Jump to content

Jarls


196 replies to this topic

#21 N0MAD

    Member

  • PipPipPipPipPipPipPipPip
  • 1,757 posts

Posted 22 March 2018 - 06:24 PM

View PostMischiefSC, on 22 March 2018 - 05:43 PM, said:

The stats themselves however are very useful for looking at how you perform.

How you perform where? in solo, group or CW, because these are all very different, but all are mixed in one lump?.
when these Qs are separated then in may give you a metric till then they are as useful as b oobs on a bull.

#22 justcallme A S H

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • CS 2020 Referee
  • CS 2020 Referee
  • 8,987 posts
  • LocationMelbourne, AU

Posted 22 March 2018 - 06:58 PM

View PostRustyBolts, on 22 March 2018 - 01:51 PM, said:

I fail to see how a WLR is a viable measurement. The WLR depends solely on the team you are on. If you end up pugging, your WLR will be lower vs an organized team. This means you can still be a very good player, just a pug player. I also see he has a formula that determines adjusted match score. Anyone can make up a formula and use that as a measurement, but it does not really mean anything, unless you believe in that formula. What is the Percentile supposed to mean?


Percentile is where you fit - perhaps a google of the term and its use in statistics will better help you understand.

WLR is an excellent measurement as is KDR and then matchscore. I solo most of my matches and my WLR is above 2.0, counting only the last few seasons (after learning the game/doing more comp/improving) it is much higher.

Also it is reasonably easy to tell a good player that does more GroupQ or SoloQ.

SoloQ will generally have:
- Lower WLR
- Lower KDR
- Higher Avg Match score

GroupQ will generally have:
- Higher WLR
- Higher KDR
- Lower Avg Match Score

Things become skewed when player mixed Solo and Group queues a lot, then it's hard to tell but you can work it out easily enough if you understand what the stats actually mean.

View PostN0MAD, on 22 March 2018 - 06:24 PM, said:

How you perform where? in solo, group or CW, because these are all very different, but all are mixed in one lump?.
when these Qs are separated then in may give you a metric till then they are as useful as b oobs on a bull.


Jarls does not count Faction Play.

Only QP leaderboards that are available to scrape. Faction Play is not.

#23 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,208 posts

Posted 22 March 2018 - 07:39 PM

Majority of playerbase just don't understand, that average MS can't be used as stat for leaderboard. Because skill - is relative thing. I.e. 250MS in Tier 5 and 250MS in Tier 1 - are two completely different things. Only PSR itself - could be only viable stat for leaderboard, if only this system wouldn't be broken. And PSR is still hidden from us.

But it's the best, we have now. PSR is broken - it's just an XP bar. Devs are incompetent and assume, that players can improve with time almost infinitely. That there is no physical skill cap. And MM is almost inexistent - it mixes all players with different ratings together. And therefore avg MS - is really the only thing, that shows your real performance vs other players.

WLR can't be used for performance measurement, as ELO-like MM usually simply tries to keep it at ~1 level for all players. KDR - may be, but it doesn't account many other ways of being useful for your team.

Edited by MrMadguy, 22 March 2018 - 07:42 PM.


#24 Y E O N N E

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Nimble
  • The Nimble
  • 16,810 posts

Posted 22 March 2018 - 07:47 PM

The fact that players with different ratings are mixing together is precisely why you can use average match score for the leaderboard.

#25 InspectorG

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Boombox
  • The Boombox
  • 4,469 posts
  • LocationCleveland, Ohio

Posted 22 March 2018 - 07:49 PM

View PostMrMadguy, on 22 March 2018 - 07:39 PM, said:

Majority of playerbase just don't understand,...


Amen.

#26 Bombast

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 7,709 posts

Posted 22 March 2018 - 08:06 PM

View PostMrMadguy, on 22 March 2018 - 07:39 PM, said:

WLR can't be used for performance measurement, as ELO-like MM usually simply tries to keep it at ~1 level for all players. KDR - may be, but it doesn't account many other ways of being useful for your team.


MWO does not have a proper ELO system. That's one of the reasons why stat swing is so wide.

Even if it did have such a system, WLR would still be a valid, and probably still best, stat metric. From a stat perspective, all that would happen would be overall WLR being 'compressed' towards the middle, reducing the spread of the stat. Instead of 2.0+ being the 'Really good' mark, something like 1.2 or 1.3 would be. The good players would still rise to the top, they just wouldn't rise quite so far.

