On paper it doesn’t seem like it’d be a long term fit- i have a cat, she has a dog. I like a lot of metal, she likes jam bands. Lot of room for the algorithm to predict we’d have a few dates and then it wouldn’t work out.

It's almost as if those preferences are not really significant in long-term relationship outcomes.

My wife and I differ on a large number of preferences, but we align in a few that are the most important to us. If were were a 100% match on superficial characteristics, that'd be boring. We've been married a very long time.

Dating algorithms are fundamentally unsound because many (most?) preferences are non-commensurate, so weighted scores are impossible to derive from pairs of them. There is no tradeoff of dog vs cat relative to metal vs shoegaze, and people and (and often do) enjoy both dogs and cats or metal and shoegaze. There's no useful information to be found in how people self-report their scores on unrelated false dichotomies. It's just shit experimental design.