3

I'd like to model Tennis results in a logit regression. How can I include head-to-head data?

Simply taking the share of won matches between two players isn't ideal, because it doesn't take the number of matches into account. E.g., a H2H of 20-1 would be worse than 1-0.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
ljmc
  • 31
  • 1
  • 1
    What is it you are ultimately trying to do? What do you want from a model of tennis results? Are you including both players from each match? – gung - Reinstate Monica Mar 03 '21 at 18:16
  • We need more details, but you must give individual results in data, not just proportions! Look at for instance https://stats.stackexchange.com/questions/509956/analysis-of-binomial-variables-with-random-effects/510265#510265 and its links. – kjetil b halvorsen Mar 04 '21 at 13:15

0 Answers0