The Dord of Darien

Musings from the Mayor of the Internet

This is why you’re a former GM

You remember Jim Bowden. He’s the former Nationals GM who is most famous for getting ridiculously roaring glass-eyed drunk, signing Cristian Guzman to a $34 million contract, and then driving really fast. Well, now he’s blogging at ESPN, and he’s helping us the home viewers to understand the logical processes that helped him transition from a career as a general manager in Major League Baseball to a career as a blogger at ESPN. Let’s see if we can catch him while he’s still sober enough to type.

Simple stats to evaluate teams, players

I think a bit of definition of terms is in order here:

1: free from guile : innocent
2a : free from vanity : modest
b : free from ostentation or display <a simple outfit>
3: of humble origin or modest position <a simple farmer>
4a : lacking in knowledge or expertise <a simple amateur of the arts>
b (1) : stupid (2) : mentally retarded
c : not socially or culturally sophisticated : naive; also : credulous

I’d like to thank Merriam-Webster for calling Jim Bowden names so I don’t have to.

I get some flack for this from time to time, so, as Ubaldo Jimenez would say, let me be clear: if you want to head out to the ballpark with your buddies and have a few beers and race-bait Milton Bradley and you don’t give a great goddamn what his WAR is or how his BABIP might be impacting his WPA, that’s cool with me. I find that understanding how the game of baseball works improves my enjoyment of it; if you don’t care, I have no beef with that. There’s room for both of us to do our things.

However. If one happens to be a former general manager of a baseball team — and here I’m talking about a real professional team, and not like in MLB Front Office Manager — and one is currently employed by an outfit that declares itself the "worldwide leader in sports" to write about one’s experience generally managing a professional baseball team, and if one’s article includes this:

Baseball also has a simple to side to it. I get asked all the time which two or three common statistics I would pick to evaluate a team or players. My quick answer would be the following:

1. For a team: Run differential
2. For a hitter: OPS + RBIs, or OPSBIs
3. For a pitcher: ERA, WHIP, SO

Then one should expect to have the shit ridiculed out of onesself on my blog.

Now, before I get accused of unfairness, I’ll admit that Bowden starts out pretty strong. Run differential is pretty much the right stat to use to do a quick-and-dirty evaluation of a team. Though he does crap it up in the calculations by doing it as Runs Scored – Earned Runs Allowed. What the eff, Jimbo? Do unearned runs not go up on the scoreboard these days? I know OBP doesn’t. The best thing about the way Bowden calculates run diff is that it will result in the league having a positive run differential against itself, which is one of those existential paradoxes that only Commander Data can solve.

So, yeah, that’s the good part. Let’s move right along into the next bit, where Bowden says:

2. For a hitter: OPS + RBIs, or OPSBIs

I had to quote that again so you could absorb the full impact of the ridiculous junk stat Bowden has just told you he relied on when making decisions as a GM. What the shit? Seriously, Jim Bowden. My goodness.

Okay, I’m coherent again. Let’s step through this. OPS + RBIs, huh? "OPSBIs" is a horrible neologism that doesn’t even parse properly; "on-base plus slugging batted in" indeed. But is it any good as a statistic? Pretend I haven’t already called it junk, and we’ll find out!

Well, it starts with OPS. OPS is not a very good statistic. I mean, it’s not very bad — if you’re just looking for a quick-and-dirty way to tell if a player sucks, it fits the bill. But it has the non-trivial flaw of weighting OBP and SLG equally, and OBP is way more important. If you’re a GM, and you use OPS to evaluate players, you’re liable to do things like give Juan Uribe 3/$21M, not realising that his career OBP of .299 will absolutely murder your offense. And only a fool would do that!

RBIs, unlike OPS, is a very bad statistic. Why? Well, you tell me. Which one of these players is the best?

Player X: 50 RBIs
Player Y: 44 RBIs
Player Z: 30 RBIs
Player F: 10 RBIs

Clearly it’s Player X, right? Good, we agree. Now let’s flesh them out a bit more:

Player X: 50 RBIs, 293 PA, .262 / .314 / .461, 109 OPS+, 1.4 oWAR. Had a total of 203 baserunners while at the plate.

Player Y: 44 RBIs, 278 PA, .332 / .486 / .678, 218 OPS+, 4.5 oWAR. Had a total of 176 baserunners while at the plate.

Player Z: 30 RBIs, 226 PA, .315 / .389 / .523, 133 OPS+, 1.2 oWAR. Had a total of 139 baserunners while at the plate.

Player F: 10 RBIs, 225 PA, .301 / .409 / .419, 124 OPS+, 1.2 oWAR. Had a total of 93 baserunners while at the plate.

Oh. So actually Player X is kind of shitty — he’s just had a ton of PA and a lot of inherited runners. Player Z has been almost as valuable to his team, and only has 77% as many PA, while Player F never ever inherits any runners to drive in. And Player Y is the best player in baseball. This is why RBIs are crap. They are too team-dependent.