#27 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,208 posts

Posted 22 March 2018 - 09:40 PM

View PostBombast, on 22 March 2018 - 08:06 PM, said:


MWO does not have a proper ELO system. That's one of the reasons why stat swing is so wide.

Even if it did have such a system, WLR would still be a valid, and probably still best, stat metric. From a stat perspective, all that would happen would be overall WLR being 'compressed' towards the middle, reducing the spread of the stat. Instead of 2.0+ being the 'Really good' mark, something like 1.2 or 1.3 would be. The good players would still rise to the top, they just wouldn't rise quite so far.

WLR would be valid performance meter in one case only - is there wouldn't be MM at all and matchmaking would be about "all vs all". Same for AvgMS.

And most of you just don't understand, how MM works. Or should work, but it doesn't... You should understand, that your performance doesn't depend on your pure skill only (that's, what we want to measure) - MM itself affects it. There is negative feedback system inside MM. Playing against Tier 5 isn't the same, as playing against Tier 1. What MM actually tries to do - is to bring your WLR and AvgMS to some average level. If your performance is lower, than this average - your rating drops, you start being matched against weaker enemies - your performance rises. If your performance is higher, than average - your rating rises and your start being matched against stronger enemies - your performance drops. And at the end every single player in this game SHOULD PLAY ON WLR = 1 AND AVGMS = 250 LEVEL. This doesn't happen only because MM doesn't work properly. That's why this stats can't be used for skill measuring - at the end they should be the same for all players, except may be some tiny minority of top 1% and bottom 1% players.

TL;DR Only PSR matters - stats are affected by MM itself.

Edited by MrMadguy, 22 March 2018 - 09:44 PM.


#28 MischiefSC

    Member

  • PipPipPipPipPipPipPipPipPipPipPipPip
  • The Benefactor
  • The Benefactor
  • 16,697 posts

Posted 22 March 2018 - 09:44 PM

View PostN0MAD, on 22 March 2018 - 06:24 PM, said:

How you perform where? in solo, group or CW, because these are all very different, but all are mixed in one lump?.
when these Qs are separated then in may give you a metric till then they are as useful as b oobs on a bull.


Group queue has a smaller population than FW, at least in certain times of day. GQ matches are about as relevant. They can skew results a bit but not nearly as much as people think. Their impact affects a small number of players out of the whole.

#29 meteorol

    Member

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

Posted 22 March 2018 - 10:01 PM

View PostMischiefSC, on 22 March 2018 - 09:44 PM, said:


Group queue has a smaller population than FW, at least in certain times of day.


Yeah australian primetime, maybe.

#30 Xiphias

    Member

  • PipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 862 posts

Posted 22 March 2018 - 10:05 PM

View PostMrMadguy, on 22 March 2018 - 09:40 PM, said:

TL;DR Only PSR matters - stats are affected by MM itself.

This is only true up to T1. Basically, there are five buckets and any T1 player is considered just as good as another T1. If you are in a low tier and you win you will move up, matches will get harder and it will try to drive your WLR down. If you are in a high tier and lose a lot you will (maybe) move down and play easier opponents, which will try to drive your WLR up. To this extent you are correct.

Once you get to T1 that changes though. Everyone in T1 is playing everyone else at an equal level, it becomes, ""all vs all"" within that tier. For that reason comparing a T1 and a T3 player might not be accurate, but comparing two T1 players is absolutely fair, and any player that doesn't get to T1 (baring insufficient matches) is going to be worse than those T1 players by default (since if they were better MM would push them up to T1).

#31 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,208 posts

Posted 22 March 2018 - 10:12 PM

View PostXiphias, on 22 March 2018 - 10:05 PM, said:

This is only true up to T1. Basically, there are five buckets and any T1 player is considered just as good as another T1. If you are in a low tier and you win you will move up, matches will get harder and it will try to drive your WLR down. If you are in a high tier and lose a lot you will (maybe) move down and play easier opponents, which will try to drive your WLR up. To this extent you are correct.

Once you get to T1 that changes though. Everyone in T1 is playing everyone else at an equal level, it becomes, ""all vs all"" within that tier. For that reason comparing a T1 and a T3 player might not be accurate, but comparing two T1 players is absolutely fair, and any player that doesn't get to T1 (baring insufficient matches) is going to be worse than those T1 players by default (since if they were better MM would push them up to T1).

