Long live baseball analysis

Is Baseball analysis dead?

Gary Huckaby, in a recent article for Baseball Prospectus, announced that “baseball analysis is dead.” Gary is the founder of Baseball Prospectus, a site that prominently features a lot of baseball analysis. In fact, we analyze baseball here at The Hardball Times, too. So what’s going on? Should we all hang up our figurative spikes and move onto other causes, like world peace or something? Is baseball analysis truly dead?

First of all, let’s acknowledge that baseball analysis didn’t start in the late ’70s, with Pete Palmer, John Thorn and Bill James. It began about the same time the game was first played, as chronicled in Alan Schwarz’s outstanding book, The Numbers Game: Baseball’s Lifelong Fascination with Statistics. People like Henry Chadwick, Ernie Lanigan, F.C. Lane and Alan Roth made baseball analysis their lifetime pursuit before most of us were born. And future statisticians will continue to do so as long as the game is played.

Analyzing baseball adds to our enjoyment. It gives greater meaning to something that is, let’s face it, just a game. Baseball analysis allows us to invest more than just our competitive spirit and emotions in the sport; it allows us to engage our minds, too. It deepens the relationship.

But that isn’t what Gary was really writing about. Once you got past the eye-catching opening line, Gary said this:

I’m strictly talking about activities like developing value metrics, forecasting, and all the other stuff we do with the massive yarn-ball of data we’ve all put together over the years.

Baseball analysis is dead because its utility has pretty much vanished.

Gary isn’t saying that everyone has bought into the findings of baseball analysis (John McLaren critics, take note), but he is saying that most baseball front offices have managed to build their own statistical systems without a lot of effort, and have bought into the fundamental findings of baseball analysis. Hence, baseball analysis is dead. There’s nothing left to gain.

This is a reprise of a theme he presented in Baseball Prospectus 2006: The BP Team of Experts on Baseball Talent, in which he interviewed an anonymous general manager (or was that a literary device?) and summarized his thoughts this way:

Performing analysis at a high enough level to help a club, in some way, shape or form, just isn’t that hard to do. Investing some time in the basics of statistics, data design, and operations research is something that literally millions of people can do, and most ex-players, coaches and even clubbies can achieve. The hard part is selling the information as the core of a decision-making process.

I don’t mean to read into Gary’s words, but it seems to me that he is totally focused on selling baseball analysis to major league clubs. Yet there are other audiences and uses for the numbers we plumb.

Baseball Prospectus has been down this road before. Back in September of 2004, Nate Silver wrote:

It is also probably true that the pace of discovery within sabermetric circles will slow as more and more data is analyzed and more and more conclusions have been proclaimed. Baseball, while a wonderfully complex game, is nevertheless a closed system, and the returns on further research efforts are likely to diminish.

A year and a half later, Nate wrote:

Of all of the sentences that I’ve written on behalf of Baseball Prospectus, (those) are the two that I regret the most.

Nate came to regret his words because he began working with Dayn Perry on a chapter for Baseball Between the Numbers (subtitle: Why Everything You Know About Baseball is Wrong—now, isn’t that a strange subtitle for a book whose subject has been declared dead two years later?) and found that pitching and defense really do matter in the playoffs (something our own Vinay Kumar had also uncovered). Nate then wrote:

You can make plausible arguments for the death of the novel, or the sitcom, or the great American rock band, but the genre of baseball analysis is far from exhausted. In fact, I now believe that we’re really just at the beginning stages of the discovery process, and that destinations like Baseball Prospectus are poised for a breakout year.

So you’ve got Gary on one side of analytic relevance, and Nate on the other. Now, the cynic in me wants to point out that Nate was selling a book while Gary was selling to ballclubs, which may have skewed their perspectives. And yes, I do get cynical at times. But while I’m not going to argue with Gary’s point of view—I have no idea what major league clubs think or do—I think Nate has the more pertinent perspective.

