@Adje word thanks a ton this is exactly the type of thing I was looking for, I can just manually double check the RGL ones and try to find non-combined versions for any of them which are combined matches
as a side note if anyone has ideas they’d be interested in someone analyzing feel free to post or msg me
current ideas:
- which metrics are the most accurate predictor of wins for a certain class (e.g., does KD actually not matter on roamer, which stats are actually most important to look at for evaluating meds)
- make a tool to generate additional insights on a single log: which players had the most impact/least impact on that particular game (potentially making use of the ‘events’ in a log to count things like who dropped a med, vs the overall summary stats of total kills)
- train a model to predict the outcome of a pug or match just by the players in it and their classes (could be used for automated pug balancing like mixchamp/faceit)
- make some tool where a player can look themselves up and find stats over all logs like avg dpm for a given class/map, as well as seeing their “elo”/rank according to the model above
- limit the data just to pugchamp (easy to identify bc the name) and find the most valuable player (delpo, obviously) whose inclusion in pug teams made the most impact on their teams winning (would be interesting to compare what this comes up with vs pugchamp’s elo system if I could get that data from erynn/tsc/whoever has access to it)
- minor contribution but was thinking it would be cool to include stuff like forces, times you dropped a med, backcaps on logs.tf through a browser plugin (similar to the one that gives DPM - DT/M), maybe even minor things like times flashed w/ uber or dropped from uber altho that might be harder to determine, thinking it could be useful for casts