This may be helpful:
However it's only an estimate. There's no straightforward way to just take two players' winning percentages against a general field and apply it to them playing each other. The reason is that an e.g. 60% winrate could mean at one extreme that I always beat the bottom 60% of players and always lose to the top 40% (step distribution), or at the other extreme that I beat every player with a probability of 60% (uniform distribution). Or something in between.
So you need to know more than just the winrate against the overall field if you want to have an accurate answer.