tag:blogger.com,1999:blog-2418171424349595525.post6818248835659218116..comments2010-10-19T12:04:12.440-07:00Comments on Basebology (The Study of Baseball): The Uncanny ValleyUnknownnoreply@blogger.comBlogger3125tag:blogger.com,1999:blog-2418171424349595525.post-9024837798352899072008-07-23T05:42:00.000-07:002008-07-23T05:42:00.000-07:00What I'm saying is that right now the statistical ...What I'm saying is that right now the statistical models that we use are really coarse-grained. They deal with large chunks of information. Someone's OBP is an aggregate of a whole ton of plate appearances. It is useful precisely because it doesn't seek to capture a whole bunch of minute details.<BR/><BR/>I'm wondering if we will reach a point where our models are trying to capture so much minute detail that, despite approaching real baseball even more than our current models, they will tend to lead to gross errors when used to make conclusions about the minute details, precisely those areas that the more detailed model is supposed to succeed over the coarse model. If the model can't help us in the fine-grained areas, what good is it? We've already dealt well with the coarse-grained areas quite well.<BR/><BR/>Note that if this truly were an uncanny valley these models would NOT be so close to baseball as to be indistinguishable. Those models would exist on the other side of the valley.John Lynchhttps://www.blogger.com/profile/09331641225948161347noreply@blogger.comtag:blogger.com,1999:blog-2418171424349595525.post-12912113304517088522008-07-22T11:01:00.000-07:002008-07-22T11:01:00.000-07:00Or the idea is that because we've gotten so close ...Or the idea is that because we've gotten so close to real baseball, like you said, where the model is the same we'll only learn obvious stuff and where the model is different we won't learn anything at all?<BR/><BR/>I think I'm starting to get it. You're wondering if baseball models will ever reach the point where they become super complex and cease to actually tell us anything right?Robert Lynchhttps://www.blogger.com/profile/10268815170382205995noreply@blogger.comtag:blogger.com,1999:blog-2418171424349595525.post-30082454738581659262008-07-22T07:54:00.000-07:002008-07-22T07:54:00.000-07:00So you're saying our models may get so good that t...So you're saying our models may get so good that they replicate baseball so well that there isn't really any benefit of actually having the model anymore?Robert Lynchhttps://www.blogger.com/profile/10268815170382205995noreply@blogger.com