I have been in the process of looking at this url here - https://www.betfair.com.au/hub/news/tennis/tennis-modelling-point-based-models/ to look into a model of finding out the probability of a player winning a tennis match.
The thing I don't understand though is that the formula to me doesn't make any sense.
[fi - fav] = the players probability of winning a point on serve minus the tennis average
[gi - gav] = the players probability of winning a point on return of serve minus the tennis average.
This I all understand, however when I do the calculations, the numbers don't come up the same as in article.
For instance fi (70.44% for Djokovic) - fav (64%) = 6.44
gi (44.6% for Djkovic) - gav (36%) = 8.6
Now for Federer
fi (72.3 for Federer) - fav (64%) = 8.3
gi (41.1 for Federer) - gav (36%) = 5.1
Now that we have those calculations I now do the entire formula.
ft which is (67.2%) + (fi - fav) - (gi - gav) = 70.4 for Federer
ft which is (67.2%) + (fi - fav) - (gi - gav) = 65.04 for Djokovic.
Totally different numbers than mentioned on the paper where it says.
Putting all this together using Barnett and Clarke’s equation, we find that it predicts serve-winning probabilities of 67.0% for Roger Federer and 68.6% for Novak Djokovic.
I don't know where they got those numbers from as I believe I have done the calculation correct. None of this makes sense.