Combining players statistics

[Mulkis]

Hi All, I'm currently working on some betting system for tennis. I have got a lot reliable statistic for ATP. It allows me to calculate very precisely each player's point won/lost on serve/return, surface, etc. For example my data gives me exact % of points player wins on serve and return. I am going to feed all this data to Markov chain model later. But for the moment I cant made my mind how most efficiently and "scientifically" to combine 2 players' statistic playing h2h and come out with combined serve/return percentages. Any suggestions will be most appreciated. Thank you!

[Datapunter]

Re: Combining players statistics Serve % is an absolute measure, i.e. it does not depend on anything else, the serve either goes in or it doesn't, it totally in the players hands. The return % however is not. Player X will have a good return % against weak players and a bad one against good players. How do you deal with the fact that return % is dependent on the opponent ?