quote:Whatever you think is important, that's fine. Put it in your own little WAR-like system of evaluation, use whatever stats you want, but three things: (1) be consistent, (2) be ready to defend your decision to include some things and not others, and (3) don't work backwards from the result you want; start with the things you think are important and work TOWARDS the result they lead you to and stand by the results. Don't tweak the system because it spits out a name you don't like.
This is not meant to be critical:
Agreed with point 1. Point 2 is not necessary, you don't need to defend if you can predict who is going to win a game with some degree of accuracy, to the past and to the future. People will COME to you, instead, for advice. For example if you were Tom House, people COME to you for advice, he does not need to argue which throwing mechanics is the best. Point 3, no sure what you mean, but in order to fit the model, you need past data to fit the model but you need have a solid criteria how you fit your data. A past win or a past lost is the most definite criteria. Whether you like Babe Ruth better than Shoeless Joe, that is not a good criteria. If you can predict the past, let say 65% accuracy with 99% confidence level, then you could predict the future. Time is just another variable in the equation, just remember to be consistent as you advised when applying the rules for the past and for the future.