Q. Can you make any sense of spring training statistics?
A. Different stats have different levels of "noise" in them. Ichiro's career batting average has noise in it, but it's about 2%, not counting park context. Ichiro's career range factor has noise in it, too, but the noise is something more like (?) 30%.
Baseball sabermetricians are really cruddy at estimating the AMOUNT OF noise in a particular statistic. A sense of proportion is almost completely lacking in the industry. Sabermigos are great at math; they're not so great at common sense.
We mean it in a good way.
Q. How much noise is in there in spring training stats?
A. Oh ... what would it be? 60%? Something like that.
One of the few sabermigos who does have a sense of proportion: Jeff Sullivan. Read this article by him. He points out various [huge] sources of noise in ST stats.
It is very, very challenging to make sense of a phenomenon when your metrics are diluted with 60% noise. When you've got 60%, 70% noise, and you run a "study" (sic!) asking about the value of the stats, your study is going to suggest NO value to the metric unless you know how to design it.
But I'll guarantee you that there is some "study" that will show the phenomenon being reflected at some level of consistency. Give me all of the home run leaders in Arizona, past twenty years -- even first-week leaders -- and I'll guarantee that their presence is reflected in MLB regular-season slugging leaderboards, at some level of confidence. It might be only a .20 correlation, but there will be a correlation. Which is the same thing as saying that ST stats are not worth zero.
Q. SSI thinks that spring performance is worth something? How much?
A. Not much, usually. In a few cases it becomes critical.
Kendrys Morales might be just tinkering. Felix' results are unimportant, unless he walks a guy per inning.
Jason Bay vs Casper Wells is going to be important; both guys are playing for a job and both guys are out there competing. Jon Garland and Jeremy Bonderman are going to be competing.
The 2nd-worst thing you can do with such metrics is to assume that the noise is small - and to treat the metrics as though they are reliable with no further thought. Many casual fans do this.
I doubt any SSI or LL or USSM readers do this, and I hate to see the kind of condescension that assumes a web-literate fan to be unaware of the noise in ST stats.
The VERY worst thing you can do is to assume that since you don't like a metric, and you don't know how to edit out the noise, and you don't know how to design a study in which correlations are found --- > then the metric must be worth ZERO.
ST stats aren't worth MUCH, in BULK. But to imagine that every player's ST performance is irrelevant is incredibly naive. If Jeremy Bonderman walks more guys than he strikes out, he's going home. If he fans 20 guys and walks 2, then with the time off and re-boot, that means that he's probably going to have a good year by his own standards.
Q. When are MLB teams going to wake up and ...
A. Hold on right there, Turbo. Just. Stop.
Consider this statement, from an author who higher on the very same web page had strictly forbidden any and all "appeals to authority":
Or, you know, you could use a projection based on more than just the last season’s data point. Which is what every good organization in baseball does.
I'll tell you something else
every good organization in baseball every organization in baseball does. Consider. Spring. Training. Performance.
We might slow down, draw a breath, and ask a simple question.
Why does every MLB organization watch ST performance carefully? Why do all "good" MLB organizations (at least occasionally) make roster decisions based on spring results?
Org's don't just watch tools and skills. They also watch performance and results. All of them do. They understand about noise, understand with 100% crystal clarity. They're out there watching, aware that ST very often misleads you, and still factoring ST performance as one variable in the equation.
Why would that be?