[steve2005]

I was not playing much with wife and had a low compatibility rating.
playing much more and rating hasnt moved much

[sfi]

Maybe BBO knows more than you think.

Oh, different sort of compatibility rating. No idea about that. But it seems pretty random even for regular partners.

[barmar]

Playing together doesn't change the factors that are used to compute compatibility.

[Rain]

I would agree the comp. score algorithm needs some major work. It's one of those features we added in last minute for fun and then didn't get back to it. I forgot whose idea it was but it could be a lot more useful.

I can't reveal (and don't know!) the algorithm, but if you're curious, skill similarity is one of the major components right now. Of course that leads to the question of how we actually compute skill similarity (very crudely!).

[wynsten]

Rain, on 2016-October-13, 13:45, said:

I would agree the comp. score algorithm needs some major work. It's one of those features we added in last minute for fun and then didn't get back to it. I forgot whose idea it was but it could be a lot more useful.

I can't reveal (and don't know!) the algorithm, but if you're curious, skill similarity is one of the major components right now. Of course that leads to the question of how we actually compute skill similarity (very crudely!).

Rather than (or in addition to) your weighted sum of the various components, it would be helpful if each profile contained the raw data itself (number of hands played, total IMPs scored, etc. etc whatever you use to determine compatibility) and let us decide whether that makes us "compatible". Then you won't have to work on an algorithm to determine compatibility OR skill level.

[1eyedjack]

Heh. In the time that it took you to analyse all that data you could have played 5 hands with them and have a far more reliable impression of compatibility than any algorithm would provide.