I had my brother write a very similar program before I knew this one existed, one way we tested our results was by using the stats we calculated to predict what the results of matches would be.
In our program offensive score had some predictive power but defense had no relation to reality.
Did you find differently when you tested your results?
[our graphs if you want to analyze them yourself]
http://kamocat.com/scouting/moredata.html