The Sabermetric Revolution: Assessing the Growth of Analytics in Baseball (22 page)

Figure 12. Estimated Labor Market Valuations, 1986–2012

Each dot represents the estimated increase in ln(Salary) for each additional unit of Eye in the previous season. A trend line for Eye is indicated. Eye is walks plus hit by pitch per plate appearance. See
Appendix
for fuller discussion.

Another way to consider the shifting labor market inefficiency is to evaluate the relative return to the walk rate versus the return to batting average. In
Table 14
we present the results of a regression of the log of Runs Scored (RS) on the logs of Walk Rate (WR), Batting Average (BA), and Power.
¹⁶
The resulting coefficients yield estimates of the elasticity of each batting skill on runs scored;
that is, the percent change in RS divided by the percent change in WR (or other hitting skill). For instance, the 0.267 coefficient indicates that during 1985–1997 a 1 percent increase in a team’s walk rate increased the number of runs the team will score by 0.267 percent.
¹⁷
In
Table 15
, we present the results of a regression of the log of individual player salary on the logs of lagged WR, BA and Power. The 0.138 coefficient, for instance, indicates that during 1985–97 a 1 percent increase in a player’s walk rate increased his (next year’s) salary by 0.138 percent. For our purposes, it is interesting to follow row 1 in
Table 15
across the different time periods. It shows that the salary return to a 1 percent increase in a player’s walk rate increased in each of the next two time periods, 1998–2002 and 2003–2007, by 6 percent (from 0.138 to 0.146) and 64 percent (from 0.146 to 0.239), respectively, before turning down in 2008–2012. These results, of course, are consistent with the argument that teams began to assign greater value to a good eye and patience at the plate in the late 1990s and first years of the 2000s and that this growing return accelerated after the publication of
Moneyball
.

Table 14. Impact (Elasticity) of Walk Rate (WR), Batting Average (BA), and Power on Runs Scored (RS)

Table 15. Impact (Elasticity) of Walk Rate (WR), Batting Average (BA), and Power on Player Salary

It is also relevant to look at the returns to batter skills in relation to the productivity (runs produced) by each skill; that is, the ratio of the elasticity of salary to the elasticity of RS for each skill, as depicted in
Table 16
. Here we see that a similar pattern holds up over the periods. The salary return to WR relative to the run scoring contribution of WR accelerates during 2003–2007, and then falls off during 2008–2012. This pattern is consistent with our earlier contention that what might have been sound sabermetric advice about labor market strategy (exploiting the undervaluation of walks) ten years ago might not be sound advice in recent years. This insight must be kept in mind when evaluating the saber-intensity of team front offices.

The particular relevance of this pattern to our discussion is that, after the market correction, we should not expect saber-oriented teams to continue to emphasize player attributes that were previously undervalued. To wit, we would not expect the Oakland A’s necessarily to continue to be among the top teams in walk rate or the ratio of OBP/BA (
onbase
). And, indeed, during the period 1998–2008, the A’s led the major leagues in
onbase
with an index of 1.041, but then abruptly fell to an average of 1.005 during 2009–2011. Similarly, the A’s strategy with regard to baserunning shifted abruptly as their index of runs created from a stolen base jumped by 60 percent from the 1998–2008 period to 2009–2011.

As presented above, however, we have developed six indicators of saber-intensity, and it is useful to consider their overall effect. To do this, we introduce a composite index of saber-intensity. In particular, we combine the six indicators by weighting them according to the contribution that each element of skill or strategy makes to a team’s win percentage. To estimate this contribution we run a multiple regression of the log of win percentage on the logs of OBP, ISO, DER, FIP, baserunning (BRUN), and sacrifice bunts (SACBUNT). The resulting coefficients tell us the percentage contribution that each skill or strategy element makes to a team’s win percentage. All the coefficients are statistically significant at standard levels,
¹⁸
and we weight the individual metrics according to these coefficients in our composite index, as follows.
¹⁹

