My Guide to Stats: Offense
Last week, I made a mistake on Twitter. That’s a pretty common sentence, I suspect. In this instance, I was talking about how I will almost certainly (and, I suspect, stupidly) buy the iPad 2 within the first couple of days, and I said that this is because I’m a “technology geek.” I meant this as self-mockery. I meant geek in the textbook definition of the word, geek being “a person with eccentric and unhealthy devotion to a particular interest.” The trouble is, geek has taken on new definitions in 2011 America. Best Buy has a Geek Squad. It is often said that the Geeks — Bill Gates and that Facebook Guy being the most obvious examples — are taking over the world. Computer geeks are viewed as the kinds of people you want as friends, or at least friends when your computer screen turns bright purple.
Geek has come to mean “somewhat socially inept but incredibly brilliant person when it comes to one subject.” Well, I’m not that kind of geek I don’t know squat about technology. I just like buying the overpriced latest thing. It is why my wife and I owned what I have to believe was the third or fourth DIVX machine ever built (or, certainly, one of three or four LAST DIVX machines ever built), it is why I have about 50 stupid and pointless gadgets stacked around my house, it is why the other day I made a specific run to the Verizon store so I could spend a half hour looking at the new XOOM tablet even though I ALREADY HAVE an iPad and ALREADY DECIDED I’m going to get the new one as soon as possible. I have an unhealthy obsession for buying new technology though I know absolutely nothing about it. There’s no word I know for “Dumb Geek.”*
I bring this up because I have been at spring training in Florida for a while, and I thought it would be a good time to explain again some of the sabermetric baseball terms that I use quite often in these blog posts and the baseball theories that I am fascinated by. But I need to first make it clear that I am not a sabermetrician. I’m not even an amateur sabermetrician. I know quite a few of these people, and I can tell you that my own efforts to add anything of any worth to the sabermetric community have been comically inept, and my own understanding of some of these sabermetric principles is pathetically simple and probably only about 40% right.
Mozart’s genius was that he could create the brilliant music.
Salieri’s genius was the he could hear the brilliance of the music.
I’d say that I enthusiastically but barely even know what Salieri’s talking about.
But here we are, and it’s baseball season, and I do write a lot about BABIP and WAR and John Dewan’s plus-minus, and OPS+, and I do often mock wins and RBIs and batting average, and while this doesn’t get me within three European countries of Cuttingedge, it’s all I’ve got. Just remember — like it would be possible for you to forget — I’m not a baseball geek. I’m like a dumb baseball geek.
* * *
What’s the matter with batting average?
I have to admit that it stuns me when I hear prominent baseball executives and scouts publicly quote a player’s batting average like it means everything. LIke they will say: “This guy hit .305 last year, so he obviously had a really good year.”
Look there’s a very good chance that if the guy hit .305 last year he had a really good year. But as Bill James has said, ranking someone by batting average is like being a movie critic who ranks movies after only watching the first two-thirds. Hal Morris hit .309 in 1998 and, though he remains one of my favorite people, I must say that he was almost useless. Felix Fermin hit .317 for Seattle in 411 plate appearances in 1994, and was out of baseball within two years. Juan Pierre hit .327 in 2001 and led the league in stolen bases and was a thoroughly unhelpful offensive player. We can go on and on.
The problems with batting average are so obvious that it seems kind of stunning that we have overlooked them for more than 100 years. It probably says something about how once we all get going in one certain direction, it’s hard to change course. I think we would all agree that the goal of a big league baseball team is to win games. On the offensive side, this revolves around scoring runs. On the defensive side (including pitching) this revolves around preventing runs. If you score more runs, you win. If you score fewer runs, you lose. This is baseball at its simplest.
So how does batting average tell you almost ANYTHING you really want to know?
I’ve made the point before about how batting average SEEMS simple, but it is really one of the most advanced stats we have if you consider “advanced” to mean “bizarrely complicated and obtuse.” WAR and xFIP have NOTHING on batting average.
How do we figure batting average? Well, start with a players’ number of plate appearances. That would be the number of times the player comes to the plate.
Now, subtract the walks. No, seriously, just subtract those. We don’t care about those.
Now, subtract the hit-by-pitches. Get rid of them.
Now, subtract the times that the player hit a fly ball that allowed a runner to tag up and score from third base.
