*The Sabermetric Revolution*

That is the new book by Benjamin Baumer and Andrew Zimbalist and the subtitle is Assessing the Growth of Analytics in Baseball.  It is an excellent and well-written look at where sabermetric knowledge stands today, here is one excerpt:

…once the ball has been put into play, it doesn’t seem to matter all that much whether it was put into play against Roger Clemens or Roger Craig.


…an extra win will bring more revenue to a team in New York City than in Kansas City.

This is a very useful book on its chosen topic.


Disclosure: I won't read this particular book because I have previous experience with Zimbalist and his work.

In general, sabermetrics is one of the fields in which the published books are roughly a decade behind the state of the art within the community, which exists mostly on the web, centered on hubs such as FanGraphs and Baseball Prospectus. (Not to mention the proprietary research conducted by MLB teams themselves, often by hiring experts from the existing freelance community, including Bill James himself, Keith Law, and Tom Tango.)

…an extra win will bring more revenue to a team in New York City than in Kansas City.

Stunning. In another section they find that balls hit with more force fly further than those hit with less force. Further, balls caught in the air by fielders result in more outs than those caught by fans in the bleachers. Amazing stuff.

Don't know why you mock this.

This is precisely the factor that the popular sports press consistently miss. People are always asserting that the Yankees can "afford" to pay their players more than the Royals because they have higher total revenue, when it is actually *marginal* revenue that matters.

And though the statement is not really surprising, it is no obvious either. And what are the magnitudes of the differences? The fact that this is discussed at all makes it a big advance over most sports writing.

To be fair, we have known that larger markets get higher marginal revenue from an additional victory for quite some time. In the sports economics literature, this has been known for more than 50 years, since at least Rottenberg's piece in the JPE in 1956.

its pretty obvious

Maybe it is not obvious. Maybe we needed all this mathematical horsepower (or horse something) to discover that which I knew as a five year old in the 1970's. Or, it was obvious.

I think the fun with numbers crowd gets far too much respect. Their contributions to the game of baseball are trivial[1]. I think the biggest thing we get from FanGraphs is our TV's are now littered with communications majors pretending to be quants.

[1] For the economics majors I was exaggerating for effect.

Were all the "obvious" things you knew at five later turn out to be just so? Could that explain why we still need studies to verify the "obvious"?

Haha I love the idea of measuring computers in horsepower.

How smart is my iPhone? By some measures (quick math), it is like 1000 horsepower. By others (recognizing wolves or mares), it's like .001 horsepower.

So my iPhone has like 1 horsepower of computing, I think.

Trivial? If you really think that then you are hopelessly uninformed in the subject matter.

Baseball is being transformed by this stuff. It's drastically changing the way front offices makes decisions.

For those not familiar with sabermetrics, "put in play" in this context does not include balls hit off of the respective pitcher which become home runs by traveling over the fence.

This is a very good point that is often misunderstood.

Although even those balls in play are subject to similar forces given current research - i.e. HRs per flyball will regress to around 10%.

I've begun reading the book and attended a presentation by Prof. Zimbalist. I look forward to the later chapters in which sabermetrics is applied to measure the effectiveness of sabermetrics itself.

Agreed that it is well-written. It also clears away the many misconceptions created by *Moneyball*.

The discussions of the evaluation of defense and the construction of the "Wins Above Replacement" ("WAR") metric would have been improved considerably if the authors had been aware of my 2011 book, *Wizardry: Baseball's All-Time Greatest Fielders Revealed* (Oxford University Press) (the link above is the OUP website for the book). Here was the Wall Street Journal review: http://online.wsj.com/news/articles/SB10001424052748704893604576200792720588366. Notwithstanding that favorable review, the book and its ideas have had trouble breaking through, possibly because Bill James wrote to his blog readership that he "would not" read the book and other well-known bloggers such as Tom Tango have declined to discuss the model with their readership.

