I spend part of the long holiday weekend reading Dean Oliver’s Basketball on Paper (aff link). This book was published in 2004, and I had seen a few references to it around the internet, but just now got around to reading it.
Oliver is a former basketball player, coach, and scout. He is now a NBA statistical consultant who has done work with the Supersonics and Nuggets. The book is ambitious in scope, covering a range of statistical topics related to basketball–from charting individual possessions to Pythagorean winning percentages to individual player ratings to whether “defense wins championships.”
I’m going to shirk a bit here and review the book bullet-point-style. Here are the pros:
- The book is engaging and quite readable, if a little “bloggy” in places.
- It provides a nice survey of various statistical aspects of basketball, serving as a sort of primer of available research on the topic.
- Oliver is very up-front about the limitations of statistical analysis–recognizing the need for a balance between scouting and statistics.
Here are the cons, which are fairly nitpicky:
- The math is a little heavy in places (even for a stathead such as myself)–although most of the heavy lifting is relegated to the appendices. It does take some work to figure out exactly how Oliver’s individual player ratings work. What he’s done is basically construct the best metrics of individual player performance that can be accomplished based on the statistics now widely available.
- The focus is on the NBA (and WNBA). This make sense from a statistical standpoint since the NBA playing field is more even, but make the analysis slightly less relevant to us college basketball junkies.
- The organization of the book is a tad scattershot. For example, discussion of the “four factors’ method of analyzing basketball statistics is spread out across the 350+ pages of the book.
- The book may have tackled a few too many topics. The chapter on parity on the NBA, for instance, wasn’t very well developed and probably could have been saved for further study.
These minor critiques aside, the book was definitely enlightening and I’d recommend it to anyone who considers themselves a serious student of the game. At minimum, it provides a solid framework for thinking about what exactly basketball statistics measure–and what they don’t. The concluding chapter of the book provides a nice summary the various findings of the book. Oliver’s first summary statement is:
In trying to understand basketball, get to know the team first and the player second.
This fits with my own experience: Over the last basketball season, I’ve delved into basketball statistics much more heavily in the past due to the creation of this blog. I’ve slowly realized that I enjoy analyzing team basketball statistics more than individual basketball statistics. And that makes sense, since basketball is ultimately a team sport.
Thinking about three major team sports in the U.S.:
- Baseball is fundamentally a sport of individual performance. Team performance can basically be constructed from individual hitting and pitching stats (setting fielding aside).
- Football is fundamentally a sport of coordinated team effort. Players have specialized roles and it’s very difficult to interpret individual player stats outside the context of their particular team.
- Basketball strikes a balance: Players all have to do the same things on a court (in different proportions) and their stats have some meaning on their own. But the team aspect of the sport is equally, if not more, important as the different proportions of things players do result in the development of unique roles.
The evolution of statistical metrics in the three sports follows suit. Because they measure individual performance so well, baseball stats have become increasingly sophisticated over the last several decades, to the point that learning about all the stats that are out there is the rough equivalent of earning a graduate degree. Baseball statistics were my first love, but at times the statistical rigor that can be applied to the game can become excessive from the standpoint of simply enjoying the game.
Football stats are just now becoming more sophisticated, but are necessarily extremely complex, making them hard to approach for amateurs. Plus the sample sizes are so small–particularly at the college level where a good team may only play 7-8 games against comparable opposition.
Basketball stats, meanwhile, are at the same time (1) simple enough for us commoners to understand and analyze and (2) in a relative stage of infancy in terms of being utilized to explain the game. That was part of the appeal of starting this blog. I don’t expect to make major contributions to the field of basketball statistics, but I do like knowing what I have to say about MSU basketball is based on the most up-to-date statistical tools available (to the public, at least).
So here’s to tempo-free basketball statistics–providing us with enough data to intelligently discuss the game while still yielding to the ineffable nature of teamwork.
Thus concludes the first book review (and related thoughts) posted on the Spartans Weblog. Next time I do one of these, I’ll try to make it’s more timely–posted within, say, the same presidential term as the release date of the book.