Jump to content

Correlation Graphs From Entire Playerbase


29 replies to this topic

#1 Tarogato

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Civil Servant
  • Civil Servant
  • 6,558 posts
  • LocationUSA

Posted 13 July 2016 - 06:25 PM

This information is not really that useful, but it's pretty to look at, so here it is! The leaderboard was exported to spreadsheet by So1ahma and posted here. Removing players with fewer than 30 matches played leaves us with 21,120 players worth of data to work with.

Analyses in italics written by this dude here, GearsAndSuch.



Posted Image

KDR vs WLR
Most players have a KDR/WLR around 1. Without doing real stats, you know you're a real, better than 95% if your KDR is 1.8 or better or your WLR is 1.5 or so. BUT! It looks like the WLR dependence on KDR is low: so there's a lot of folks with sky high KDR that don't have great WLR, but not the other way around.




Posted Image

Match Score vs WLR
Match score of 200 and a WLR of around 1 is normal. Average match score of over 275 or so is exceptional, but even then the WLR is only around 1.5. There is a lot more scatter in WLR; there are players with high WLR but low match scores.





Posted Image

Match Score vs KDR
Strong, though none linear relationship here! (Raise your hand if you're surprised. Oh I see...) Obviously the KDR and match score center around 1 and 200 respectively (based on my prior comments). There's an inflection point at around match score of 225, where slope steepen, meaning that generally, higher scoring players win more. At 100, the KDR is about .25, but at 300 it's 2.1. There are very few players with a match score over 350 and there is considerable variations in their KDR.




Posted Image

Death Rate vs WLR
Inverse cubic relationship, meaning, that if you the lower your death rate, the exponentially greater your WLR. I'm pretty sure this concept is Church of Skill Gospel, but not something that's embraced by the workaday PUG. The average player looks to have a death rate of about 0.67, meaning they're more likely to die in a match than not. Interesting, the ~95% confidence death rate is still 0.55, meaning even good players are dying more than not. At that point, their WLR is 1.4 or so. The shape of the trend is dictated by a small group of outliers that have deathrates of less than .5 and WLR that ranges widely from 0.5 to 5... in other words, there's some people who manage to never win and never die, probably over drugged baluga whales in Urbies who power down on a building at the first sign of trouble, and some people who are the MWO embodiment of Connor Macleod (Andrian Paul version of course).

It might be worth noting the neat geometric patterns that show up in the Death Rate and Kill Rate graphs. They appear because these figures (kills or deaths) are divided by match count and a significant portion of players have sub-100 matches played, which makes those numbers (<100) a very common divisor. Why exactly the shape is in the form of this geometric curvy pattern? I have no idea. Maybe it was aliens.




Posted Image

WLR vs Kill Rate
Weak relationship between these two... huge amount of scatter in kill rate, but WLR is constrained between 0.1 and 2 for most of MWO. Average kill rate is like ~0.6, with exceptional players having Kill rates between 1-1.25 (too much scatter for my stats-squints to help me). The scatter in the WLR starts to increase quite a bit after a kill rate of 0.5, meaning that there are few players with high WLR and low kill rates, buit there are some players with fairly high kill rates that tend to loose. There are no players with a WLR less than after a kill rate of 1.5, so at some point, pure brutality takes on a quality all it's own. Conversely, there are a few datapoints with huge WLR with a Kill rate between 0.65 and 1.25. Real team players? Owners of cap accelerator? Someone with time should solve this with multiple regression.




Posted Image

Illuminati confirmed! lol, that geometric pattern is more obvious than ever. Wicked cool, haha.

Kill Rate vs Death Rate
Stronger, negative relationship here... the more you kill, the less you die. MWO immortals with a kill rate of 2 have an average death rate of 0.45. The average player (as you could surmise from above) will have an average kill rate of ~0.65 and and average death rate of, you guessed, about ~0.67 (See how those are related? Good! even you can claim mastery of basic statistics when you're on the internet). The scatter is pretty well constrained: there's no suicide jockeys that are regularly taking out a 2 mechs, then overheating from excitement.

Edited by Tarogato, 13 July 2016 - 06:31 PM.


#2 Deathlike

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 29,240 posts
  • Location#NOToTaterBalance #BadBalanceOverlordIsBad

Posted 13 July 2016 - 06:31 PM

So... learn to shoot and focus fire?

Who knew working together to kill things would improve your W-L + K-D ratio!

For the high K-D ratios with low W-L - the reality is that there are people that don't bother doing/knowing the objective at times (though it doesn't mean the Assault should be the first to cap on Conquest), so loss by cap can occur.

