Devizier said:
Generally speaking, crappy stats are more predictive of crappy outcomes than they are of great outcomes.
Now how much you can predict from a 100-PA sample is another question. Maybe if you're Eric Van...
You've got the principle right, but you are being overly broad in using it in this instance.
I do not think anyone here would deny that if you had a 100 PA, or 200 PA sample of the first at bats of a player in MLB, the players with better stats are probably going to have ended up being better players, on average, than those with worse stats. That's fairly self-evident once you realize that there is data there and, at some point, the law of large numbers kicks in. What that does not tell us is that the first 100 or 200 PA sample of a young player's MLB career is at all meaningful if we are trying to predict their individual outcomes. Because, it might be that the difference between a great start and a horrible start, once you bring it back to the individual level, is pretty much meaningless.
Once accepting that, there are a couple things worth noting. The first relates to your flippant EV reference. What EV generally argued, in regard to small sample sizes, was that small samples of aberrantly strong performance were meaningful in suggesting that a player was better than previously thought. That is not to say that such a player is as good as that performance, but that a really strong performance, even over a small sample, can raise a baseline. This is especially true for prospects. However, a really bad performance over a small sample is far less informative. The reason for this is simple. We know that all baseball players are really bad for stretches and that these stretches are not all that uncommon. We, similarly, know that there are certain levels of performance sustained, even over a small sample, that a player of a given talent level is incredibly unlikely to achieve. This latter point is why it's now a legitimate debate whether Iglesias is a better hitter than was previously thought of just got incredibly lucky for a while. We know he got incredibly lucky, but, given how good he was, it's reasonable to argue that some of the change in performance was skill rather than luck.
That point applies directly to prospects that struggle for a stretch upon promotion. Any baseball player who struggles over a small sample isn't providing meaningful data with the possible exception of a player you think is the next Mike Trout. Basically, the distinction I acknowledged in my first paragraph stems entirely from the fact that the players who have good starts are providing meaningful data on the margins, not that those who are struggling are. But, it goes further than this. Players adjust to level jumps at different rates and in different manners. It is fairly common for players to struggle after a level change, get their feet under them, and go back to being the player they looked like prior to that jump. So, even more so than how ordinarily a small sample struggle is pretty much meaningless, it is especially true for a player playing at a new level for the first time.
