Author here. You're absolutely correct in that there are two levels of skill indicator.
Something like a rank or rating is the gold standard for determining how good a player is against the rest of the group rated by the same system. It's the gold standard because it's verifiable: it translates to a probability of winning and this can be calibrated over time to increase faith in the rating. But by design, it's a fairly slow-moving indicator.
Faster feedback is interesting for figuring out in-game decision rules or improving more effectively, and that's where these per-match ratings like the Leetify rating comes into play. They are naturally more difficult to verify because they are working on a very small, noisy sample, and since the outcome of the match (win/loss) is attributed to an entire team, it can be hard to detangle the individual contributions to that, without resorting to subjective judgment. Leetify have, from what I understand, tried to take a quantitative/verifiable approach to it, but they haven't shared any details around exactly how they verified it – if at all.
I'll take a look at tracker.gg and see what else I can learn about this. Thanks for sharing!
Something like a rank or rating is the gold standard for determining how good a player is against the rest of the group rated by the same system. It's the gold standard because it's verifiable: it translates to a probability of winning and this can be calibrated over time to increase faith in the rating. But by design, it's a fairly slow-moving indicator.
Faster feedback is interesting for figuring out in-game decision rules or improving more effectively, and that's where these per-match ratings like the Leetify rating comes into play. They are naturally more difficult to verify because they are working on a very small, noisy sample, and since the outcome of the match (win/loss) is attributed to an entire team, it can be hard to detangle the individual contributions to that, without resorting to subjective judgment. Leetify have, from what I understand, tried to take a quantitative/verifiable approach to it, but they haven't shared any details around exactly how they verified it – if at all.
I'll take a look at tracker.gg and see what else I can learn about this. Thanks for sharing!