*Wizardry* presented the first ever (and still only) open-source statistical model of baseball defense (pitching and fielding) that can be applied throughout baseball history using whatever public data is available. The model is essentially a forced-zero-intercept, two-stage multi-variable regression model that results in linear equations for rating each pitcher and fielder. The model is called Defensive Regression Analysis or DRA. Directly and indirectly it has resulted in a greater change in the evaluation of players throughout history than any statistical model since Pete Palmer's Linear Weights formula for offense, published circa 1980.

One of the important points at the conclusion of *Wizardry* is that, pace Baumer and Zimbalist, estimates of fielder value provided by DRA have no more 'noise' in them than the measurement of batting performance for plate appearances not ending in a strikeout, walk or home run--that is, at-bats in which the batter hits the ball into the field of play and either grounds out, flies out, gets on base on an error, or hits a single, double or triple.

In other words, the evaluation of how well players handle balls hit 'near' them in the field is now approximately as accurate or meaningful as the evaluation of how effective players are when they _hit_ a ball into the field.

DRA is the open-source component in a composite fielding metric created this past year for determining Gold Glove winners: the SABR Defensive Index, or SDI, which also aggregates various proprietary fielding evaluation systems based on privately collected records of the observed location and trajectory (or hang-time) of every batted ball and, to a great extent, mostly undisclosed calculations with respect to such proprietary data.

*Wizardry* analyzed in depth these proprietary metrics based on their published results. The authors of *Sabermetric Revolution* are understandably frustrated by the 'black box' nature of these systems.

I am currently working on an updated version of DRA and hope to blog about it and more generally about open-source methods for evaluating fielding defense, which remains the only significant part of major league baseball player performance evaluation that is at all controversial.

Anyone with any questions should feel free to contact me directly at Michael.A.Humphreys1@gmail.com.

So does Jeter suck or what?

Yes. Subject to one new piece of information I have just come across, if you subtracted from Jeter's _batting_ statistics a single for every hit he 'allowed' as a shortstop compared to what an average shortstop would likely have made, Jeter would be a lifetime .270 hitter with a roughly league-average on-base percentage and slightly below average slugging percentage. He would not have had a single season of even borderline MVP quality. He would be merely a good and durable player with a few to several all-star seasons.

The one new piece of information I have heard from sabermetrician Sean Smith is that for some reason nobody has figured out, Jeter was much, much worse at Yankee Stadium and only merely bad elsewhere. I have no idea why there should be a significant park effect for shortstops.

This really isn't new. We already have wins above replacement (WAR), in which a player's contributions in different aspects of the game are all quantified in the same unit. So we already know how Jeter's defensive value dragged down his overall value and by how much, subject to how much stock you put in defensive metrics. We don't need to subtract "hits allowed" and whatnot to determine some adjusted batting average. I'm not sure why we would want to anyway.

The park effect for fielders that you mention is sort of interesting though.

The park effect was for Jeter not for generic fielders. Which makes it completely understandable, since Cap'n Jetes is always scanning the NY crowd for his next ... friend, and consequently a little slow to react to the on-field action.

It's quite different if he's claiming his model turns Jester from a 75 fWAR player (hall of fame lock level) to a maybe 25 WAR level player.

That turns Jeter's defense from costing the Yanks a few wins over these 19 seasons to two wins a year.

Zeezil, DRA would drop Jeter from his current ~70 WAR career to about 60. Borderline Hall of Fame. But more importantly, it would deprive him of any even borderline MVP-type seasons and leave him with only a handful of all-star quality seasons. Usually that would not get you into the Hall. Stated differently, imagine he was an average fielder with a twenty-year career who hit .270. Nobody would consider him a Hall of Famer.

I admit that the only thing I know about your measure is what I've read here, but it sounds as though your system is based on a concept of production above average. So of course 'ol Jetes would look worse, as would every player since you're measuring relative to a higher bar.

We don't have to imagine a scenario where Jetes was an average fielder who hit .270 because we can just look up his WAR (whichever iteration you fancy) and compare it to such a player. This is the utility of WAR. However, we can already look at the composition of WAR and make our own hypotheticals if we must. According to Fangraphs, he accounted for about 140 runs below average on defense. So, imagining him as an average defensive shortstop would boost his career WAR from about 74 to about 87. If we want to see what kind of offensive production is equivalent to 13 wins, we can figure that out too. If your system accounts for something current WAR formulas are missing or doing incorrectly, that's one thing. But that doesn't appear to be so.

