Jump to content

Better Metrics To Track Performance


No replies to this topic

#1 Haji1096

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 339 posts
  • Twitter: Link
  • LocationAlexandria, VA

Posted 24 March 2014 - 09:08 AM

View PostRoland, on 06 January 2014 - 01:38 PM, said:

There are various weapons in the game, and configurations of those weapons, which really do fall into the same category. They are simply BAD builds. An issue with this, once you get to that level of complexity, is that you start to get into evaluations which may have (or are at least perceived to have) more of a subjective component. The problem is, a lot of the time the idea that it's subjective stems entirely from ignorance on the part of the observer. They THINK that a given build is effective, but it's not, and can be effectively proven as such. Various weapons and builds fall into this category, and folks run them because they don't really understand the game to the degree where they can see how inferior those builds are. Often, they are misguided by focusing on incorrect metrics, or not really fully considering what is going on. For instance, an LBX 10 will result in high damage numbers... but it's mathematically an inferior weapon. You can clearly prove that it is going to be less effective at the only thing that matters, which is killing mechs. But many users continue to use them, because "they do well with them", failing to really do a full comparative analysis of those builds against other alternatives. We've even seen folks post videos that they believed showed them using LBX effectively, which actually clearly showed the inherent ineffectiveness of those weapons. They simply weren't noticing the right things.. For instance, they were killing targets in the video, but it was taking FOREVER to do so. To the extent that almost any alternative build would have killed the target sooner, and really the only reason they didn't die to the target was because the target was incompetent and running an even worse build.

So, the long and short of it is that no matter how hard you wish for a build to be effective, if wishes were horses then beggars would ride. There is an actual science to building mechs, and as a result some builds will be good and some will be bad.


Roland brings up a very good point about players looking at the wrong metrics to evaluate their performance with a particular build. Taking it a step further, PGI does not give us any mathematically correct metrics to judge a players performance, beyond damage, kills, assists, kdr. Each of which is can be dubious given a certain context.

For example:

Damage: You could have 1000 damage but it could be because you were spreading your shots around and not hitting the same component. You might have have been shooting at a mech that was neutered.

Kills: You might not have as many kills if you leg a mech, than move on. Or get a mech down to critical red ct, notice your heat is too high for a kill shot, leave him for a teammate. Then you relocate to kill the next threat while you cool down. Over a long period of time you might lose many kills, even though its the best way to help your team.

Assists: You could graze 12 mechs with an ERLL and hit them for 1 point of damage each. But you didn't really do anything useful other than scratch some paint.

KDR: You can inflate your KDR by playing in a 4 man against PUGS, powering down against insurmountable odds or suicide. I leave kill stealing out because I don't believe in kill stealing. If a team mate "steals" my kill it is almost always good. It saves me damage taken from that enemy and heat used for a shot I didn't have to take.

With regards to player performance here are some additional metrics I would like to see:

Number of times I overheat in a given mech per match

Damage per component destruction

Damage per Kill

Damage Taken

Effective Damage Taken = Damage Taken / Total Damage of Weapons shot at you within a percentage of its max range

Average Survivability = Average of time survived in a match / total match time

With regards to mech builds, I would like to see:

Effective armor = armor per square area of component protected. Why does an Atlas and an Awesome feel squishy ? Because their ct armor protects a large area.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users