Implementing a Moneyball approach in complex team sports

B. Gerrard, "Is the Moneyball Approach Transferable to Complex Invasion Team Sports?", International Journal of Sport Finance, 2: 214-230, 2007. [Citation]

This paper analyzes reasons for the success of Major League Baseball's Oakland Athletics as described in the best-seller Moneyball using a benchmarking technique and investigates the development of a knowledge-based strategy in conjunction with systematic analysis of player performance data.  The study goes on to ask whether such a strategy could be effective in invasion team sports like soccer which have low degrees of separability between team and player performance.  A hierarchical structural model is developed to analyze player and team performance in the English Premier League during the early 2000s.


The issue of whether the principles illustrated in Michael Lewis' best-seller Moneyball: The Art of Winning an Unfair Game can be extended from baseball to other sports in an intriguing one.  This question goes all the way back to the start of this website and the founding of my company. With the nationwide release of the motion picture Moneyball this weekend and the appearance of Billy Beane at the upcoming Leaders in Performance conference in London, it's time to reexamine those ideas.

The publication that I will discuss is one that I mentioned in the early days of Soccermetrics, but I was not able to find the full text of the publication until recently.  (You can find the raw text on, but I like to view the original article in case there are any figures that are more illustrative.)  The author, Bill Gerrard, is Professor of Sport Management and Finance at Leeds University Business School (the title Professor is a much stronger one in the UK than in the USA and indicates a department head) and has done work in sports finance over 25 years, including assessment of football club valuations. In the years since he wrote this article, Gerrard has collaborated with Oakland A's general manager Billy Beane on developing extensions of his approach to soccer.  This 2008 article in the UK Guardian is a good read of Beane's intentions, but I don't know if the Beane-Gerrard collaboration was tied specifically to Tottenham Hotspur. (You will see the name Damien Comolli appear, however.)

In the journal article, Gerrard attempts to do three things:

  • Quantify the "Moneyball effect" of the Oakland A's
  • Describe the "Moneyball effect"
  • Search for an extension of the "Moneyball effect" in complex invasion team sports

Two definitions are in order; first, the "Moneyball effect", and second, the idea of complex invasion team sports.

The "Moneyball effect" is best described as a knowledge-based strategy in which player performance data are used systematically to inform decisions on player recruitment, player valuation, and game tactics.  Gerrard calls this a "David" strategy that small-budget teams are more likely to implement, but there is no reason why large-budget teams could not implement such an approach — and some of them have, in fact. Alternative strategies center around the acquisition of resources by either developing unique strategic resources that are difficult to replicate by competing firms (teams) or by purchasing as much of the high-quality resources as possible. Such strategies often result in varying levels of achievement depending on the quality of the management team, as described in the figure below.


Gerrard illustrates the effect of the knowledge-based strategy for the A's by developing models of team performance as a function of payroll costs over a period of nine seasons starting in 1998 (the year Beane became A's GM).  In constant 1998 dollars, the A's marginal performance as a function of payroll, quantified as marginal cost per win, was significantly more efficient than the rest of the league, and about 50% more efficient than the MLB average.  Gerrard developed two regression models that estimated the expected number of wins based on payroll in order to assess the contributions of team experience and team innovation and obtained more dramatic results.  One, the A's won an average of 10 more games than expected each season over the nine-year period, which can make the difference between a marginally winning season and a playoff contender.  Second, the bulk of their wins were attributed to innovation effects, which indicates that they were continuing to innovate and refine their selection and tactical strategies. Given the level of turnover in their organization due to the pressures of being a small-budget team, they pretty much had to!

There are some issues with the results.  The first one is that most of the World Series winners during that period had marginal payroll costs per win that were in line with the league average (Florida, St. Louis, LA/Anaheim, and the Chicago White Sox were the only teams that were below the average). Even the Arizona Diamondbacks, who won the World Series in 2001, had marginal payroll costs above the league average.  Phoenix is definitely a major city in the USA, but the D-Backs aren't a big-money team in the sense that the New York, Boston, and LA (Dodgers) teams are.  The second is that innovative teams, more often than not, advanced to the playoffs and appeared in World Series (the lone team out appeared to be Atlanta).  Big-spending teams won World Series in that period, but there are a number of championship teams that were able to innovate and overachieve.  They didn't do so to the extent that the A's did, but the randomizing effect of the playoffs made those differences negligible.  I can understand better why Billy Beane hated the playoffs as a measure of the best team in the league, and his attraction to a sport competition whose champion is decided solely on a double round-robin.