The OUP website did not show up on the prior message. Here it is: http://global.oup.com/academic/product/wizardry-9780195397765;jsessionid=BC44DC8E60CF6AA498C661DA1D702E4F?cc=us&lang=en&

While there is plenty more to learn about baseball analytics are way past the low hanging fruit of evaluation of players. Really the more interesting sports analytics are happening in basketball. Also, other sports like NFL are just beginning to better understand what makes players good and teams successful. Hockey analytics are at the cuspnof something interesting. To me, soccer knowledge is poised to make the most gains as the analytics movement in that sport is still seen as a bunch of nerd outsiders to be dismissed.

Truthfully, baseball has lost the most interest do me as the sabr movement has made me look at the game in its most simple quantized moments seperated by luck. I find that I much prefer the fluid interactions of multiple players much more interesting and the challenge of separating skill and luck in those endlessly changing circumstances far more engrossing a field of study.

Rather than watch a spheroid move through the air pursued by largely similar athletes wearing different colored jerseys, try following chess. Some of the events broadcast online can get quite exciting to the aficionado (go to chessbomb.com), especially in a playoff tiebreaker involving blitz. But to appreciate the game you have to play it, so you can anticipate moves (turn off the chess engine!) which is not true with the spheroid games. And there's chess sabermetrics too, with opening theory (% of wins), individual stats and of course Elo.

You can learn to anticipate moves in either sports or chess by either playing or watching. And chess is more easily calculable than baseball, not less, so it doesn't sound like it would have what errorr is after.

Probably won't be long before the computerization of baseball gets to Las Vegas betting parlors and bettors will be able to put money down on each pitch, betting against computerized odds on balls, strikes, walks, hits, stolen bases, home runs, etc. according to the players involved, game situation etc. What do you suppose the odds would be against Prince Fielder getting a triple off Clayton Kershaw with a man on first and one out in the bottom of the third inning?

If it's in the playoffs, astronomically high.

Cheap computer power and even cheaper programmers have already changed Las Vegas. Heck, advanced math changed Vegas back in the 1970's when Edward Thorp started beating the house at blackjack. In the 90's gambling syndicates started attacking the sports books with the application of behavior science, cheap processing power and communications, as well as good old probability models. The bookies had to change as a consequence. That lead to the growth of teasers and bundling.

What I'm talking about is a table layout with all kinds of props on it that players make bets on during the actual course of the game, with a play on each pitch and at bat, kind of like super craps. On the first pitch of the game one could make a play on a ball, a strike, a ground out, fly out, single, double, etc. A put out by any particular defender, and so on. Imagine the excitement around the table with the home team trailing by 3 in the bottom of the ninth with the bases loaded, their biggest hitter up and the league's top closer on the mound. Chips would be all over the place. Maybe they'll get it together next season.

Would be a lot easier to get someone to ringtone outcome on an individual event of that sort. A ball on the second pitch of the fourth inning is unlikely to draw much suspicion.

Wow. "Rig." I should probably disable auto correct.

However, several hundred thousand dollars bet on a ball on the second pitch of the fourth inning would.

Thank God the sabermetricians were standing guard in the 1980s to 2000s to detect attempts to corrupt statistical norms through steroids. Remember how Bill James got an executive job with the Boston Red Sox and then immediately turned in Manny Ramirez and David Ortiz for being obvious juicers, costing the Red Sox the 2004 World Series? Remember all those hard hitting columns Nate Silver wrote about how sabermetrics demonstrated that Ramirez had to be cheating?


Remember how the Sabermetrics community rose up as one to defending the awarding of the 2001 AL MVP to Ichiro Suzuki on the grounds that the main rival Jason Giambi was clearly taking so many PEDs that he'd probably develop some kind of tumor?