Table 16. Relative Return to Batting Skills (InSal/InRS)

Saber-Intensity Weights:
   .237 OBP,   .037 ISO,        .290 FIP,

.433 DER,   .001 BRUN,   .002 SACBUNT

Note that DER has the highest weight, followed by FIP in a distant second, and then OBP close behind. Although ISO, BRUN and SACBUNT are all statistically significant at the .10 level or better, the magnitude of their impacts is much smaller. We then apply these weights to our six saber-intensity skill elements to generate our composite SI index. It is also of interest that the defensive and pitching components to the A’s success in the early 2000s were virtually omitted from Lewis’s analysis in
Moneyball
. This provides further corroboration of our contention that more of the A’s success was attributable to their superior pitching staff than to their penchant for walking.

In
Table 17
we present the top thirty scores in our saber-intensity index since 1985. Most of the names are familiar. The Atlanta Braves of the 1990s captured first place as well as four of the top eleven places in our composite index. While these teams were not known for embracing the sabermetric philosophy, the intelligence of their front office personnel was impressive and these teams were characterized by extremely strong pitching (with three Cy Young Award winners on the same team in Greg Maddux, Tom Glavine, and John Smoltz) and defense (with Otis Nixon, Rafael Belliard, Andruw Jones, Terry Pendleton, Ron Gant, and Marquis Grissom).
²⁰
The 2007 Red Sox and the 2001 A’s, two successful teams widely recognized for their saber-intensity, are in a near dead heat with the 1995 Braves. The championship 2004 Red Sox, the Oakland A’s in 2003 and in 1990 are in fifth, seventh, and eighth place respectively. The 1990 A’s were, of course, led by Sandy Alderson, with sabermetrician Eric Walker by his side, and featured players like Rickey Henderson (high walk rate) and
Dennis Eckersley (high strikeout-to-walk ratio) who would later be identified as sabermetric darlings. Beyond eighth place, there are few surprises with most of the teams inclined toward sabermetrics or with dominant pitching and tight defense (e.g., the 2008 Tampa Bay Rays). Our composite metric appears to correlate well with the prevailing wisdom on sabermetrically oriented clubs.

It is also instructive to consider how the teams rank over different time periods. For instance, during the 1992–97 period Atlanta received the highest average score, followed by Baltimore and Montreal, as shown in
Table 18
. During the immediate pre-
Moneyball
period, 1998–2002, the saber-intensity leaders were the New York Yankees, Boston, and Oakland (although their scores were practically identical.) During 2003–2007, the top three remained the same with the Yankees sliding below Boston and Oakland. Notably, during this period, the Tampa Bay Rays were in twenty-ninth place in the ranking. After the 2005 season, the Rays were purchased by Stu Sternberg. Sternberg brought in an Ivy-educated, investment banking team of executives and Andrew Friedman as GM and began to rebuild the team. The Rays rose from next-to-last to first place in the SI ranking during 2008–2011.

Table 17. Top Thirty Scores in Saber-Intensity (SI) Index, 1985–2011

It is also notable that the A’s slid from the top three down to tenth during 2008–11. This is consistent with our observation that Billy Beane appeared to switch strategies as formerly undervalued skills became fully valued. It bears noting as well that the A’s surprising first-place divisional finish in 2012 is marked by a reemphasis on saber-intensity: in 2012, the A’s ranked second in
obp
, third in
der
, and fourth in
iso
among all major league teams (and twelfth in
fip
).

Our next step is to see if the teams that had high SI scores in particular years also had high winning percentages in those years. To do this, however, we must first adjust for the fact that some teams benefitted from much higher payrolls than other teams. We do this by first estimating the impact of payroll (PAY) on win percentage (WPCT). PAY is considered as each team’s payroll relative to the average major league payroll in each year. Because we expect there to be some diminishing returns to higher and higher payrolls, we estimate the following equation, again for the years 1985–2011:

Table 18. Top Three Teams, Average Saber-Intensity (SI) by Period