The rise to prominence of columnists such as Rob Neyer in baseball has created an atmosphere in which many clubs are hiring according to objective, statistical analysis. Football, although not to the same extent, is following the same pattern. Why isn't basketball? It seems that in basketball, too many GM's still operate on the glitz principle, and ignore the possible benefits of deep statistical analysis. I know that stats can be misleading in basketball more than in baseball, and even football, due to the fluid nature of the game, and individual match-ups.... yet, I'm sure that if you layered stats deeply enough, you could come out with several formulas which would be effective. If these exist, you rarely hear about them. Heypartner... any chance you know much about this? Dreamshake, popeye?
There are several statistical methods of analyzing basketball, but to put it bluntly, stats to really analyze things such as the correlation between player interactions don't exist to my knowledge. Yes, we have things such as assists and 2nd chance points, but I haven't been able to find individual player data for stats such as this on a quarter-by-quarter basis or in correlation with other players concurrently playing together. I tried compiling a database of statistics during this season to track stats such as "When Francis and Mobley play more than 30 minutes in a game, what is the Rockets' record?" These type stats are easy enough to do with something simple like an Access database or some web-based database (Clutch may be using something similar now). Unfortunately, there is no sport that I can think of that statistically captivates quite like baseball (and I hate baseball ). If you really want to consider hypothetical player valuations or team performance based upon statistical analysis, check out The Journal of Basketball Studies (JoBS) and its creator, Dean Oliver. http://www.tsoft.net/~deano/frontpg.html Click on the "Methods" link at the bottom of that URL. Interesting reading and fairly easy if you're interested in basic numerology. Two interesting formulae he uses that some of you may be interested in would be the "Approximate Value Method" of evaluating a players' season and the "Trade Value" method to determine a player's trade value. Of course just about all statistics basically forgoes subjective analysis in favor of empirical/concrete interpretation of events, so take any of these formulae as being one person's effort at "modeling life". In other words, don't expect it to be perfect 'cause it ain't. ------------------ trade them all and fire the coach.
Yes, but it would likely still be more effective than a strategy based on subjective analysis. GM's that recognize this in baseball are generally managing successful teams. GM's that ignore this, are becoming less successful, in general. I think it's sort of bizarre that this type of analysis DOESN'T exist much in basketball, when I think it would almost certainly lead to at least some improvement in free agent acquisitions, such as an OBJECTIVE "cost-benefit analysis." As you said, it's not THAT difficult to do w/ a relatively sophisticated search engine. I doubt it would cost a team more than $100,000 year, including a good statician... which would more than be paid for by corresponding increases in salary efficiency. BTW, thanks for the link... it's very interesting. [This message has been edited by haven (edited February 16, 2000).] [This message has been edited by haven (edited February 16, 2000).]
Haven, I'm sure basketball teams employ statisticians that delve into these numbers. They probably have access to much more numbers than any espn, sportsline, or nba.com site can offer. As for my comment about it being relatively easy to do, what I meant was, stats based upon what the public can easily obtain are easy to do. For example, if you give me the number of minutes Francis and Mobley played in a game and the number of points they scored, I can generate something simple such as "The Rockets have won X games when Francis and Mobley have scored more than Y points combined and played for Z minutes in a game". What I can't tell you from that is a stat such as "What percentage of points in a game have the Rockets scored when Mobley and Francis were in the backcourt together?" I have no idea if anyone outside of a team's statistician keeps track of something like that. I think this is what you're trying to say as well. That's a good stat, but how does one obtain it? I've scoured the 'Net for player stats based on quarters, but nothing... Stats, Inc doesn't have anything for basketball like they do for baseball in terms of stats. ------------------ trade them all and fire the coach. [This message has been edited by Dr of Dunk (edited February 16, 2000).]
