either you believe random events play some role in the outcomes of games — and self-evidently more so in close games — or you believe they play no role. One or the other.
Of course randomness plays some role. I’m not disputing that. What I’m disputing is the notion that Orlando didn’t play better than Cleveland.
I read Dr. Berri’s original post to mean that we had been ‘fooled’ by randomness — we meaning fans, of course, but the management of Cleveland in particular; the post was predicated on an article about trade rumors in Cleveland. I don’t think we were fooled at all.
In a few words, this is my entire point:
Orlando played better over the course of the series, and they won the series as a result. Orlando shot the ball as well as they could possibly shoot, and Cleveland’s heralded defense couldn’t handle both Orlando’s outside shooting and Howard’s inside presence.
Yes, both teams benefited last-second heroics (LBJ and Lewis) or “luck” or “randomness.” But that’s my point — BOTH TEAMS benefited. It wasn’t just Orlando, and that “luck” doesn’t detract from the FACT that Orlando played better than Cleveland over the course of the series.
Further, my understanding of the nature of sports is that the team who plays better than the opposition on any given night will be the winner of the competition…
You are making your own definition there. Nobody can argue with you when you just decide that.
My understanding, and that of Bill James and other sports stat analysts who’ve followed him, is that teams that play better (worse) than their opponents win (lose) one-sided games, games not close at the end so chance doesn’t determine them, while teams that play equal to their opponents play close games determined largely by luck. The difference is that this understanding is backed up by the ability to accurately predict the percentage of times underdogs will win.
Why is it that I’m creating my own definition, but you (or other statisticians) are not? And frankly, my point was to question the definition of “better” that we were using.
With respect to the rest of the NBA over the course of the 82-game season, Cleveland was “better.” They had a better record and better efficiency differentials. We agree on that. And I expected Cleveland to win in six, FWIW.
By that definition, Cleveland was “better,” and it’s reasonable to predict that they would win.
However, when we look at when the teams played each other this season, Orlando beat Cleveland much more often than they lost to them.
Or in other words, when Orlando and Cleveland have played each other this season, Orlando has played better than Cleveland more times than the converse and as a result has beaten Cleveland more than they have lost to Cleveland.
And to me, (and it seems like to you as well), that’s the excitement of sports — that on any given night (day), the ‘underdog’ can win.
The irritation for me is that Dr. Berri’s article implies that Orlando’s win can be attributed to randomness. Or, Orlando just got lucky. But as others noted, the article completely ignores how “lucky” Cleveland also was. If not for a miracle shot from LBJ, Orlando likely wins the series 4-0 or 4-1. Or in others words, if not for LBJ getting extremely “lucky” the series would have been even more lopsided than it already was.
And that’s the basis for my objection, really. Dr. Berri’s article, especially the title, implies that Orlando just got lucky. It even cites how Orlando got lucky. But, it ignores the FACT that Cleveland also had their fair share of luck.
Was randomness involved in some way? Sure it was. But it was involved for both teams, and doesn’t entirely explain how Orlando beat the ‘better’ team so frequently in both the regular season and playoffs.
In my mind, the fact remains that Orlando outplayed Cleveland for the majority of the games in which the two teams competed against each other this season. And to me, a much more interesting conversation would have been an analysis of each player who played in these games.
Such an analysis would demonstrate, one way or the other, how much luck was actually involved — since that’s what the book and this website are about. Orlando played better basketball (was more productive), and they won the series. The statistics would demonstrate that. And since WoW measures players (and consequently teams) in terms of their productivity, doesn’t it make a heck of a lot more sense to explain the series that way instead of by a random appeal to randomness?