[dlg]

The change you describe from N+1 to N+2 wasn’t made to defeat your hack. It was made because we got lots of complaints that people thought the filters were buggy/broken when they saw people who didn’t match their filters and because they were seeing irrelevant people, lowering the chance of a match. The set of people who already liked you were being served out of a different service than regular recommendations and it was, iirc, just a Redis list until fixed. (More generally, we never purposely made the recommendation algo worse to increase boosts or because people would only stay if they didn’t meet someone during my era, even if everyone thought that’s we did. I haven’t been involved in several years however.) In any case, sorry!

[GameOfFrowns]

>More generally, we never purposely made the recommendation algo worse to increase boosts or because people would only stay if they didn’t meet someone during my era, even if everyone thought that’s we did.

Would anyone admit this publicly? That's a surefire way to destroy your career or getting sued.

Also begs the question for which performance indicators did you optimize if not engagement and retainment?