@SS - not to mention the baseball slugger Babe Ruth who took amphetamines, a PED that no doubt raised his game. Heck I remember a guy in college who took some speed before an exam and aced the exam while the rest struggled. But, like athletes on PED, he was smart and had studied the material ahead of time, but supposedly the drugs made him sharper. PEDs in chess? I don't know of any but maybe they exist.

You are probably thinking of Hank Aaron.

> PEDs in chess? I don’t know of any but maybe they exist.

Adderall, for one. Probably Ambien too, for the nights before. There are rumors people use things like Beta Blockers in golf for calmness in putting, not sure how that would interact with Adderall.

In fact, "How many ELO is Adderall worth?" would make a fascinating study.

Way back when no one knew anything about anything a friend and I talked about your article while watching McGuire in Wrigley. Anyone who played sports like football in the 1980's knew enough about steroids to see what was happening. What we found puzzling was why it took so long for hitters to take advantage of what had been around for decades. Our conclusion was that baseball players were probably scoring a lot lower on standardized tests than other athletes.

Yeah, I wrote about that back in 2004 in AmCon: baseball is a traditionalist (i.e., kind of lazy sport). Honus Wagner lifted dumbbells in the 1900s and Babe Ruth hired a personal trainer and worked out all winter after his terrible 1925, but most ballplayers traditionally spent the winter primarily exercising their drinking muscles.


But, in retrospect, maybe it was less "lazy" than "sporting" that American ballplayers up into the 1980s didn't cut every corner the way Olympic athletes did.

It is a good question. To get the desired benefit from steroids, you have to learn a fair bit about biology and pharmacology. If you spend any time around body builders you quickly figure out that "bro-science" is quite sophisticated. If you spend time around baseball players, you quickly figure out why the Cartoon Network is on in every clubhouse. But, maybe my bias is showing here and it is more cultural.

The minor league experience does not encourage intellectual sophistication. I knew a guy who played baseball at Stanford, then had to ride the bus for a couple of seasons in the minors before making the big time. He found the minors just stultifying compared to being at Stanford. (Conversely, Billy Beane passed up a Stanford football scholarship to play in the minors and has regretted it ever since.)

> To get the desired benefit from steroids, you have to learn a fair bit about biology and pharmacology.

I think it's more common to know someone who has these skills than to have them yourself. Having a "trainer" who is a nurse is pretty common. Schwarzenegger trained with a medical doctor in the early days.

I've wondered how much of the volatility of athletic careers when people are traded stems from moving away from their sources. Eric Gagne comes to mind.

I don't know enough about baseball, but hockey was hit by stimulants before it was hit by steroids. The coke-fueled 80s were led by the Oilers playing fast, exciting, and offense-oriented games, while the 90s were characterized by the Devils - plodding players taking up space and implementing various trap schemes.

Good point about being traded and losing touch with your steroid connection. The rise of the Dominicans in baseball probably has something to do with steroids being sold in drug stores without prescription in the Dominican Republic, so Dominican players probably had more stable connections.

If we're talking about baseball in the 80s, what's the need for a "connection"? Steroids weren't scheduled until 1990.

We're mostly not talking about baseball in the 1980s. Canseco was more or less of an anomaly back then and he takes credit for personally introducing many of the big sluggers of the 1990s to PEDs.

I suspect that Canseco's trade from Oakland to Texas in the middle of the 1992 season (signed off on by George W. Bush, by the way) is when the steroid genie really started to get out of the bottle.

You might think this would be a good historical topic for sabermetricians to research, just like Bill James put a lot of research effort into understanding the demise of the Dead Ball era around 1919-21, but they've largely been averse to thinking about steroids, and get annoyed when anybody asks about it.

Or maybe colleges encourage information to flow between sports, leaving baseball out in the cold?

Right. The college track team was probably historically a main conduit. The Soviets started using steroids around the 1956 Olympics, so the American Olympians in the throwing events started using them pretty soon afterwards. It used to be common for football players to be on the track team in the spring (e.g., O.J. Simpson was on a world record sprint relay team at USC) as sprinters or throwers, so that could have been a conduit for steroid expertise from Olympic track & field to college football.

