28 replies to this topic

[Sephlock]

[Ghogiel]

BigBadHarv, on 01 June 2013 - 11:18 AM, said:

so the consensus seems to be that it stats dont matter because the player base is so bad?

why do we have elo then?

So that the steering wheel under hive can exist.

[p00k]

Edustaja, on 01 June 2013 - 09:30 AM, said:

I did my stats similarily:


588 games with all mechs that have more than 20 matches played.
Mostly solo or 8-man drops included.

Not a lot of correlation between win/loss and kill/death.

actually, short of those 2 outliers around a k/d of 1, that's a pretty reasonable correlation. also you've got a very narrow k/d range, from 1-2, and narrow w/l range, 1-1.8. if, for example, you only look at the k/d's on my graph that are under 5, it won't look like a good correlation either. you just don't have the sample size.

but, tell me if any of the other stats available to you correlate more strongly

Mawai, on 01 June 2013 - 08:04 PM, said:

I've observed that it is easier to obtain a high K/D ratio in two ways

1) Mount high damage weapons ... AC/20, Gauss, PPC ... when striking a vulnerable area (like red center torso) ... high damage weapons have a greater chance of delivering the killing blow if your aim is good.

2) Intentionally hunt mechs with red torso sections ... CT for all mechs, LT/RT for ones with a likely XL engine. Lights with a red leg. If a person concentrates their attacks on the weakest opponents the odds of getting the killing blow go up.

the meta dictates most people doing (1) anyways
if you're not doing (2) you're doing it wrong. you should always be hunting mechs with red torsos, lights with red legs, etc. it's not playing for k/d, that's what you should be doing

Mawai, on 01 June 2013 - 08:04 PM, said:

Of course the K/D is pretty meaningless since it only indicates who struck the final blow and not significance of the contribution to the kill.

spoken like a scrub, to be blunt. yes k/d only indicates who struck the final blow, but unless you only have 50 drops under your belt it evens out in the end. you're gonna kill some mechs other people did most of the "work" on, and others are gonna kill some that you did most of the "work" on.

a lot of people like to say damage done. you saw how k/d correlates with w/l for me. here's damage per match, for example:



some people say assists. mwo doesn't track assists anywhere visibly. i suspect assists should correlate better than damage, but not as well as kills. why? because i need only hit an enemy once in the arm, and as long as that enemy dies, i'll get an assist. of course i can't say that for sure since i don't have the numbers handy

anyways, bottom line is, killing mechs is the most direct way to help your team, ESPECIALLY when pugging. dying is the most direct way to hurt your team. scouting helps, yes, as does being the sacrificial distraction, but you're a better distraction if you stay alive, and a better scout if you kill the enemy scouts too. focusing on maximizing kdr as a number is dumb, but playing to kill more effectively and die as rarely as possible helps your team.

and kdr isn't meaningless. its meaning simply has to be taken in context - how much time you pug vs 4man vs 8man, how much time you spend eliting/mastering new mechs, etc. so comparing 2 people's kdr's may be meaningless, but comparing your own kdr to your kdr 2 weeks ago, and knowing how you've played for the last 2 weeks, holds quite a bit of meaning

edit: here's xp/match vs w/l

still not as strong as kdr. and remember, xp per match is DIRECTLY dependent on whether or not you win or lose

Edited by p00k, 01 June 2013 - 09:19 PM.

[Edustaja]

Could you post your dataset, eg. your mech stats page?

[Mawai]

p00k, on 31 May 2013 - 05:15 PM, said:

a "good" kdr is subjective. maybe your kdr isn't that good after all. my kdr correlates with my win/loss quite closely, R2=0.85. so yeah, kill more, die less, and you'll win more.

alternately, win more and you should die less, which should inflate your kdr. which is cause and which is effect? who knows?

Actually, it seems to me completely normal that W/L vs. K/D are correlated. They aren't independent variables and there are clear dependencies between each of the 4 variables involved.

1) Usually, when you lose you die ... not always but most of the time. Similarly when your team wins the odds of being a survivor are higher. So L and D are fairly closely correlated in many cases.

2) Usually, matches your side wins provide the greatest opportunity for kills. The probability of obtaining more kills go up when you survive and the entire opposing team is eliminated.

With these implicit correlations in the raw data ... there should always be a strong correlation between W/L and K/D when plotted. The difference between players, that might depend on pilot skill, would be the slope of the relationship. If you manage to kill more while dying less and maintaining a specific W/L ratio, across a range of mechs, then the slope of the plot would be different.

[Wispsy]

p00k, on 01 June 2013 - 09:13 PM, said:

numbers and reasonable arguments

Well after giving us so many of your stats you might as well screenshot that page which tells us what mechs they are on...I think I can guess but I am curious to know if I am right

[Chavette]

Edustaja, on 01 June 2013 - 10:05 PM, said:

Could you post your dataset, eg. your mech stats page?

Didn't you know you're not supposed to ask someones K/D? Its worse than asking the age of their wife.

[p00k]

eh, people who've dropped with/against me know what i usually run

[Sephlock]