Review of "Basketball on Paper: Rules and Tools for Performance Analysis" By Dean Oliver
Review by Tom Lyons
"Basketball on Paper" is a recently published book (Brassy's, Inc. 2004) that attempts to use statistical analysis to distinguish what is truly important to winning basketball games. The book also provides ways to analyze individual player performance. The jacket describes the author, Dean Oliver, as "a former basketball player, assistant coach, and collegiate scout" who founded the online "Journal of Basketball Studies" (www.rawbw.com/~deano/) and who has contributed to Basketball Digest and the Association for Basketball Research. Oliver attended college, played basketball and coached at Cal Tech. He has a Ph.D in engineering from North Carolina. Oliver worked for Bill Bertka's scouting service for four years. He has paid the bills as a computer consultant and risk analyst. Oliver is now with the Seattle Supersonics putting his basketball theories into practice. (As of the 2005 All-Star break, the Sonics were 35-15 and 10 games up in the new Northwest Division. Last year, they finished 40-42, 19 games back in the Pacific Division). Referring to Michael Lewis' book about the Oakland Athletics' use of statistical analysis to run their team, Oliver has called his book the "Moneyball" of basketball.
It is to some extent a compendium of material previously published on his website or elsewhere. One article on the website seems to be a revision of one chapter of the book. It is well written and appears to be a very thorough attempt to use a lot of data, mostly from the NBA, and, in some cases, very complicated formulas* to arrive at its conclusions. In that sense, the full-blown analytical methods themselves are not very useful for coaches unless you have a ton a data about your team and individual players as well as the ability to draft a computer program that uses that data. The book provides a detailed explanation of how to record data about a team manually. Oliver says on his website that he has developed an analytical program named "Roboscout" that he uses with the Sonics but it is generally not for sale.
Oliver believes his conclusions apply equally well to women's basketball, but he does not have the wealth of data there that he has for the NBA. Oliver has some WNBA data and states he is involved in a long-term project analyzing WNBA data but he does not have many conclusions from that analysis yet. Nonetheless, nothing in the data he has indicates his conclusions do not apply to women's basketball. His conclusions are useful and enlightening, at least to me. (Even if you already know them, you can now say science confirms them). Here are Oliver's main points.
Based on statistical analyses, the four most important keys for team success in basketball and their relative weights, in parentheses, are:
- Shoot a high field goal percentage (10).
- Do not commit turnovers (5-6).
- Get offensive rebounds (4-5).
- Get to the foul line frequently (2-3).
Teams that consistently win basketball games do at least three of these things well. If a team doesn't shoot well, it better do the other three things very well. I should note that Oliver says these factors should be considered on the basis of the number of a team's possessions compared to its opponent, not in absolute terms. In other words, he looks at efficiency, for example, the number of made shots per possession, or the number of turnovers per possession, not total points or turnovers, which can vary greatly depending on the pace of the game.
Obviously, these keys could also be stated from the defensive perspective, i.e., prevent easy baskets, cause turnovers, do not give up offensive rebounds, do not foul, etc. However, I have put them in the offensive perspective because of another of Oliver's conclusions: In the NBA, at least since the 1970's, offense wins playoffs and championships more often than defense . The difference in success between strong offensive teams and strong defensive teams is not large, but the data contradicts the old adage that defense wins championships. I suspect this may be even more accurate at lower levels of basketball where offensive skills are more rare than defensive skills. [On the other hand, in a one-game playoff, it is possible for a team or a critical player to have a bad shooting night, whereas defense may not be as variable].
But, if defense does not necessarily win championships, where did the idea originate? With Bill Russell's Celtics who were the best defensive team ever and won 11 of 13 championships during his tenure. They had Hall-of-Famers coming off the bench. Of course, basing the idea on those Celtics is like looking at Michael Jordan's Bulls who won 6 titles in 8 years and saying shooting guards win championships. The fabulously exceptional team or player is still the exception.
Oliver offers some comments on why he thinks these factors appear so important. The benefit of a high shooting percentage is obvious, especially since most missed shots result in defensive rebounds (and potential fast breaks). Oliver cites one study that found that the NBA team with the higher field goal percentage won 79 percent of the games.
Turnovers are important because a team cannot even shoot, never mind score, if it turns over the ball during a possession. The NBA team with fewer turnovers wins about 58 percent of the time. If field goal percentages are about equal, the team with fewer turnovers wins 69 percent of the time. In an email to me, Oliver commented that turnovers are even more important below the NBA level.
