The Sabermetric Revolution: Assessing the Growth of Analytics in Baseball (18 page)

In the NFL, with the possible exception of the New England Patriots, there appears to be little correlation between the adoption of analytics and playing field success. Football metricians, however, do seem to have contributed important insights with regard to game strategy. The most significant, and one on which there is unanimity, is that coaches are too conservative on fourth down. Yet, according to a recent study by Football Outsiders, over the period 1992–2011 there has been no increased tendency for coaches to be more aggressive on fourth down.
³³

Football game strategy can also be informed regarding the best choices for approaching overtime, when to challenge a call, or when to rush three versus four linemen. Tabulating the experience of other teams against various opponents can help inform a team’s tactics. Still, opening up these questions to statistical analysis also geometrically increases the potential processing of data and threatens to overwhelm GMs and coaches alike. In a world where the assimilation of such mountains of information and its analysis is new, the obstacles to its effective use are manifold. Thus, even though developing analytical insights in football with regard to game strategy may be more forthcoming than with personnel decisions, the cultural divide between the coaching staff and metricians in the front office is still wide and has held back what appears to be desirable change in game strategy.
³⁴

Ben Alamar and Vijay Mehrotra have aptly commented on the cultural chasm in NFL front offices: “Most decision-makers have little to no experience or training in the methods and tools of analytics, and as such are not well-equipped to evaluate the landscape of options. The result is that some organizations start small, at best thinking very incrementally about analytics and at worst simply adding a small amount of staff and/or software as window dressing. Meanwhile, other organizations have absolutely no idea of how to begin and thus simply do nothing.”
³⁵

Of course, much of this observation also applies to baseball. Yet, outside baseball, analytics got a later start, tends to be more proprietary, and struggles with the less tractable nature of the sports themselves.

One football GM told a
New York Times
reporter: “Ideally, you want the
objective and subjective to match up. The NFL is about resource allocation—you have a certain number of salary cap dollars and draft picks. If you found any area of the market that may be undervalued, you want to keep that information. At the end of the day, the tape is going to be our first choice. They have to look good on film.”
³⁶

Analytics outside baseball, while growing, still seems to be taking its baby steps, mostly outside of public view.

Quantitative analysis has been increasingly introduced over the last two decades to understand the business of baseball (and other sports). It has been applied to a variety of issues, such as efficient ticket pricing, regulation of secondary ticket market policies, impact of stadiums and teams on a local economy, threshold city size for hosting a team, franchise valuation, the relationship between player development expenditures or major league payroll and team performance, labor market institutions, and optimal competitive balance, among others. Entire books have been written about these subjects, so in this brief chapter we only wish to illustrate an area of this research that directly affects the sabermetric concerns of player performance and team strategy.

It is the received wisdom that sports leagues, in order to command the interest and attention of an adequate fan base, must attain a certain degree of competitive balance across all the teams in a league. Without such a modicum of balance, the uncertainty that drives the suspense and excitement of league games and championship seasons is compromised, and game attendance as well as television ratings will suffer.
¹

Exactly how much balance is needed to maximize league revenues in the short and long runs, however, has never been established. Moreover, it is clear that the optimal degree of balance varies considerably between open and closed league systems, as well as from league to league within each system.

Notably, the soccer leagues in Europe, which operate with an open, promotion/relegation structure, demand less balance. This is because several contests during the league year are of interest to the fans: winning the league, not being relegated, gaining a berth in the following year’s Champions League or Europa competition, along with various domestic competitions. In contrast, in the U.S. closed leagues, gaining a postseason berth and winning the championship is all that really matters.

Two other factors are also prominently at work. First, for a fan of Manchester United, Bayern Munich, Barcelona, InterMilan, or any other team, the two biggest prizes are winning the domestic league and winning the pan-European Champions League. If there is too much balance in the domestic league, then the domestic league’s best teams will not have the requisite resources to dominate in the pan-European competition. Therefore, the European football leagues generally engage in much less revenue sharing than the U.S. leagues.

Second, the promotion/relegation leagues contain an automatic balancing mechanism that is not present in closed leagues. In open leagues the number of teams that play in a particular city is ultimately determined by market forces; thus, there have been between five and seven teams from greater London in the English Premier League in recent years. In the closed leagues, the market power of large city teams is protected by administrative controls exercised by the league monopoly. No team in the English Premier League contains the natural geographic advantage that, for instance, the New York Yankees or Los Angeles Dodgers have in Major League Baseball.