Baseball's one on one nature is much more conducive to statistical analysis. In baseball, one can sufficently isolate a players preformance such that you can compare it to other players statistics and recieve valid information. Because every player faces more or less the same pool of pitchers, the same number of day/night games, that it is easy over 400 or so at bats to isolate the performance of the individual batter. You'll also notice that people place far less stock in the statistics of people with 100 or 150 at bats. This is because these people don't have a representitive sample of at bat conditions from which to isolate the preformance of the batter. For instance, if a backup infielder only plays against the 5th starters for the other team he might have an exceptionaly high average, which would be discounted because of the ease of the pitchers whom he faced. In basketball, a players statisticaly measurable performance is much more heavily influenced by a myriad of unrelated factors. For instance, if you have Jason Kidd as your PG, your FG% is likely to be positively influenced as a result of all of the easy layups that you get. Also, figure in how well a player fits a teams offense, half court team vs. running, etc. In short order, with all of the uncontrolled and unquantifiable variables, your data becomes unreliable, which is dangerous because alot of people will implicitly believe statistics, even misleading ones. In other words, garbage in, garbage out.
One of the major differences are that in Basketball hot-dogging (glitz) sells tickets, sells merchandise and gets national TV attention. There is a correlation between the top level players in both leagues selling tickets. A perfect example is the Sac. Kings. These guys were one of the worst NBA franchises a few years ago. Then they added White Chocolate, & Webber (who work so well on the offensive end of the court). If you look at the Kings on paper they may seem to be so succesful (I think they are the 4-5 seed in West) and they are in the top 5 in the league now in merchandise sold. However their problem appears to be they still haven't learned how to play defense and they are terribly inconsistent. However I don't think that really matters to the Kings now, because they are selling tickets, selling merchandise, will surely make the playoffs (ad I think will make it farther than the 1st round), have star power in Chocolate & Webber and have generally turned a losing team into a well known NBA franchise. Basically fans of the NBA like this show-boating. I don't think you can say the same of most MLB fans. Sure fans like homeruns, but I don't think a MLB team could get alot of positive attention if all the infileders tried to barehand every ball, or if all the outfielders dove at every fly ball. To summarize everything glitz sells the NBA, but not MLB.
It is an interesting idea to use stats in basketball to evaluate players, but I don't think it would produce better results than the current method. As far as evaluating talent goes, I'd put my stock in an intelligent man, heavily experienced and totally immersed in the basketball industry, with an eye for talent. I think any mathematical model would have a hard time competing with the human brain, properly trained.
I concur with Ottomaton's comments. In baseball, you really have 2 main variables- a pitcher and batter. There are a few other variables involved--the defensive players out there, wind, sun/lighting, ballpark features, etc.)--but they are much more minor. Also, over time most all batters face the same qualities of pitchers with comparable frequency (unless you bat FOR the Braves). Thus with a lot of observations (at bats), stats between players become pretty comparable, though sometimes you should also consider things like pitching or hitting in 1/2 your games at Coors Field. Basketball however is extremely context driven. There are lots of variables--other players on the court both playing with you and against, various game situations--involved in success. Hockey uses a goals scored/goals against figure, which I think is a decent guide to have about a players performance. The nice thing about this figure/number is it comes closer to judging what really matters-team performance, over individual performance in a team sport. I don't know if anyone uses it but a points team scored/ points team given up while the player was in the game figure. Such a figure could be a useful statistic, as would be a similar one focusing on +/-'s based on individual match-ups (how you did relative to your guy). You still would also have to consider some other reference points--like the overall teams performance, does the player always take the easiest/toughest defensive assignment, play in meaningful minutes or mop-up, etc., but it would be a nice compliment to existent individually-based statistics in basketball. As far as statistics being replaced by solely non-statistical judgements, that wouldn't make much sense. Just as you want your doctor checking your cholesterol figures, blood pressure, etc. in developing a plan for your care, statistics are an instrumental tools most evaluations of persons or organizations (if you have a stock advisor don’t you hope he/she looks at company profits and other figures?). Statistics just need to be looked at in context--often many statistics should be subjectively consider at the same time, to me most informative about a basketball player. [This message has been edited by sir scarvajal (edited February 17, 2000).]