Offensive rebounding can make up for a poor shooting percentage, particularly as shots off of offensive rebounds generally are more likely to result in scores than initial field goal attempts. Oliver refers to a study of WNBA offensive rebounds done by an official with the Charlotte Sting. The study found that offensive rebounders increase their field goal percentage from 41 to 48 percent and their points per play from about .80 to .94. That improved the team's points per play from .80 to .90, which Oliver says is a huge difference. In the NBA, if shooting percentages are about equal, the team with more offensive rebounds wins 63 percent of the games. (This also emphasizes how important it is for the defense to box out).
Oliver discounts to some extent the value of defensive rebounding. He notes that 70 percent of missed shots result in defensive rebounds so they do not take much special effort or ability. Also, defensive rebounds are to a large extent the result of good defense making the offense shoot a low field goal percentage. In that sense, factoring in defensive rebounds is like double-counting the same affect.
Interestingly, it is more important for teams to get to the foul line frequently than it is for them to hit a high percentage of their foul shots. Perhaps, this is because lots of foul shots means the other team's starters are in foul trouble. Again, if shooting percentages are equal, the NBA team that commits fewer fouls wins 67 percent of the games.
Oliver also has some other observations based on his NBA data. Good offensive players are more important than good offensive systems. Oliver offers numerous examples of good scorers who succeeded regardless of the system in which they played and "good" offensive systems that floundered without a good scorer(s). Think of the Bulls' triangle offense with and without Michael Jordan. Think of Shaquille O'Neal in Orlando and Los Angeles. On the other hand, Oliver does note that a few NBA coaches have had offensive success with more than one team: George Karl, Phil Jackson and Don Nelson. This may be because they are good at recognizing and using scorers well, not necessarily because of any particular system. In other words, adapt the system to the players rather than vice-versa. However, I should note that Oliver believes offensive systems are more important at the college ranks and below but he cannot prove it.
Offensive efficiency is more important than pushing the pace of the game. Oliver quotes Dean Smith who said: "The biggest reason I'm against simply running the ball down and shooting the first shot available is that the defense doesn't have time to foul you." And, as noted, getting fouled and shooting lots of foul shots is a key to success. Slowing the game (reducing the total number of possessions) down does not seem to increase defensive efficiency (the percentage of stops), though, as will be described below, it does make the outcome of individual games more variable.
Good offensive teams have both inside and outside scoring, however, no single position is key to offensive success. A team does not have to have a great point guard or a great center to score efficiently. However, it does have to pass well. Most offensively successful teams have a high percentage of assists on their scores. Assists lead to higher quality shots and a higher shooting percentage. The NBA team with the higher number of assists wins about 72 percent of games.
Height is more important to offensive success than defensive success, though it is not critical. Statistically, the top offensive teams in the NBA have been somewhat taller than average, whereas the top defensive teams have been shorter than average. [Of course, the reason could be that shorter teams had to play tougher defense because they had a hard time scoring]. While there is a correlation between having a top shot blocker and having a very efficient defense, there is no correlation between a high team total of blocked shots and defensive efficiency. Oliver surmises that height may be more important at the high school and college level where there are greater variations in size among the players.
Contrary to what some coaches and commentators think, it is not easier for offenses to rebound three point shots. Oliver looked at numerous NBA games between 1999 and 2002 that included several thousand missed shots. The offense rebounded 33 percent of two point misses and 31 percent of three point misses. Generally speaking, the rebounds on three point shots do not go farther from the basket than two point shots. Also, many two point field goals are shot close to the basket and the shooter often has the best chance to get the rebound.
There is no such thing as a "hot hand." Oliver cites two studies of NBA players and one of college players that showed no indication that any particular players were more likely to make a shot if they had made the previous shot. The NBA studies were of the Philadelphia 76ers' field goal attempts during the 1981 season and the Celtics free throw attempts during the 1981 and 1982 seasons. If anything, the players were slightly more likely to make a shot if they had missed the previous one. Of the twenty-six college players only one was significantly streaky.
Underdogs should adopt a "risky" strategy even if no particular strategy is clearly better. A risky strategy is one that changes the dynamics of how the game would normally go. Examples of risky strategies are pressing, attempting many three point shots, slowing down the pace of the game, playing a zone, fronting the post, sending guards to the boards to rebound, releasing the guards on defense for a potential fast break, and playing oversized or undersized lineups. If one such strategy is clearly better, such as pressing a tall team with poor ball-handlers, then the underdog should use that strategy. However, even when no particular risky strategy is better, the underdog should still use one.
Oliver notes that the slow down strategy works by reducing the number of possessions each team has. This reduces the better team's advantages and makes the outcome more variable. Similarly, the other risky strategies make results more variable in different ways. For that reason, the otherwise superior team should avoid risky strategies. Oliver describes how Rick Pitino's talented Kentucky teams may have lost some games by pressing.
