The Sabermetric Revolution: Assessing the Growth of Analytics in Baseball (21 page)

But what about David Ortiz—the team’s offensive leader in 2004, with an OBP of .380 and an OPS of .983, along with his forty-one homers and clutch hitting? It turns out that the signing of Ortiz had nothing to do with sabermetrics. Red Sox standout pitcher, Pedro Martinez knew Ortiz from playing with him in the Dominican Republic winter leagues and urged the front office to sign him, despite Ortiz’s history of knee problems. (Of course, the front office exercised the judgment to act on Martinez’s recommendation.) The rest is history.

Or, consider the case of Dustin Pedroia. Pedroia was the sixty-fifth pick in the 2004 amateur draft. His physical attributes were unimpressive, but a careful analysis of his numbers commended him to some. Traditional scouting based on the five tools would have had him fall well below the second round of the draft.
⁴
While the Red Sox sabermetrician-in-chief Bill James liked Pedroia, he did not recommend that the Red Sox spend a second-round pick on him. That decision was made higher up. Pedroia, of course, went on to become one of the best second basemen and premier overall players in baseball over the past decade, winning the American League’s MVP Award in 2008. Many would have credited sabermetric insight for Boston’s prescient selection of Pedroia in the second round of the draft. But Nate Silver says what distinguished Pedroia is something that only intelligent scouts could have seen—his exemplary attitude and mental attributes.
⁵

Thus, attempting to judge saber-intensity from outside an organization and based on an organizational chart or media guide is problematic. We choose instead to develop objective indicators of a team’s approach to rosterbuilding and game strategy.

As discussed in earlier chapters, one of the most fundamental of the sabermetric insights, and one emphasized in
Moneyball
, is that walks are almost as good as singles; or, more precisely, both walks and singles share the essential and important characteristic that they do not use up one of the three outs per inning and they each result in a man reaching first base.
⁶
For this reason, along with the fact (detailed in
Chapter 3
) that OBP is more closely associated with runs scored, sabermetricians have long argued that OBP (onbase percentage) is a superior metric to BA (batting average).

OBP, then, is a useful measure of how sabermetrically inspired a team’s roster is in practice. Higher OBPs can result both from the signing of players from outside the organization and from coaches emphasizing the importance of being selective at the plate (although not all hitters can assimilate this skill). It is not sufficient, however, to simply compare one team’s OBP to another’s to identify each team’s saber-intensity. This is because teams with a higher payroll are likely to have players with higher BAs as well as higher OBPs. Instead, to neutralize the effect of payroll, we consider the ratio of OBP to BA for each team. We call this metric
onbase
.

Since ballparks have different dimensions with regard to the distance of the fences from home plate, the amount of foul territory, prevailing wind patterns, effect of the batter’s eye in centerfield, and so on, teams that play half their games in certain parks will either be advantaged or disadvantaged with regard to certain metrics. For both OBP and BA, as well as for other performance metrics used in this chapter, we adjust for park and other factors; our adjustment methodology is described in the
Appendix
.

It will likely come as little surprise that Oakland dominates the list of teams that score highly in the
onbase
metric, claiming seven of the top twenty-five spots, as shown in
Table 13
. (The values are normalized; the average score for an MLB team is 1.00, so the 1999 Oakland A’s, with the highest score during 1985-2011, were 7.2 percent above the overall average.) It is also noteworthy that the
Moneyball
year of 2002, as designated by Michael Lewis, does not make it onto this list, although the teams from Oakland’s three previous seasons do.
⁷

Table 13. Highest
onbase
Scores

Although OBP is probably the best-known indicator of saber-intensity, there are several other candidates. In the area of game strategy, sabermetricians have long observed that, under more circumstances than many people think, it does not make sense to sacrifice bunt, and thereby give up one of the three outs in an inning (see discussion in
Chapter 3
). Both the expected run matrix and linear weights methods show that on average using the sacrifice bunt does not increase run output.
⁸
Because of the fact that pitchers bat in the National League and not the American League, the NL has a higher rate of using the sacrifice bunt. Hence, as with our other metrics, our sacrifice bunt metric is calibrated relative to the average number of sacrifice bunts in the league. We call this metric
sacbunt
.
⁹

Similarly, sabermetricians have maintained that unless a baserunner can steal bases with a high success rate (generally over 65 percent), the extra base gained is not worth the potential loss of an out. Accordingly, other things equal, high saber-intensity teams will tend to use the sacrifice bunt and stolen base strategies more judiciously. Naturally, some teams find themselves with superlative baserunners on their rosters who may be successful in 90 percent or more of their stolen base attempts. There would be no reason to restrain the aggressive instincts of such stolen base leaders. Accordingly, our metric
for saber-intensive base theft strategy is based on the difference 0.18 SB - 0.32 CS, where SB is stolen base and CS is caught stealing, which equals the expected run value gained from a successful stolen base minus the expected run value lost from an unsuccessful attempt. Our stolen base metric is
brun
.