Now, subtract the times the batter bunted a runner from first to second base, or second to third, or third to home but still made an out. Do not subtract the plate appearance if the batter successfully made it to first base. Do not subtract it if he hit a hard smash that accomplished PRECISELY THE SAME THING as a bunt. Do not subtract it if he hit a check-swing dribbler that was KIND OF like a bunt but did not seem from the press box to be a purposeful bunt.
Remember to include the times he reached base but only because of a defensive blunder.
OK, you have that number? We call those “at-bats.” Now, what you want to do it take the number of hits and divide those by at-bats. What is a hit? Any time someone hits a ball that allows him to reach base. No, we don’t care what base he reaches. Double … triple … home runs … they’re all just “hits” when it comes to batting average.
Of course, if the batter gets on base because of a defensive error, that doesn’t count as a hit. That counts as an out. Even though he didn’t make an out. How do we determine if the defensive player made an error? Someone in the press box we call the “official scorer” will watch the game and make the determination based on whatever he happens to be thinking at that moment.
OK, now you divide the hits by at-bats. And that is your hits percentage. We call it batting average even though it is not an average of anything. And the person with the highest average will be named the batting champion, even if we have to carry out the division to five or six or seven decimal points. The team with the highest batting averages will be listed on top of the charts even if they scored 200 runs less than another team.
It seems at least possible that there’s a better way
* * *
Why isn’t OPS+ instead called “Special Ops?”
The two most basic statistics that seem to best define hitting are on-base percentage and slugging percentage. I don’t think I need to explain them here, but I will. On base percentage is times on base divided by plate appearances. It is basically “the percentage of time the batter did not make an out.” It is not exactly that — there are a few quirks revolving around errors and sacrifice hits — but it’s pretty darned close. Of all of the basic offensive stats, OBP is probably the most important because, as has been said many times, baseball doesn’t have a clock. Outs are the clock. In football, you get 60 minutes to score as many points as you can. In the NBA, you get 48 minutes. In baseball, you get 27 outs. Every out is one more step to the end. A batter’s job is largely to not make outs, and on-base percentage measures that.
Slugging percentage is total bases divided by at-bats. It is a good measurement of how much power a player offers. If a player gets 187 hits in 623 at-bats, he’s a .300 hitter. If they are all singles, his slugging percentage is .300. If they are all home runs, his slugging percentage is 1.200. And his slugging percentage can be anywhere in between.
In 2008, Justin Morneau got 187 hits in 623 at-bats. That’s a .300 average.
In 1958, Nellie Fox got 187 hits in 623 at-bats. That was a .300 average then too. By batting average that was exactly the same offensive season.
Morneau though hit 47 doubles to Fox’s 21 doubles. Fox actually hit more triples, 6-4, but Morneau hit 23 home runs and Fox hit, um, zero. Justin Morneau had a .499 slugging percentage. Nellie Fox had a .353 slugging percentage.
Because those are the two basic stats that seem to tell us most about the players, there have been several efforts to mash them together. Bill James multiplied them and then multiplied that by plate appearances to come up with what he called “Runs Created,” which is still a great way to judge the raw offensive contributions of a player.
Last year’s Top 5 in runs created:
1. Joey Votto, 144
2. Albert Pujols, 142
3. Miguel Cabrera, 141
4. Jose Bautista, 139
5. Josh Hamilton, 134
The more famous effort to mash on-base percentage and slugging percentage is simply adding of them together, a sum which we have come to call OPS — (On-base percentage Plus Slugging percentage). The 2010 leaders in OPS are the same as the leaders in runs created, only in different order:
1. Josh Hamilton, 1.044
2. Miguel Cabrera, 1.042
3. Joey Votto, 1.024
4. Albert Pujols, 1.011
5. Jose Bautista, .995
There are several problems with OPS, one of them being that apparently you should never add together two fractions that have different denominators (on-base percentage works with plate appearances; slugging percentage works with at-bats); another is that on-base percentage is actually much more important when it comes to scoring runs than slugging percentage is but in OPS actually counts for less (because on-base percentages are usually smaller). But I think OPS, even with its flaws, is a pretty good way to measure offensive contribution, certainly better than batting average, and its become pretty popular, and if that’s our best shot to get out of the batting average dark ages then I am all for it.