With respect to evaluating individual player's performances, Oliver says the goal is obviously to determine how well a player helps a team do the things that win basketball games. From an offensive perspective, the player should:
- Shoot a high field goal percentage or help teammates shoot a high percentage through assists and playmaking.
- Avoid turnovers.
- Get offensive rebounds.
- Draw fouls.
At the defensive end, the player should:
- Make offensive players take low percentage shots.
- Cause turnovers.
- Box out and prevent offensive rebounds.
- Avoid fouls.
Oliver discusses several different ways of evaluating individual offensive and defensive performance including his own methodology. Suffice to say, most of them, including his own, require much detailed data about what a player does in the game and a good computer program. However, if a coach does not have those resources, she can still pay close attention to these four factors.
Even using Oliver's methodology, individual defensive assessments in particular require a lot of subjective judgment. Which defender gets credit for causing the offense to take a bad shot, especially if the defense is in a zone? Which defender caused a turnover, the one harassing the passer or the one who stole the pass? Who failed to box out the offensive rebounder? Whose fault is it if a help defender picks up a foul? Should the credit or blame be split between two defenders? And, to what extent does the coach want to delegate that judgment to someone else?
Team chemistry is very important though it is more difficult to quantify. Chemistry has two parts: having a team that has well-balanced skills and having players who are willing to contribute to the team's good balance. If you already have several good shooters, you may need more ball-handling and defensive skills, otherwise, you will have too many turnovers and give up too many points. On the other hand, if your team dribbles and defends well, but cannot put the ball in the basket, you need shooters. Oliver discusses some statistical analyses to determine your team's needs.
The second part of chemistry, players with good attitude, is very important but very difficult to quantify. Oliver says that an important issue is giving players credit for the contributions they do make. He describes psychological studies about people's concepts of "fairness" that are not purely rational. He notes that very few players are willing to play on a team where one player takes all the shots and gets all the credit no matter how much better he or she is than the others. Good coaches have to find ways to balance players' abilities and give credit to maximize individual efforts and team success.
While the second part of Oliver's book may not provide much specific help to coaches who lack a substantial support staff, it does provide a lot of fodder for basketball buffs to debate. He has a huge amount of statistical analyses of individual players. The first part of the book has a large amount of data comparing NBA teams, past and present. For example, his analysis indicates that the most dominant professional basketball player in the last 25 years is not Michael Jordan, it is Cynthia Cooper. (Who volunteers to tell Mike?). Arguably, the single best offensive season for a male professional player in that period belongs to…..Reggie Miller in 1991. As great as he was, a good argument can be made that Jordan did not shoot the ball enough. Ditto for Cooper. Allen Iverson should shoot less. Who was a better NBA player, Bird or Magic? Well, statistically, it's a dead heat, for second place, behind Jordan. The top offensive NBA teams, compared to their competition, have been the Denver Nuggets of 1982, the Chicago Bulls of 1997 and the Dallas Mavericks of 2002. The top defensive teams were the Knicks of 1993, the Bullets of 1975 and the Jazz of 1989. Cooper made the Houston Comets simply unbeatable at either end of the court.
Oliver's book and website are very useful for coaches at all levels. Coaches who have the resources to gather substantial data on their teams and opponents and analyze it will have very helpful information. They can use that to evaluate their players, pick optimal offenses and defenses for their teams and decide good match-ups or combinations of players on the courts. Even coaches who do not have those resources can use this information as guidelines for making more intuitive decisions.
©2005 Thomas W. Lyons, no claim as to Dean Oliver's original material.
*For example, Oliver's formula for measuring the points produced by an individual player: Points Produced=(FG Part + AST Part + FT Part) x (1- (TMOR÷TMScPoss) x TMORweight x TMPlay%) + OR part, where FG Part is 2x(FGM+ ½ x FG3M) x (1- ½ x (PTS-FTM÷2xFGA x q ast ); AST Part is 2 x [TMFGM-FGM = ½ (TMFG3M-FG3m)] ÷ (TMFGM-FGM) x ½ x [(TMPTS-TMFTM)-(PTS-FTM)] ÷ 2 x (TMFGA-FGA) x AST; FT Part is [1-(1-FT%) 2 ] x 0.4 FTA; and, OR Part is OR x TMOR weight x TMPlay% x TMPTS ÷ TMFGM + [1-(1-TMFT%) 2 ] x 0.4 x TMFTA. [As a former history major, some parts of the book caused me severe MEGO where MEGO is "My Eyes Glaze Over."]