So what happens when you add RBIs and OPS together, then? You get shit soup is what happens. So we’re not here all day, I’ll pass on another histrionic demonstration and just mention that the league average OPS is .710, while the league average RBIs is, who knows, something like 25. So, if you’ll excuse the presumption, Mr. Bowden: what the fuck is the fucking point? Why not just use OPS? He even makes this point for me, though he seems to be too drunk to figure it out, when he adds the ridiculous chart. Is it sorted by OPS or by "OPSBIs?" you tell me — the order’s the same either way.

But wait! Bowden isn’t done being dense about this:

The key for me is breaking down the RBIs to differentiate the ones that came against pitchers who throw with the most velocity, change in velocity, late-breaking action, change in planes or deceptive deliveries.

Which you didn’t do. You just, like, added RBIs — a counting stat — to OPS — a rate stat.

How often do the RBIs occur in one-run games against Mariano Rivera in the ninth inning as opposed to a 12-0 blowout in the fourth inning against a mopup reliever. It is essential to break this down in detail.

But maybe not so essential that he actually did it. Because he didn’t. Also not essential: caring about how many runners a hitter inherited.

Statistics can be misleading unless you blend them with the scouting aspect of baseball.

Statistics are more likely to be misleading if they’re garbage to begin with.

When you get to October, hitters need to hit Jon Lester, CC Sabathia, Roy Halladay or Tim Lincecum, so you need to have hitters who can hit the best pitchers in the game rather than hitters who put up good numbers against mediocre or below-average pitching.

Do you have any evidence for the existence of this fabled hitter who absolutely murders R.A. Dickey and Randy Wolf but is awful against Lincecum and Halladay? Of course you don’t, because there’s no such hitter. You might as well say that what’s important is getting hitters who can teleport so they don’t make outs on the basepaths.

Look. I get the emotional appeal of "clutch." I remember when Sammy Sosa hit a ninth-inning, two-out, two-strike pitch out into Waveland Avenue in game 1 of the 2003 NLCS to tie it up at the last possible moment. That was awesome. But "clutch hitting" simply is not a repeatable phenomenon. It is not a skill. It is a product of luck. David Ortiz is probably the most famous clutch hitter ever, right? Go here. Look at his WPA. See how it’s up-and-down? Now look at his "Clutch." Positive, negative, positive, negative. No year-over-year correlation at all. This is because it is mostly luck. You cannot get players who will be "clutch" in the future by going after players who were "clutch" in the past, or by any other means besides plain, ordinary luck.

The fact that you don’t understand this — and that you choose to define "clutch" as "RBIs in a blindingly arbitrary set of situations" — is terribly, terribly funny.

Still, OPS + RBIs gives me a general feel for the level of player.

Only about 7% worse than OPS by itself, even.

So what remedy do I suggest? What stat is best for evaluating a hitter at a glance? Well, Bowden makes a big stink about "simple" stats, so I’m assuming I can’t use something mildly esoteric like oWAR. In that case, I’d go with OPS+ — it’s like OPS, but it’s normalised for park effect and differences among eras, and the context is built-in, so you don’t have to try to guess at what the average is (it’s always 100).

(Of course, with leadoff hitters, OBP+R+SB would be a better barometer, so you have to know the type of hitter your looking at).

I… or, yeah, I guess you could do that. Just a quick question: since OBP already correlates to runs better than any other basic stat, is it really helpful to add runs? I mean, even assuming that adding a counting stat to a rate stat made sense to begin with. Which, by the way, it doesn’t. And are you sure you don’t want to control for caught stealing at all? Don’t even want to throw in a -CS for the sake of form, as though the 8 or 11 or whatever will be discernable against the 350 or so points of OBP you’re randomly adding it to?

Also, sorry, but: [sic]

Bowden doesn’t really say much about pitching. He says this:

3. For a pitcher: ERA, WHIP, SO

And he gives leaderboards for those three stats, but he doesn’t do anything as comical as create a whole new garbage statistic and then write three paragraphs making excuses for it. All he says is:

These are my three favorite quick-look pitching statistics. Combined, they give me a snapshot of their abilities and talent, although statistics are best analyzed in concert with video and scouting reports in a complex system.

Now, for being so short, it’s amazing how wrong that is. Why?

Well, ERA don’t tell you much about a player’s ability or talent, because it’s polluted by external influences and bullshit judged outcomes, and it’s not even adjusted for park factor, for fuck’s sake. WHIP is a bit better, but it still gets fouled by defense and judgment calls. As for strikeouts (which is what Bowden means by SO, even though most people — though, notably, not baseball-reference — use K for that and SO for shutouts): just, like, raw strikeouts? Not even going to consider the number of batters faced?

If you had to pick three simple stats to get a good idea of a pitcher’s true level of ability, the correct stats to choose are K/9, BB/9, and HR/9. These stats correlate well with themselves over time, have been shown to be a better predictor of future ERA than ERA itself is, and have been shown to be the only things a pitcher actually controls anyhow. Pretty much all advanced pitching statistics are built on those three.

Not that I’m saying scouting isn’t valuable, but, come on. perhaps you wouldn’t find statistics so "misleading" if you were bright enough to figure out which ones are worth using. And perhaps you wouldn’t be a professional blogger at this point in your life.

1: a new word, usage, or expression
2: a meaningless word coined by a psychotic

No points for guessing.

June 17th, 2011 Posted by | Baseball | one comment