For these reasons, the U.S. closed leagues all depend on the equal distribution of 100 percent of central television, radio, Internet, licensing, and other revenue sources generated by the league office. They also all now have a supplementary revenue sharing system. In the NFL, which has had extensive revenue sharing since the league’s inception in the early 1920s, the policy is
the most penetrating, encompassing about 75 percent of all league-wide revenues. In the NHL, NBA, and MLB, where either little or no revenue sharing was present at the beginning of the leagues, the amount of shared revenue is less than 30 percent of league-wide revenues. Naturally, team owners from large cities in the latter leagues have a tendency to resist additional sharing because they paid a much higher price for their franchises than did the owners from small cities. They tend to view such sharing as confiscatory. The conflicts among owners over revenue sharing frequently spill out into labor conflicts and other inefficiencies.

Among those leagues that did not begin with revenue sharing, MLB has the most extensive sharing. MLB’s supplementary system was introduced after the devastating 1994–1995 strike and has been expanded steadily, from under $30 million being shared in 1995 to nearly $400 million in 2012. Via various formulae, MLB has transferred these quantities of revenue from the high-revenue teams to the low-revenue teams. The basic idea has been twofold: on the one hand, by taking revenue away from the rich teams, to reduce high team payrolls, and, on the other, by giving revenue to the poor teams, to raise low team payrolls. That is, the system is supposed to result in payroll compression, which, in turn, is supposed to promote greater performance balance across the teams.
²

Note that one issue here is that the explanatory power of payroll on win percentage in MLB, while usually statistically significant since 1990, tends to vary between 10 and 35 percent in any given year. This means that between 10 and 35 percent of the variation of a team’s win percentage in a particular season is explained by that team’s major league payroll. Or, stated differently, between 65 and 90 percent of the variation in win percentage in any year is explained by factors other than team payroll. This implies that even if MLB’s system of revenue sharing resulted in appreciable payroll compression, the impact on the compression of win percentages may be less noticeable. Nonetheless, it is reasonable to assert that greater payroll compression will raise the probability of more compression (reduction in standard deviation) in team win percentages.

It is also interesting to note for our purposes that the use of inferior performance metrics (e.g., batting average instead of OBP or OPS) will tend
to lower the correlation between payroll and performance, while the use of improved metrics will tend to raise this correlation. That is, since improved metrics, by definition, show a tighter relationship with win percentage, their adoption will enable teams to allocate payroll more efficiently and increase the correlation between payroll and win percentage. If the spread of sabermetric knowledge within baseball has resulted in improved metrics over the last decade, then, other things equal, we should witness an increased correlation between payroll and win percentage. In
Table 6
, we show what has happened to the relationship between payroll and win percentage since the publication of
Moneyball
in 2003.

Table 6. Relationship Between Team Payroll and Team Win Percentage, 2003–2011

The first line in the table shows the R
²
or the percent of the variation in team win percentage that is explained by the variation in team payroll.
³
The second line is a three-year moving average of the same in order to smooth out the inevitable random year-to-year fluctuations. It is very difficult to identify a clear trend from the data on either line.
⁴
This, of course, doesn’t mean that the new metrics are not valuable or that the spread of sabermetric knowledge is not happening. It simply means that these tests are abstracting from other pertinent information and are inconclusive.

To properly grasp the empirical effect of revenue sharing on MLB’s competitive balance, it is instructive to consider the trend before supplementary revenue sharing was introduced in 1996.
Table 7
depicts the ratio of the standard deviation of win percentages to the idealized standard deviation of win
percentages (RSD) since 1903. The lower the ratio, the greater the competitive balance. (The idealized standard deviation is that which would obtain if talent were equally distributed across all teams, given the number of games played in a season.
⁵
The RSD is the most commonly used measure of competitive balance, because it facilitates comparisons within a league over time or between leagues. It explicitly controls for the fact that the more games in a season’s schedule, e.g., 154 versus 162 over time within baseball, or 16 versus 162 between the NFL and MLB, the less important chance will be in determining the distribution of team win percentages. Thus, we would expect leagues playing more games on the yearly schedule to have a lower standard deviation of win percentages.)

By the RSD ratio, there is a clear historical trend toward greater balance in MLB that runs through the 1980s. The ratio remains constant in the 1990s, halting the historical trend, the decade during which MLB’s supplemental revenue sharing system was introduced. The ratio then proceeds to increase in the 2000s, reversing the historical trend, a decade when MLB’s revenue sharing roughly tripled to some $400 million annually.

The principal reason for the secular downward trend through the 1980s (i.e., toward greater balance) is long-term talent compression. This has occurred as the percentage of the population (from the United States and other baseball countries) that plays in the major leagues has decreased over time, since the population has grown considerably faster than the number of teams.
⁶

Table 7. Ratio of Standard Deviation to Idealized Standard Deviation (RSD)