I have one criticism of the author’s analysis. In many examples where stats are provided, the analysis is not robust. He does not consider how two samples he is comparing should be different in a vacuum. For example, the author defines a group of women’s servers as “bad”, and the group holds serve 56.6% of the time. When a bad server is serving to stay in a set (e.g., down 4-5) the serve hold rate drops to 54.9%. The author concludes that laying a bad server in such a situation may be advantageous. A better analysis would have estimated the bad server hold rate in that situation (given the player is down a game, that player on average is slightly worse than the opponent, and would be expected to have a slightly lower service hold rate for all games in the match, and not just that particular situation). This omission of conditional groups occurs throughout the entire book.