Another area of sabermetric focus, although the initial insight dates back to Henry Chadwick in 1872, has been fielding. The traditional measure of fielding proficiency is fielding percentage (FPCT). As indicated above, this metric does not take account of a fielder’s ability to get to a ball, nor in many instances to make a strong, accurate throw. A standard sabermetric measure is DER, or defensive efficiency rating. DER encompasses a team’s fielders’ ability to reach balls in play and to make the necessary throws. Our metric here is the ratio of DER to FPCT, or
der
.
¹⁰

As discussed in
Chapter 4
, initiated by the work of Vörös McCracken during 1999–2001, sabermetricians have argued that the traditional measure of pitching performance, earned run average or ERA, is significantly limited. ERA is dependent on team fielding and good fortune. McCracken observed that the typical pitcher’s BABIP (batting average on balls in play) showed only a very low correlation from one year to the next. In contrast, the typical pitcher’s home runs allowed, strikeouts, and walks showed high consistency over time. Accordingly, sabermetricians developed a new metric, called Fielding Independent Pitching (FIP), as a truer measure of pitching skill and performance. Because FIP is constructed according to a formula that typically compresses its values between 3 and 4, it is not possible to derive a mathematically meaningful measure from the ratio of FIP to ERA.
¹¹
In this case, since the saber-savvy metric FIP is lower for better pitchers, we take its inverse, 1/FIP, to arrive at our pitching index
fip
.

Our final metric of saber-intensity represents the ability to hit for power. The standard sabermetric measure of a batter’s power is Isolated Power (ISO). In contrast, the conventional measure of power that incorporates all extra base hits is slugging percentage (SLG). As is well known, SLG gives one point for a single, two for a double, three for a triple, and four for a home run and divides the total points by the number of at bats. This relative weighting of base hits does not accord with their relative contribution to runs produced; moreover, it includes singles which do not represent power hitting and it
weights triples as representing 50 percent more power than doubles, even though the evidence suggests that triples are not necessarily hit harder or further than doubles. ISO, in contrast, usually scores doubles and triples equally and does not include singles.
¹²
By the same logic as used for earlier metrics, our measure of saber-intensity with regard to power is ISO divided by SLG, represented by
iso
.

On the basis of these six metrics (
onbase, sacbunt, brun, der, fip, iso
), we construct a composite index of saber-intensity for each team. Before we present our findings, however, it is necessary to take a short detour on shifting labor inefficiencies to provide a context for their interpretation.

One of the central tenets of the moneyball philosophy lies in the relationship between the value of certain skills (such as having a good eye, as manifested in the walk rate) in producing wins and the value of these skills in baseball’s labor market. If there is a disjuncture between these values, then there is a market inefficiency that the astute general manager can exploit.

On one level, it is surprising that baseball’s labor market would have such inefficiencies. After all, we are better able to measure an individual’s productivity in baseball than in practically any other industry. There are measurements in baseball for just about everything. Yet there is a good deal of evidence that such inefficiencies have existed.

What is true, however, is that the abundance of publicly available data on player productivity, team payrolls and team success makes it unlikely that the inefficiencies will persist. Other teams will soon notice the imbalance between value and price, and change their behavior. Thus, the A’s apparently discovered the inefficiency that walks were undervalued in the late 1990s. The evidence we present below is that the undervaluation of walks began to adjust prior to 2003.

As we consider the evidence, it is important to keep in mind that baseball’s labor market has built-in rigidities. First, during the first three years of major league experience,
¹³
players are under reserve to their teams. That is, players are not free to go into the labor market to receive salary bids that reflect
their performance; rather, they belong to the team and usually receive MLB’s minimum salary ($490,000 in 2013) or a salary close to it. Until the player has six full years of major league experience, the player cannot go to the labor market as a free agent, but he is eligible for salary arbitration. In the arbitration process, the player’s salary is expected to approach a free-market level. Many teams sign arbitration-eligible players to long-term contracts. After six years of major league experience, players become free agents and the better ones sign long-term contracts, many lasting for four, five, six, or more years.

Because of the fact that players under reserve have salaries divorced from their productivity and a large share of the other players have salaries that are based on their productivity from earlier years, it is not to be expected that salaries will adjust immediately once a market inefficiency is discovered. Thus, we would not expect, even if the publication of
Moneyball
were the catalyst to change the behavior of baseball front offices, that there would be a sudden, dramatic change in labor market skill valuations in 2004. Rather, the adjustment would come gradually, beginning in 2004. Yet, as we discussed in
Chapter 1
, change was already under way in baseball front offices prior to the publication of
Moneyball
and our evidence suggests that the labor market had already begun to shift prior to 2004. When we consider the return in salary to a player’s walk rate (share of plate appearances when a player reaches base with a walk or hit by pitch), we find that the uptick begins in 2002 (
Figure 12
). Recall that if
Moneyball
were the catalyst for this correction, we would not even begin to see the correction before 2004. Also, as shown in
Figure 12
, the correction did accelerate in 2004, but then it seemed to reverse in the next two years.

We must be cautious, however, in placing too much emphasis on the year-to-year changes because there are random fluctuations in the annual labor market conditions (for instance, the number and quality of free agents available in any given year can vary considerably) that could impact these results. The final picture that emerges is the one described earlier: a market correction begins prior to the publication of
Moneyball
, accelerates after the publication probably to the point of overcorrection, and then adjusts moderately.

Nevertheless, if
Moneyball
did not initiate the labor market correction, it certainly contributed to it. Indeed, the return to the skill of working a walk
increased by around 64 percent between the 1998–2002 and 2003–2007 periods.
¹⁴
Our evidence also indicates that the market may have overcorrected slightly, as the return to the skill of walking fell by approximately 12 percent during 2008–2012. As we shall see, the A’s strategy shifted, suggesting that Billy Beane perceived that he could no longer take advantage of the undervaluation of the skill of walking.
¹⁵