Let me tell you up front that I don’t really care what baseball clubs find useful. I’ve never wanted to work for a baseball club. I just like to think about baseball and share the results with you. Actually, lots of people like to do the same thing. And we’re not going to go away.

I guess there really is nothing fundamentally new to discover about run estimation formulas, such as runs created, base runs and linear weights. The fundamental offensive insights have been made, and any new insights will probably be so marginal that they won’t make a wave in the baseball firmament, outside of those of us truly passionate about the margins. But as I will attempt to show below, there are layers and layers of insightful analysis still to come.

Likewise, Voros McCracken’s “DIPS theory” seemed to cement our understanding of pitching by articulating and reinforcing the primacy of basic pitching stats: strikeouts, walks and home runs allowed. However, recent research by Tangotiger (and others) has provided a much more nuanced (and valid) view of the DIPS phenomenon.

Tango found that pitchers do indeed have some control over their batted balls in play, but it takes several seasons before that talent becomes manifest. In fact, it can take seven or more seasons for the stats to truly point out which pitchers excel at inducing outs from batted balls. The talent is there, it just lays overwhelmed by statistical random variation for some time.

Now maybe major league teams aren’t interested in these nuances. But I am, foolish me. And there are enough readers out there who seem interested, too (including some from Major League Baseball). In fact, I think this is the golden age of baseball analysis. Let me give you an example of some pretty darn interesting analysis that is only now possible, thanks to the widespread availability of baseball stats. At least, I think it’s interesting.

Some baseball analysis

What has happened to David Ortiz this year? His batting average is running at a career-high .318, but that’s mostly because his BABIP is also running at a career-high .349. He’s not hitting more line drives; his high BABIP is the result of outfield fly balls falling for hits more often (30% of the time, not including home runs. The major league average is 16%), and his ground balls have found holes more often, too (19% have been hits; the major league average is 15%).

His slugging percentage is down and, most critically, his home run rate per outfield fly is down (16% vs. 27% last year). That’s a real concern to Red Sox fans, but it’s not what I’m wondering about.

Ortiz hasn’t been winning ballgames as he has in the past. Why do I say that? Because his Win Probability Added (WPA) is only 3.35, in line with his ’03 and ’04 seasons, but not nearly as good as his past two years (8.93 and 8.03). For the past two years, Ortiz has been hailed as a great clutch hitter, but Fangraphs pegs his current clutch “score” at -1.4, meaning he’s given up 1.4 wins less than average by not hitting “in the clutch.” The past two years, his Fangraphs clutch scores were positive 2.9 and 1.5.

On the other hand, we also track a statistic called “clutch” here at THT, and it shows that Ortiz has added 2.6 runs due to his situational hitting (he has batted .346 with runners in scoring position). So has Ortiz been clutch this year, or not?

The concepts of WPA and clutch hitting are mighty controversial. I wouldn’t dare raise them in a bar full of angry Primates. But the two concepts are very useful for baseball analyses like this.

Fangraphs tracks the WPA of each player, and it also tracks another stat called WPA/LI. WPA/LI is an awesome stat because it basically neutralizes the game situation in WPA. See, many people object to the fact that WPA sometimes values a ninth-inning single more than a first-inning single (or, in some cases, even more than a first-inning home run) in the same game (close and late situations, for instance). While this doesn’t bother me as much, I understand their concerns. WPA/LI fixes this.

LI stands for Leverage Index (invented by Tangotiger), one of the most important analytic concepts of recent years. LI has several uses but, in this case, it’s being used to “neuter” the game impact of a WPA situation. In other words, a batter’s WPA/LI calculates how much he has helped his team score runs, not win games. It essentially mimics the approach taken by Tom Ruane in this research, calculating the batter’s impact on the base/out/runs situation, but not the inning or margin.

Is this too analytic for you? I apologize, but it’s important for understanding David Ortiz’s contribution to the Sox.