Also, killing things makes it easier to win the match, regardless of the results. If you're not really holding your own (at around 1 k-d or better), then you're not really helping the cause.

Edited by Deathlike, 13 July 2016 - 06:31 PM.


#3 El Bandito

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • Big Daddy
  • Big Daddy
  • 26,736 posts
  • LocationStill doing ungodly amount of damage, but with more accuracy.

Posted 13 July 2016 - 06:33 PM

I'll christen them, the Pug Nebulae.

#4 Chuck Jager

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 2,031 posts

Posted 13 July 2016 - 06:55 PM

View PostDeathlike, on 13 July 2016 - 06:31 PM, said:

So... learn to shoot and focus fire?

Who knew working together to kill things would improve your W-L + K-D ratio!

For the high K-D ratios with low W-L - the reality is that there are people that don't bother doing/knowing the objective at times (though it doesn't mean the Assault should be the first to cap on Conquest), so loss by cap can occur.

Also, killing things makes it easier to win the match, regardless of the results. If you're not really holding your own (at around 1 k-d or better), then you're not really helping the cause.

That would be 1kmd not 1 kill. I actually had a game in group where I had 75 damage and 3 kills (the game was over in less than 3 minutes). Kills are many time pure luck if you are focusing fire with others as a habit.

#5 Deathlike

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 29,240 posts
  • Location#NOToTaterBalance #BadBalanceOverlordIsBad

Posted 13 July 2016 - 07:05 PM

View PostChuck Jager, on 13 July 2016 - 06:55 PM, said:

That would be 1kmd not 1 kill. I actually had a game in group where I had 75 damage and 3 kills (the game was over in less than 3 minutes). Kills are many time pure luck if you are focusing fire with others as a habit.


It averages out in your favor though. If it's not an assist, it's probably KMDD if you're doing a lot of work.

Worry less about your own stats sometimes - it's better to worry about winning the match in whatever method possible.

#6 Tarogato

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Civil Servant
  • Civil Servant
  • 6,558 posts
  • LocationUSA

Posted 13 July 2016 - 08:28 PM

I may have gotten a little carried away with that weird pattern that kept showing up in the scatter plots...



Posted Image


\m/ (>_<) \m/

#7 FuhNuGi

    Member

  • PipPipPipPipPip
  • Shredder
  • Shredder
  • 182 posts
  • LocationMendocino California

Posted 13 July 2016 - 11:54 PM

Nice job on the statistical analysis.
I am curious about the "spread" on the pattern and if there was much expansion /contraction if you adjusted the threshold for games played to a higher number of games played... in other words did more "experienced" players make for a tighter grouping?

Can you give the guys working on the leaderboard a hand now?

#8 meteorol

    Member

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 2,848 posts

Posted 14 July 2016 - 12:06 AM

View PostDeathlike, on 13 July 2016 - 06:31 PM, said:

For the high K-D ratios with low W-L - the reality is that there are people that don't bother doing/knowing the objective at times (though it doesn't mean the Assault should be the first to cap on Conquest), so loss by cap can occur.


That aside, if you only play solo queue, you will get a fair amount of games in which you drop like 5 solo kills 3 kmdd on 1k+ damage and still lose the game. It happens more often than one would like. Futhermore there are plenty of people that are able to get 2-3 kills on average no matter if they win or lose the match, which can lead to k/ds of 5+, but the vast majority of players won't be able to drag solo pugs to a w/l of 5+. I'll just go ahead and mention kaffeeangst because he will mention himself anyway if no one else does. Even with a k/d of 10+ and ~1000 average damage with the KDK-3 his w/l is below 5. There is only so much one person can do in solo queue.

#9 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,221 posts

Posted 14 July 2016 - 12:40 AM

Good thing
Stabilization level at W/L = 1, K/D = 1 and MS = 200, I have been talking about for ages, but nobody has believed me, is confirmed!

Bad thing
Spread of players' performances is way too big - matchmaker doesn't do it's job, i.e players with bad and good stats are being mixed in one match -> unbalanced matches. Majority of players has W/L and K/D < 1, which means minority of Meta/Hardcore players stomps casual majority, while they should be separated from each other. In case of good matchmaker, that evens players in teams better, results should be closer - closer to K/D = 1, W/L = 1 and MS = 200 for all players, except tiny minority of players with lowest and highest skills, who simply don't have competitors to even them out.

Posted Image

Edited by MrMadguy, 14 July 2016 - 12:48 AM.


#10 PhoenixFire55

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,725 posts
  • LocationSt.Petersburg / Outreach

Posted 14 July 2016 - 01:08 AM

Nice work by all the guys who contributed in making these graphs. Sadly unless we can separate solo/group queue results its not saying much apart from the good idea about what the "average for the average" is.

