Player rating for Valorant?

hmm that could be an issue since Riot aren't a fan of that kind of stuff. Hopefully they go a bit more relaxed with it or make their own

Sure that would be interesting. Obviously applying this to Valorant unlike CS like i said would come with changes but this could be the right way to go. Maybe utility success or something along those lines could also be interesting as an input

wouldn't it better to set agent ratings ???

If you mean combining agent comparison with K/D, I agree, that would be closer to a more specific, meaningful comparative than just ACS. But then you run into the question of the role of that player in the team and what agents the others in that team tend to run. Some players simply get support from their team to perform well on that metric. Sentinel/Sova baiters and superstar duelists are the dominant players in this category. So ultimately it is not a good measurement of how good a player is compared to other players, it's just selective of that category of players. Is a Jett main OPer, like Shahz, Jamppi or Wardell comparable to a Jett initiator/fragger like cNed, babybay or russ? What about someone who creates space for their team, like Derke? Difficult to say.

Overwatch compares the skill of players based on the hero they play and how divergent they are statistically compared to other people playing that hero across many metrics. The reason this works for OW is because the role of a hero and what you do with that hero is more or less well defined and reflected by the metrics. The roles of agents in Valorant are not always so well defined, so agent comparisons are only useful in some cases. Either the roles of the agents are well defined and are comparatives in and of themselves. Or there are no "heroes" or "agents" and the players effectively play under the same constraints, making them mostly comparable, like in CS.

If you don't have either of those qualities in the game, creating a model to compare performance of players requires some heavy assumptions, which create biases, which ultimately only make the results of that model reflect that model's assumptions. So then it just becomes a question of "what assumptions are correct", which will create a years long shitstorm debate about the "correct" way to do things. In OW, you can reasonably say what players are the best at their hero. In CS you can reasonably say who are the most impactful pros with KAST. In Valorant it's difficult to say, because people value different things & the way agents are utilized is not easily quantifiable; objectivity goes out the window.