The bottom line is that David Ortiz is 11th in the American League in batting WPA, but he’s second in WPA/LI. In other words, once you take the game context out of his WPA stats (but not the base/out context), he looks much, much better. So David Ortiz has been a much stronger contributor to the Red Sox’s wins than WPA indicates, right?

Well, hold on a second. I want to make the case that game context still counts. Maybe you don’t believe there’s a difference between a first-inning home run and ninth-inning home run in the same game, but I believe there is definitely a difference between a home run in a one-run win and a home run in a ten-run win. In general, hits in close games have a bigger impact on winning than hits in blowouts.

To prove the point, I pulled all the WPA events from 2006 and calculated the average impact of each event on the win probability of the batter’s team. Not surprisingly, I found that the same type of event (such as a single) in close games has a larger win impact than that type of event in blowouts. Based on the data, I have estimated a standard multiplier for calculating the impact of a hit or out in a game with a particular victory margin. It’s still under development, but I don’t think the final version will look much different than this:

   Margin  Impact
       1    1.38
       2    1.13
       3    0.97
       4    0.86
       5    0.76
       6    0.66
       7    0.63
       8    0.57
       9    0.51
      10    0.47

You might call this a “Game Leverage Index,” as it shows that a single in a close game is worth about 40% more (in wins) than a single in a three-run game (1.38 compared to 0.97). And it’s worth nearly three times as much as a single in a game with a 10-run margin (1.38 vs. 0.47). The same multiplier can be applied to any type of event.

I was actually surprised that the differences weren’t greater, but this table makes a lot of sense to me. I think it could represent a workable compromise between those who want to value performance in games, but don’t believe that the difference between “when” an event occurred should matter once the game is over.

Back to David Ortiz. It turns out that Ortiz has batted .294 in close games (decided by one or two runs), but .408 in games with a margin of seven or more runs. His contribution has been greater in games in which it mattered less.

One-run games tend to be low-scoring affairs, so you might expect a difference. But that’s a very large difference. To put it in perspective, Vlad Guerrero has batted .324 in close games and .297 in games with a margin of seven or more runs.

To sort out the net impact of these differences, I pulled together a little baseball analysis. First, I calculated 2007 Base Runs (BR) for Ortiz, Guerrero and David Wright (for comparison purposes). Base Runs is one of those run estimator formulas I mentioned before, and arguably the best. Then, I applied the Game Index to the stats of each game for these three players and calculated Base Runs again. You might call this “Game-Adjusted Base Runs” (GI BR in the table below). Here are the results:

                BR    GI BR  Diff
Ortiz           113     74   -39
Guerrero        101     99    -2
Wright          118    119     1

Vlad and Wright are virtually even in Base Runs and Game-Adjusted Base Runs, but Ortiz loses nearly 40 runs when adjusting for the game margin. Although he has hit well with runners in scoring position, he has produced more in one-sided games than close ones. If I’m thinking about this correctly, the impact of his production distribution has been about four wins less than expected (10 runs generally equals one win) because he has produced more in games in which his bat has mattered less.

I’m guessing that most Red Sox fans, having watched Ortiz all year, will call this obvious science. But this is what good analysis does: it affirms what we’ve observed (or not, as the case may be) and quantifies its impact. Long live baseball analysis.

So that’s an example of what baseball analysis is still capable of. I haven’t even touched some of the more complicated subjects, such as fielding or managerial strategy. We’ll be analyzing baseball for a long time here at THT, which is why you should definitely buy the 2008 Hardball Times Annual.

Hey, I may be a cynic, but I’m a self-serving cynic.

References & Resources
Just to be clear, I’m a big fan of Baseball Prospectus. Most of the BPro links included in this article require a subscription, but there’s no reason for you not to subscribe. The batting Game Logs were pulled from Baseball Reference, and the Game Multiplier stats were derived thanks to WPA statistics from Fangraphs.

Dave Studeman was called a "national treasure" by Rob Neyer. Seriously. Follow his sporadic tweets @dastudes.

Comments are closed.