Would have been really interesting if we could have same stats from scouting mode to compare how personal Kill/Death ratios affect games between smaller teams.

#11 L3mming2

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 1,304 posts

Posted 14 July 2016 - 02:36 AM

nice stats :), douse the leaderboard include stats from groop que? because those pesky team players may very well influence the resulting trends..

(a super star 12 man team will have a verry high W/L ratio but there mach score and kill ratio will be only avrage high due to being divided equaly between the 12 players...)

#12 meteorol

    Member

  • PipPipPipPipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 2,848 posts

Posted 14 July 2016 - 02:46 AM

View PostL3mming2, on 14 July 2016 - 02:36 AM, said:

nice stats Posted Image, douse the leaderboard include stats from groop que? because those pesky team players may very well influence the resulting trends..

(a super star 12 man team will have a verry high W/L ratio but there mach score and kill ratio will be only avrage high due to being divided equaly between the 12 players...)


It does. Pretty obvious that people don't get W/Ls of 28 by playing solo.

#13 Deathlike

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 29,240 posts
  • Location#NOToTaterBalance #BadBalanceOverlordIsBad

Posted 14 July 2016 - 10:51 AM

View Postmeteorol, on 14 July 2016 - 12:06 AM, said:


That aside, if you only play solo queue, you will get a fair amount of games in which you drop like 5 solo kills 3 kmdd on 1k+ damage and still lose the game. It happens more often than one would like. Futhermore there are plenty of people that are able to get 2-3 kills on average no matter if they win or lose the match, which can lead to k/ds of 5+, but the vast majority of players won't be able to drag solo pugs to a w/l of 5+. I'll just go ahead and mention kaffeeangst because he will mention himself anyway if no one else does. Even with a k/d of 10+ and ~1000 average damage with the KDK-3 his w/l is below 5. There is only so much one person can do in solo queue.


I get that. You can't really carry everyone in the solo queue. That's why people being only mech rambos/snipers at times aren't really helpful (not enough shared armor).

#14 RePlayBoy101

    Member

  • PipPipPipPipPipPip
  • Shredder
  • 408 posts

Posted 14 July 2016 - 10:58 AM

really people are only averaging between 200 and 300 dmg ... wtf

#15 Deathlike

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 29,240 posts
  • Location#NOToTaterBalance #BadBalanceOverlordIsBad

Posted 14 July 2016 - 11:09 AM

View PostRePlayBoy101, on 14 July 2016 - 10:58 AM, said:

really people are only averaging between 200 and 300 dmg ... wtf


That's better than the derping I see, with sub-triple digit damage.

#16 Mister Blastman

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Survivor
  • Survivor
  • 8,444 posts
  • LocationIn my Mech (Atlanta, GA)

Posted 14 July 2016 - 11:11 AM

View PostDeathlike, on 13 July 2016 - 06:31 PM, said:


For the high K-D ratios with low W-L - the reality is that there are people that don't bother doing/knowing the objective at times (though it doesn't mean the Assault should be the first to cap on Conquest), so loss by cap can occur.



And then there's the whole group queue effect skewing the win/loss ratio numbers.

Awesome post, OP!

#17 the hedgehog

    Member

  • PipPipPip
  • 80 posts

Posted 14 July 2016 - 11:17 AM

Wins / Losses Ratio 1.35
Kill / Death Ratio 1.37
pretty nice considering how the game can be pretty punishing.

Edited by the hedgehog, 14 July 2016 - 11:18 AM.


#18 EvilCow

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,243 posts

Posted 14 July 2016 - 11:24 AM

Thanks for the charts, this is hugely interesting.

Good Job!!

#19 dervishx5

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Workhorse
  • The Workhorse
  • 3,473 posts

Posted 14 July 2016 - 11:27 AM

Do they have stats for Faction Warfare?

#20 Bishop Steiner

    ForumWarrior

  • PipPipPipPipPipPipPipPipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 47,187 posts
  • Locationclimbing Mt Tryhard, one smoldering Meta-Mech corpse at a time

Posted 14 July 2016 - 11:39 AM

Biggest Problem I see with this:

Players will draw the wrong conclusion:

The Less I die, the better my WLr will be!

And go hide more, to "avoid dying". When the opposite is the case, the less on plays NOT TO DIE, but to kill, the better one's chance of surviving and winning is.

Good charts and all... but end of the day we know what it takes to win more than to lose, and have a positive KDr.... Balls and coordination. The majority of the playerbase will never have the former, nor attempt the latter. It's why Team Play skews one's stats so dramatically. Coordination and Aggression lead to upward trends in all stats.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users