Did any prominent sabermetrician use the tools of his trade to dig into the steroids scandal while it was happening? I can't think of any. A friend of mine who is a player's agent told me around 1993 that Jose Canseco was "the Typhoid Mary of steroids," which Canseco's autobiography confirmed a dozen years later. And I could see that just looking at the stats of players who came into contact with Canseco.

Surely, Bill James was more plugged into the rumor mill than I was, so he must have known. But if James had used his statistical tools to expose this giant scandal, would he have ever been hired by the Boston Red Sox? When will sabermetricians ever stand up and admit they almost universally flunked the first great moral test of their era?

Well, even if we don't know who passed the great moral test, at least we know who administered it.

We can make up a list of those who flunked the test, starting with Bill James, Michael Lewis, and Nate Silver.

Old-fashioned sportswriters who deserve commendation include Thomas Boswell of the Washington Post who called out Canseco in 1988 and Rick Telander of the Chicago Sun-Times. But the celebrated Sabermetricians were almost uniformly (and, of course, intentionally) blind to the biggest story in baseball during their rise.

I stood up. When it mattered.

As mentioned above, I'm the author of *Wizardry: Baseball's All-Time Greatest Fielders Revealed*(OUP 2011). I initially developed the statistical model (DRA) disclosed in that book back in 2002-03, and shared it with Bill James, who replied to my email and thereby came to know me. When Allen Barra, who is Bill James' friend, was writing his *Brushbacks and Knockdowns: the Greatest Baseball Debates of the Last Two Centuries*, he saw an email debate I was having with James about whether Barry Bonds was on steroids. Barra liked my back-and-forth with James so much he reproduced it (in a re-ordered and heavily redacted form designed to make James look smarter) in the last chapter of the book.

In the original email exchange, my final argument to James was that Bonds was the only everyday regular player in major league history to experience his sustained peak in performance during the period of his career (age 35+) that James defined as "old" in his then-recent book *Win Shares*. James never responded to that final argument. That final unanswered argument of mine was excised from the chapter, though Barra picked it up and basically claimed it as his own.

By the way, James was extremely obnoxious during our email exchange. He likened me to "paranoid" people who tend to believe in "witchcraft." And as mentioned above, either he or Barra reordered and redacted my emails with James to edit out my best arguments and make James look smarter.

Thanks. I see the book here:


with some of your debate visible [picking up at p. 272]. Is it online in whole anywhere?

Bill James' behavior regarding Barry Bonds is just shameful:


I've been thinking that the point of Sabermetrics was to produce a way to identify juicers while maintaining the ability for the team to not know.

"Did any prominent sabermetrician use the tools of his trade to dig into the steroids scandal while it was happening?"

Yes, they in fact did, at least by the early 2000s, but without moralistic prejudging. No consensus was formed as to whether PEDs caused increased performance, and by corollary, no evaluation of a player's performance could reliably determine if that player was on PEDs or not.

Funny how all the high-powered Sabermetricians couldn't decide whether there was much PED cheating in the 2000s. I did a study of Olympic track & field performances for National Review in 1997 that showed clearly the ups and downs of steroids in that sport:


If Bill James had noticed steroids cheating in, say, the five years after McGwire-Sosa in 1998, would he have gotten his nice job with the Red Sox in 2003 and his three World Series rings?

This is an economics website: What were the economic incentives facing Bill James?

It's my understanding the the stats crowd is generally agnostic about the effects of PEDs on performance and, should there be any effect, whether this is an appropriate source of moral outrage. I take it you think differently. You might think they've retroactively adopted this stance because they all collectively turned a blind eye toward the phenomenon in the hopes of scoring lucrative big league gigs 20 years in the future, but this seems implausible. Considering his stature within the game circa 1989—that is, dismissed as an outsider and a crank whose views were regarded as contrarian and silly—Bill James would have to be playing a tremendous long con. Why would he thumb his nose at all of baseball's conventional wisdom except with regard to PEDs? How could he hope this would lead to a cushy front office gig? Has it lead to a cushy front office gig? How does his salary from consulting the Red Sox compare to his pre-Sox book sales?