Adjusted OPS+ is an offensive number I might quote more than any other — it is OPS adjusted to include context … specifically the park the player hit in and the time when he hit. OPS+ is a great stat, I think, a single number that tells you so much about what the player’s season really means.
In 1995, Andres Galarraga hit .280 with 31 homers and 106 RBIs.
In 1908, Ty Cobb hit .324 with 4 homers and 108 RBIs.
Galarraga had an OPS of .842 built largely on his .511 slugging percentage.
Cobb had an OPS of .842 built largely on his .367 on-base percentage.
Who had the better year? You will probably assume it was Cobb. You may even assume it’s not close. But OPS+ tells you — Galarraga didn’t even have a GOOD offensive year. He had a 97 OPS+ … 100 is average. He didn’t walk. His slugging percentage was largely a function of the offensive time when he played and the absurd Coors Field ballpark where he played.
Cobb meanwhile LED THE LEAGUE with a 169 OPS+. He, of course, played during deadball, when runs were at a premium. This was especially true in 1908, when Cobb led the league with a .475 slugging percentage, when only two other guys hit even .300, when only two guys scored even 100 runs and Cobb’s 108 RBIs led the league by TWENTY-EIGHT. The thing is most people do not know the history of baseball well enough to know that run scoring was ESPECIALLY bleak in 1908, and soon enough few will remember the insanity of the early days of Coors Field.
But if you put it like this …
Cobb in 1908: 169 OPS+ (led league)
Galarraga in 1995: 97 OPS+ (below average)
… you will know very quickly that there is no comparison between Cobb’s season and Galarraga’s season.
And, I don’t know why we don’t call it Special Ops. That would be awesome.
* * *
WPA? Is that a new deal?
One of the coolest stats out there is WPA, which stands for Win Probability Added, which is a name that I don’t think helps the cause much. There are certain words that scare the bejeebers out of people. Linear Weights were like that for me. I would see anything mashing those words together — “linear” and “weights” — and I would kind of freak out. For years, this prevented me from reading or thinking too much about the great work of Pete Palmer and others even though the concept of linear weights — giving values to various offensive things — is really not complicated at all.
Win Probability Added is not only an offensive stat, but I’m including it for offense … the concept is that at every point in a game, each team has a certain chance of winning. Take the Pittsburgh-Milwaukee game of July 20th last year. The game started and obviously both teams had exactly a 50% chance of winning.
Milwaukee did not score in the top of the first. At that point Pittsburgh’s chance of winning moved up to 55%, and Milwaukee’s dropped to 45%.
Pittsburgh promptly scored nine runs. Yeah, nine. Each of those runs obviously increased the Pirates chances of winning the game. For fun, here is a quick chart only of the runs:
— Pedro Alvarez grand slam (Pittsburgh 4-0)
Chances before the home run: 64%
Chances after the home run: 86%
— Lastings Milledge scores on error (Pittsburgh 5-0)
Chances before run: 88%
Chances after run: 91%
— Jose Tabata hits two-run double (Pittsburgh 7-0)
Chances before runs: 92%
Chances after runs: 96%
— Delwyn Young hits run-scoring double (Pittsburgh 8-0)
Chances before run: 96%
Chances after run: 98%
— Neil Walker hits run-scoring double (Pittsburgh 9-0)
Chances before run: 98%
Chances after run: 99%
In the top of the second, Milwaukee scored three runs. This moved their winning percentage up from one percent to 5%. Alvarez homered again moving Pittsburgh’s percentage from 95% to 97%. And so on. It turned out that this was a wild game and at one point Milwaukee cut the lead to 10-9 on a Ryan Braun homer — when Braun hit that homer, the Brewers winning percentage jumped from 14% to 30%.
This is a simple concept to understand when you only talk about scoring runs. Its quite easy to understand the math when you say that the Yankees up 2-1 in the eighth have a better chance of winning than the Red Sox down 2-1 in the eighth.
What gets a little bit tougher is to realize that EVERY PLAY increases or decreases a team’s chance to win the game. If the Red Sox leadoff hitter in the eighth draws a walk, the Red Sox chances go up. If that is followed up with a single, so that there are runners on first and third, Boston’s chances chances go up yet again. If Joba Chamberlain then strikes out two, the Red Sox chances go down. If a single scores the tying run, the chances go up. And so on. Every play, from the first to the last, changes the percentages, sometimes in an almost unnoticeable way (a one out groundout in the third) sometimes in earth shattering ways (a game-winning walk-off grand slam).
