Summing up RBI

Keith Law in a chat today on ESPN:

Todd (MA): Perhaps the RBI stat is overvalued, but it's hardly meaningless. The team that scores the most runs wins the game, and the RBI stat helps to identify what players succeed at tallying runs for their team. The ability to hit with runners on base is arguably one of the most important abilities in baseball, perhaps second only to proficiency in getting on base.

Keith Law: All it does is tell us who happened to come up to the plate with men on base. This is not valuable information. And there is no evidence that "the ability to hit with runners on base" is at all distinct from "the ability to hit."

It's that last sentence that basically sums it up. Every single time the issue is really studied, one of two conclusions is made:

1. There is no ability to hit with men on base that is in any way distinct from simply the ability to hit.
2. There is an ability to hit with men on base, but its discernible impact is so small, that you're better off just proceeding as if #1 were true anyway.

I'm not sure if it can be summed up any better than Keith did.

But who will pitch the eighth inning?!?!

Over at the RLYW, SG notes that since May 29th, seven relievers in the New York Yankees bullpen have combined for 128 innings of 2.53 RA baseball. May 29th was the day that the New York Yankees moved Joba Chamberlain out of the bullpen and into the starting rotation.

It's only been a couple of months, but already Joba, with an assist from the baseball gods, has shown people why it was absolute lunacy to consider keeping him in the bullpen. Not only has he maintained his dominant stuff throughout his starts, but the bullpen hasn't really missed him. The team hasn't been hemoragging late inning leads. In fact, as far as my admittedly spotty memory can recall, they haven't blown one game that they might have won if Joba had been available in the bullpen (confirmation bias warning).

That's obviously not going to continue, but the larger point remains: it's just not that hard to pitch one inning of baseball. The difference between Joba and some random reliever pitching the eighth inning is dwarfed by the difference beween Joba and some random starter pitching the first seven. Case closed.

The Uncanny Valley

In robotics, there exists a hypothesis known as "the uncanny valley." In short, the hypothesis is that humans respond with increasing empathy to robots (and other objects) as they become increasingly human up until a point where they are almost human. At this point, humans will instead experience strong repulsion as they now perceive those characteristics that make them un-human instead of those that make them more human. However, this repulsion only lasts for a little while because as the robots become even more human we once again begin to identify strongly with them.

This hypothesis tends to ring true to me. Just look at the reaction that most people have to computer animated characters. We tend to obsess over the things that make them less human. We find them creepy, even if technically they are superbly animated. We don't have that reaction to simple hand-drawn animations. Indeed, much animation is designed to create truly unrealistic characteristics that are designed to elicit a truly empathetic response.

This, of course, has nothing at all to do with baseball. Not obviously, anyway.

I began thinking about this concept with respect to baseball as it pertains to baseball analysis. All baseball analysis is designed to simplify our view of baseball so that we can more easily extract information from it. As statistics become more and more complex, they begin to capture real baseball more and more accurately.

But will we reach a point where the models involved will be so minute that they miss the big picture? Can we create models that are too detailed? Will baseball models reach a point where they are so close to real baseball that they actually cease to tell us anything useful, instead providing us with either totally obvious or totally false conclusions?

One of the advantages of statistics is their coarse-grained nature. By not being bogged down in details, we can cut right to some important if general truths.

Obviously, I don't have an answer to these questions. And really, I probably don't need one. As long as we can measure the efficacy of the models we create, we should be able to discern between useful and useless models. Nonetheless, it's a question that's been eating at me for a few weeks. Thoughts?

A realization and another piggyback

I keep meaning to do something worthwhile in this space, to explore some area of baseball research that I'm curious about or to develop some new way of thinking about a problem. In reality though, that stuff is far more involved than I'm usually willing to commit to, so I end up piggybacking off of other people's blogs, which essentially makes me like every other blog on the Internet (not that there's anything wrong with that).

Today's piggyback is from Rob Neyer, who is unfortunately behind the ESPN Insider curtain:

The "anti-scout stat geek"? He's a straw man. Doesn't really exist, at least not in any meaningful numbers. As for scouts doing "a better job of projecting prospects than the numbers do" … Well, we don't really know that, do we? When scouts evaluate players, even Class A players, they can't help but notice the numbers. And when number geeks evaluate players, they read the scouting reports.

This is a marvelous point and it points out a flaw on both the objective and subjective realms of baseball analysis.