Sabermetrics is about evidence-based analysis. I've read many in that crowd who have said that they lack sufficient evidence to make any conclusions about the prevalence or effect of PEDs. Isn't this consistent with their views? Why do you need to ascribe bad motives?

"Has it lead to a cushy front office gig?"

Yes. And three World Series rings for Bill James. He's a beloved success, vindicated. And who is so bumptious as to point out his moral failure on the biggest baseball statistics phenomenon since the end of the Dead Ball failure? Practically nobody.

Look, read my 1997 article about track above to see what a modestly sophisticated sophisticated statistical analysis could tease out about steroids in running. Next, consider the years 1998 to 2003 in baseball: McGwire-Sosa, Barry Bonds, etc.: a freak show. And where was Bill James? He was the three monkeys who see no evil, hear no evil, and speak no evil rolled into one. And his stonewalling on steroids did indeed help him get his cushy front office gig with the Red Sox in 2003, who promptly lifted the Babe's Curse the next year, led by two obvious roiders, Ramirez and Ortiz.

I know Bill James is a hero, but isn't this documented history at least more interesting than blind hero worship?

And it's not just James -- Nate Silver and a bunch of lesser celebrities are also implicated.

Hmm. Can't seem to reply to the Rey. Let's see where this appears...

You fail to grasp my criticism. Most of the stats crowd don't care about the steroids era, don't view it as a moral failure, and are dubious of its impact on the game. I don't know whether this is Bill James's personal view, but it is prevalent among FanGraphs, Baseball Prospectus, and so forth. Bully for those who we're blowing the whistle as their conscience dictated in the late 80s and mid-90s. But, as a community, the advanced stats crowd didn't turn a blind eye to steroids—they just didnt, and still don't, care. You might say not caring was a moral failure, but I just don't see what morality has to do with it. And insisting it is a moral failure isn't going to change anyone's mind. At some point, you are just repeating Monty Python's argument sketch.

And second, the idea that Bill James adopted a particular view on steroids so that he could get a consulting gig 20 years in the future is, to put it mildly, absurd. It suggests a level of Machiavellian foresight that no person outside of the Marvel Universe could possibly possess.

I get it, you were right and Bill James was wrong. And not only was he wrong, he was MORALLY wrong. And, what's worse, he was knowingly and purposefully morally wrong so that he could profit an unknown amount at an unknown point in the future. Yes, it is all so clear now.

Yes, the sabermetricians, both celebrities and foot soldiers, were not only wrong morally, they were wrong empirically. During the steroids decades, they were like pro wrestling fans who never came to an opinion on whether pro wrestling is rigged or not. And they considered themselves super-sophisticated pro wrestling fans!

Jeter's not a cheater !!!!

"…once the ball has been put into play, it doesn’t seem to matter all that much whether it was put into play against Roger Clemens or Roger Craig."

So is all the jockeying and changing of pitchers/batters for matchups throughout a game just so much froth?

Nevermind. I see "put into play" does not include home runs, strikeouts, walks, etc. Only when ball is hit into the field of play. I guess I still get the impression from announcers that they feel there are some pitchers that are better at throwing balls that when hit are more likely to become outs (some pitchers are better at throwing groundball outs for instance?)

Tommy John

He's 10th all time in hits given up but 49th all time in homers given up. He threw a sinker and gave up a lot of ground ball singles.

Interesting that you brought John up. With all the controversy over PEDs, why no thought about surgery? The "Tommy John" procedure is now common and done to players that could continue effectively without it. Hitters with normal vision undergo surgery to improve their eyesight. How far can these surgical modifications, which actually improve rather than restore an athlete's physical condition, go before somebody objects? Is it "fair" for an established star to undergo a costly operation that improves his performance when a minor leaguer can't afford the same? Why is dope bad but surgery good?

I kargy agree, except neither the star nor the minor leaguer are paying for their medical procedures. The team is.

I'm currently reading this book.

Comments for this post are closed