What WPA does is add up all the percentages. It doesn’t only do this for hitters — it does it for pitchers and fielders too. But for now, we focus on hitters. WPA simply adds up how much a hitter changes his teams chances to win. It adds up EVERYTHING. The clutch hits. The key strikeouts. And more, much more, the mundane at-bats that our minds simply cannot keep track of.
Here were the Top 10 in WPA in 2010 by Fangraphs:
1. Miguel Cabrera, 7.42
2. Joey Votto, 6.85
3. Josh Hamilton, 6.25
4. Albert Pujols, 5.38
5. Adrian Gonzalez, 5.11
6. Jason Heyward, 4.82
7. Shin-Soo Choo, 4.59
8. Matt Holliday, 4.10
9. Delmon Young, 4.06
10. Jose Bautista, 3.93
Tom Tango is quick to point out that WPA is not a great way to evaluate the TALENT of a player, but it’s a good way to evaluate HOW MUCH THAT PLAYER CONTRIBUTED during the year. That may sound odd, but it gets another point about fairly obvious point about offense that I should make here, a point about clutch hitting.
The baseball community has long celebrated players for their ability to lift their game when the chips are down, when the moment is bleak, when the game is on the line. And the sabermetric community has for a while now scoffed at the notion that players CAN consistently lift their games in the clutch moments. The baseball community builds its case on waves of emotion and selective memory. The sabermetric community builds its case on the fact that so far nothing has been found in the numbers to suggest that players, no matter how good, no matter how celebrated for their heroics, are capable of predictably and reliably being better in the biggest moments.
So statistically, if you want to judge the talent of a hitter, you would not use WPA — would not use a statistic that rates some at-bats as being much more important than other at-bats. But if you want to judge a player based on how much he contributed to the team, there are few stats better suited for that than WPA.
* * *
BABIP stands for “Batting Average on Balls In Play” and it’s a different kind of stat from the rest here. It doesn’t tell you much about how good player is. It migt tell you how hit lucky he has been … and how likely he is to improve or fall off in the future.
To figure BABIP, you take all the balls in play and subtract the home runs. Then you figure the batting average. It’s really simple. Last year, batters hit .297 on balls in play. The number stays right around there. The year before it was .299. The year before it was .300. The year before that it was .303.
So it’s always around .300. Players who hit a lot of line drives will have a higher BABIP, of course. Joe Mauer has a career .344 BABIP. But in general, BABIP can swing wildly from one season to the next, and a lot of it appears to be Crash Davis luck — hitting one extra flare a week, just one, a gork, a ground ball with eyes, a dying quail.
Last year, Josh Hamilton had an abnormally high .390 BABIP. The year before that it was .319. His line drive percentage was almost exactly the same. He popped out more. But he hit many more ground balls, and those ground balls went through, and that was a big contributor to his massive season.
Is that repeatable? There’s is a lot of dispute about that. Some think Hamilton is due for a big drop-off in 2011. Others think he will have a huge season. It’s just something to think about.
We will include one more stat because it is prominent in a stat I will come back to in the end, WAR. The stat wOBA looks scary because any word where you make the first letter lower case and the rest upper case is scary. It doesn’t matter how harmless or happy the word really is. Look:
wOBA stands for Weighted On-Base Average. And as they say over at Fangraphs, this is the statistic that realizes that every time you reach base, it’s worth SOMETHING.
Here is an approximation of what each thing is worth:
Non-intentional walk: .72
Hit by pitch: .75
Reached base on error: .92
Home run: 1.95
Funny, isn’t it, that reaching on error is worth just a touch more than a single, or that getting hit by a pitcher is worth a touch more than a non-intentional walk. I’ll have to look more closely at that. Anyway, you multiply all that out, divide by plate appearances and, voila, you have wOBA. An average wOBA should be about an average on-base percentage — .330 or so. Last year Josh Hamilton led the American League with a .447 wOBA. Joey Votto led the National League with a .439 wOBA.
How did they get to these numbers. If you are really interested, you can read this and then look around the Internet. But the larger point is that these weighted numbers do a pretty amazing job of estimating runs scored. And it’s worth remembering one more time that scoring runs is, in fact, the goal of the team at the plate.
OK, I have no idea if I will ever have the strength to do part two, but if I do it will be on pitching.