Not only are scouts, the subjective analysts, not impervious to the influence of numbers (after all, they have to use some data to back up a projection, even if it's height, weight, age, bat speed, fastball velocity, or whatever), but they are not forced to quantify their projections into any meaningful set of numbers. Thus, unlike purely objective projection mechanisms, it's really hard to evaluate how well they've done at projection. If I were paying scouts, I'd have them rigorously quantify their projections along the lines of providing basic stat lines at various percentiles for important ages at the major league level. Then you would be able to actually grade how well your scouts do project and give them valuable feedback.

Of course, Rob exposes a problem on the objective side as well. Objective analysts too often fool themselves into adding their own subjective point of view to a supposedly objective evaluation. It drives me crazy every time I see Nate Silver hedge his PECOTA projections. Which is it, Nate? Are your PECOTA projections "deadly accurate" or do they need subjective adjustment? It completely defeats the purpose of objective analysis when the analyst starts adding his own view point.*

* Yes, I realize that BPro articles are for entertainment first, and they do entertain me, but I still reserve the right to get annoyed. I'm sure Mr. Silver understands my point without needing a correction from yours truly.

Twelve hits

I'm gonna cherry pick from the boys at River Ave. Blues again, mainly because they buried this gem in an analysis of Joba Chamberlain's brief first start.

Now, we know Cano is struggling, and we know he’s swinging at everything. But here’s the reality: At this point last year, Cano’s batting average was .050 higher than it is now, and the second baseman had just 12 hits more than he did now in the same number of at bats. 12! That’s hardly anything.

This really underscores just how much small samples distort reality. If you scatter just a dozen singles into Robinson Cano's line, he's back to where he was at this point last year. Robby can make that up with a couple of hot weeks.

This is the ultimate reality of baseball: the season just is not long enough for all the breaks, bounces, injuries, good luck, bad luck, and blown calls to even out. The difference between a .300 hitter and a .280 hitter over the course of a year has more to do with chance than it has to do with skill. That's what makes baseball fun, but it's also what makes it frustrating.

Why stats are important

Because I just watched LaTroy @!&*#%!*%#!&^$!(*@#! Hawkins blow a one run lead in the bottom of the 11th inning and statistics are the only thing capable of providing me with an objective point of view to prevent one excruciatingly painful moment from permanently warping my perspective.

Even LaTroy @!&*#%!*%#!&^$!(*@#! Hawkins was likely to convert that save. He didn't. It happens.

*(^#$*&#%*^#$@*#^!

Why don't we see complete games anymore?

Craig Calcaterra over at Shysterball has a brief post up about why we see fewer complete games in the present than we did in the past. His thesis, and that of the gentleman to whom he links, is that because teams have more money invested in young pitchers now than they did in the past, they are more careful with their arms. They don't push their arms to their limits for fear of losing their investment entirely.

I think this is mostly true, but I want to make a finer point: it doesn't really matter how much you paid for the services of a particular pitcher. All that matters is how to extract the best value from that pitcher, or rather, from your system of training pitchers.

Once you've signed a young pitcher, the money that you have paid to him is a sunk cost. You cannot get it back. Thus, the money that you have already paid to pitchers should not factor at all into the training and usage of your pitchers. If the best way to extract performance from your pitchers is to have them all throw 300 innings immediately upon leaving high school and see which ones survive, then this will be the best way no matter how much you paid them. Thus, it would not be correct to alter your system of pitcher training and usage just because your pitchers cost you more money than they did in the past.

However, there is a cost to determining which system of pitcher training and usage is the most efficient. In the past, it is entirely possible that this cost outweighed the cost of paying young pitchers. Thus, teams simply hired as many young arms as they could, worked the hell out of them, and thus found which ones would stick. This may have been cheaper than actually determining how to maximize that value of any given set of pitchers. I think that this is highly plausible when you consider the relatively high cost of data analysis (no computers) and the relative inexpensivenesses of pitching.

Now that it costs a lot more to hire a pitcher (and a lot less to analyze data), the rewards for having an efficient system are much higher. Thus, teams are more willing to invest in research to determine which systems of pitcher training and usage are the most efficient. In the process of doing so, they have apparently determined that it is better for your pitchers to not ask them to throw twenty complete games every year.

Whether or not this is the correct conclusion is a completely different question.