Med mains (and coders) help me out
Let's pretend I wanted to rate a medic's performance just off of logs.tf stats (for a coding project)
I clearly want as few drops as possible, but what's a good "range" of effective healing in a 30 minute game, 20K heals?
Deaths are important too but can be due to the team's fault, and I have little to no idea what to expect for ubers, time to build, time before using etc...
Could you guys give me your insight on how you think this kind of information matters when it comes to evaluating a medic's performance?
It's a super crude approach because stalemates are not factored in during long games and tf2 is just a very robust game, but I wanna see what I can do with just the logs.
Also, any tips on working with logs.tf API? Any recs for alternatives?
Med mains (and coders) help me out
Let's pretend I wanted to rate a medic's performance just off of logs.tf stats (for a coding project)
I clearly want as few drops as possible, but what's a good "range" of effective healing in a 30 minute game, 20K heals?
Deaths are important too but can be due to the team's fault, and I have little to no idea what to expect for ubers, time to build, time before using etc...
Could you guys give me your insight on how you think this kind of information matters when it comes to evaluating a medic's performance?
It's a super crude approach because stalemates are not factored in during long games and tf2 is just a very robust game, but I wanna see what I can do with just the logs.
Also, any tips on working with logs.tf API? Any recs for alternatives?
35k+ healing is what you should try to hit every game. (However this is mainly building so dont read too much into it)
As few deaths as possible obviously.
Deaths after charge is a underlooked one it shows your post uber positioning which is very important.
Avg time before building is probably the most important. You want this to be as small as possible. Under 60 seconds is what the top medics hit in prem hit consistently. If you see a medic hit like 45-50 thats really good. Think June hit nearly 52/53 seconds in polan GF.
Deaths near uber is important as well it shows the uber tracking is on point. If you see a medic die a lot near uber they messed up on the tracking.
Advantages lost too.
And obviously kills.
Don't fall into the trap of damage taken being the most important.
So to answer your question:
A good medic performance will have not many deaths, few drops, few deaths after charge, 30k+ healing, 0 advantages lost and a low time to build.
35k+ healing is what you should try to hit every game. (However this is mainly building so dont read too much into it)
As few deaths as possible obviously.
Deaths after charge is a underlooked one it shows your post uber positioning which is very important.
Avg time before building is probably the most important. You want this to be as small as possible. Under 60 seconds is what the top medics hit in prem hit consistently. If you see a medic hit like 45-50 thats really good. Think June hit nearly 52/53 seconds in polan GF.
Deaths near uber is important as well it shows the uber tracking is on point. If you see a medic die a lot near uber they messed up on the tracking.
Advantages lost too.
And obviously kills.
Don't fall into the trap of damage taken being the most important.
So to answer your question:
A good medic performance will have not many deaths, few drops, few deaths after charge, 30k+ healing, 0 advantages lost and a low time to build.
Basically everything Turbo said, but I'm curious - are you going to use PCA or some sort of dimensionality reduction with the logs.tf dataset?
Basically everything Turbo said, but I'm curious - are you going to use PCA or some sort of dimensionality reduction with the logs.tf dataset?
The classic.
If you die to 2 rockets your fault.
If you die to 3 its your scouts.
The classic.
If you die to 2 rockets your fault.
If you die to 3 its your scouts.