Gerrard goes on to explain the reason for the A's overperformance during that period, which was that they discovered a discrepancy between the market valuation of hitters and their respective field valuation.  The specifics aren't really that important for a soccer blog (but if you're curious, it had to do with on-base percentage), but what mattered was that once they identified this market inefficiency they exploited it through their squad selection and their on-field tactics.  That this knowledge was so valuable was illustrated by the fact that in the year after Moneyball was published, on-base percentage became a valued commodity in hitters as reflected in their salaries.  The principle of market efficiency works as expected.  Gerrard commented in wonderment that the A's organization allowed a writer such access to their decision-making progress that a competition-sensitive information would be revealed.  The hope is that the A's had some additional knowledge about the baseball market in reserve, but the moral is that organizations with knowledge-based advantages should keep that information secret for as long as possible.

The remainder of the paper attempts to transfer the knowledge from baseball to other sports, and this section presents the opportunity to explain the second major concept from this paper.  Complex invasion team sports are those sports in which a team of players attempts to gain possession of an object (a ball, a puck, or something else) in order to move it across a field toward an opposing team's goal, while at the same time preventing the opposing team from doing the same thing.  The following list are examples of such sports:

  • All football codes (soccer, rugby, American, Canadian, Gaelic, Australian)
  • Basketball
  • Hockey (ice or field)
  • Polo (field or water)
  • Team handball
  • Netball

These sports are characterized by cooperative action between players, a more complex set of individual actions, an interdependence between offensive and defensive actions, and varying levels of continuous or segmented play.  In the case of the football codes, one can think of a continuum of permitted actions and levels of continuous or segmented play.  These sports contain highly nonlinear, coupled, and stochastic actions that are difficult to model, which makes assessment of individual and team performance extremely challenging.  In soccer, the largely continuous nature of the game and the hybrid roles required of players make the sport one of the most challenging to model statistically.  The lack of (until recently) a data culture within the sport makes it more difficult.

Gerrard ran into the same problems that soccer analysts experience with the lack of availability of data, especially payroll data which European clubs keep secret.  So he is not able to develop the kind of benchmark analysis for the English Premier League that he did for Major League Baseball.  In fact, the source of his data came from the Football Yearbook which Opta publishes for the Premier League!  Gerrard develops a series of hierarchical models that cascade from "general play" (which he defines as passing statistics, crosses, dribbles, tackles/interceptions won, blocks, and clearances) to shot offense/defense and scoring offense/defense.  I have my own objections to using those data per se, but one does what one can with the data available, and Gerrard presents a useful framework for coaxing some measure of team and individual contributions to overall performance. In his study of the Premier League over four seasons (1998-99 to 2001-02), Manchester United and Arsenal are the two dominant teams that appear from his model.  He does some handwaving of the data to deduce that Arsenal's overperformance was due to an equal measure of purchasing the best players for the team in a systematic and effective matter, while Manchester United's overperformance was the result of out-buying their competitors.  The study can't explain the rise of Bolton Wanderers as an overachiever during the 2000s (they only appeared in the Premiership in 2001-02), nor can it determine the marginal payroll cost of a league point due to lack of data.  Some limited payroll data might be easier to come by when the UEFA FFP is implemented, and in general it would be useful to revisit this study based on what we learned in the previous decade.

Gerrard presents some explanation for a resource-based strategy based on the systematic use of match performance data, drawing from the experience of the Oakland A's and developing some informed guesses of who might have been pursuing such a strategy in the English Premier League in the turn of the century.  The strategies revolve around understanding the market, collecting the right data and converting it into internal information, and acting on that information.  It is the middle point that introduces issues of technology and modeling, but the last point that introduces issues of sporting culture which are more difficult to overcome.  It is true that the concepts of resource- and information-based strategies find fertile ground among the small market teams which have little to lose in a competitive sports league (Oakland, Bolton), but large-budget teams are starting to realize that this approach also makes sense for them in making the best choices with larger resources (Boston Red Sox, and almost certainly Barcelona).  Gerrard says that the Moneyball approach requires someone willing to champion the idea.  What is needed even more is for these champions to become champions.