Quote:
Originally Posted by Timofmars
[Also, I have no idea how accurately OOTP models these stats in particular. They may do a good job on making hits and HRs appear in realistic amounts, but who knows what they used to figure out how often baserunners advance on an out.
|
I have tested the OOTP18 scoring model using the 2002 Runs Created formula (see wiki
https://en.wikipedia.org/wiki/Runs_created.) OOTP18 came within 1.3% of the 2002 RC formula, so I would infer that it is modeling the correct behavior. (Keep in mind there is a separate caculation done elsewhere in-game that shows the leagues actual versus historical calculations as well.) The RC model is itself simply a theoretical model of real baseball scoring using all the statistics that are available to statisticians.
A difference within 1.3% means the OOTP model is somehow taking into account the general effect of the hits, stolen bases, advances on outs, etc. in a manner that leads to a similar result that "real" baseball generates. This doesn't prove it's working as real baseball, but it's a hurdle it has to cross in order to make that claim. It was successful. Without actually seeing the code, I would not be surprised if the modeling engine is working realistically. Again, this isn't proof, but it's a strong indicator. Mishandling the effect of outs in run generation would be relatively obvious to pick up in a test like this since outs are the most numerous event in baseball.