Ok. Skimmed the thread.
First reaction - that coach is an idiot.
Second reaction - .... but maybe if I...
Third reaction - Odd daydream combination of Hari Seldon and a World Series parade.
So, for the sake of argument, let's pretend it is possible to predict the score of a single game.
In effect, that means predicting the number of runs for both teams.
I'm sort of thinking that it theoretically could be possible to cobble Team A's offense numbers and Team B's pitching numbers into a run prediction for Team A.
C1*wOBA + C2*OppWHIP = Runs + Error
Where C1 and C2 are constants derived from working on previous seasons of data and OppWHIP represents the WHIP of the opposing pitcher(s).
Here's the problem - it would take a lot of data to calculate those constants with any degree of remote accuracy, even if such a thing were possible (and I don't think it is).