Thinking more about it. I suspect Cara is not too different to what I did manually using Arena and a spreadsheet to convert Arena to an ELO:
Above for example I did Bobby Fischer Arena Analysis ages 12 -15 and compared them to the Arena Analysis I obtained from matches Mephisto Atlanta, Lyon and London out the posts here in the Forum.
OME is the overall game strength and they came out good as averages over many games. Individual games you could have as low as 1300 (even with Fischer) or as high as 3500 ELO all depending on the complexity of the game. But the averages worked out pretty well.
I suspect with Cara you have to do the same collect many games and let them average out. Individual games can only give you a rating for the individual complexity of a game.
So I suspect it is not Cara which probably works well but you need a number of games for each player to get a real result of ELO strength.
Regards
Nick