That's, why PSR is so broken - this shouldn't happen. Players with different skill levels shouldn't have equal PSR rating. That's why quality of matchmaking is so low in this game.

But anyway. Without accurate information about ratings you still can't determine, who is true Tier 1 player and who is just on a smurf account, stomping Tier 5 noobs with 100500WLR and 100500AvgMS.

Edited by MrMadguy, 22 March 2018 - 10:14 PM.


#32 Xiphias

    Member

  • PipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 862 posts

Posted 22 March 2018 - 10:26 PM

View PostMrMadguy, on 22 March 2018 - 10:12 PM, said:

That's, why PSR is so broken - this shouldn't happen. Players with different skill levels shouldn't have equal PSR rating. That's why quality of matchmaking is so low in this game.

But anyway. Without accurate information about ratings you still can't determine, who is true Tier 1 player and who is just on a smurf account, stomping Tier 5 noobs with 100500WLR and 100500AvgMS.

You'd have to have the same number of tiers as players to do that. That's what ELO was and people complained about it so now we have this.

It's pretty easy to tell the smurphs from the T1 players. Just look at the seasons played and the matches played. Anyone with few matches played and high stats is probably smurphing. If they do it long enough they will move up to T1. While this doesn't work 100% of the time it's true for enough of the population that you can draw useful information from the leaderboard.

You aren't going to be smurphing and stat farming T5 with 1000 matches played.

#33 Grus

    Member

  • PipPipPipPipPipPipPipPipPip
  • Little Devil
  • Little Devil
  • 4,155 posts

Posted 22 March 2018 - 11:38 PM

View PostLordNothing, on 22 March 2018 - 06:07 PM, said:

i like it because it gives you an at a glance look to see if you are getting gud or not.
right there. Like my stats are going to drop a bit most likely due to the fact that I haven't been able to play a lot so I'm hoping my previous seasons performance will help balance it out a bit.

#34 Catra Lanis

    Member

  • PipPipPipPipPipPipPipPip
  • Bridesmaid
  • 1,182 posts
  • LocationSweden

Posted 23 March 2018 - 04:11 AM

View PostSjorpha, on 22 March 2018 - 02:02 PM, said:


Not true.

The impact of teams in solo queue averages out over time as the teams are random, the long term trend to your WLR is from your own impact on the matches. A hundred or so matches is where it starts being a useful stat, so for most people playing the game it's fairly reliable.

If you play more group queue than solo and drop in fairly large groups that can certainly skew your WLR quite a bit if the team you're playing with have very consistently high or very low WLR, so it would definitely be good if we could track only solo Q WLR for the purpose of ranking.

So yeah you have to look at a little context, but it's actually the most useful stat for measuring overall skill. Much more useful than secondary things like damage, KDR etc.

Very good players tend to have very good stats all over though, from WLR to damage and MS.


That would be true if everyone played the same mech and did so all the time. I've had 30 losses in a row and the majority of those games I was in the upper 70% damagewise. I've also had a lot of matches where one or even two players have been absent and we still won by a big margin. Yes you affect w/l but I want to see proof that you do it to such a high degree as stated.

Edited by Catra Lanis, 23 March 2018 - 04:12 AM.


#35 Bombast

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 7,709 posts

Posted 23 March 2018 - 04:16 AM

View PostCatra Lanis, on 23 March 2018 - 04:11 AM, said:

That would be true if everyone played the same mech and did so all the time. I've had 30 losses in a row and the majority of those games I was in the upper 70% damagewise. I've also had a lot of matches where one or even two players have been absent and we still won by a big margin. Yes you affect w/l but I want to see proof that you do it to such a high degree as stated.


First, 30 is not a large enough number.

Second... well, I'm not going to go there. I'm sure someone else will though.

#36 Tier5 Kerensky

    Member

  • PipPipPipPipPipPipPipPipPip
  • Bridesmaid
  • 2,049 posts

Posted 23 March 2018 - 04:41 AM

View PostRustyBolts, on 22 March 2018 - 02:44 PM, said:

Xmith, The stats, at least mine are not even close to what MWO says. How often does it update?


They are identical to MWOs leaderboard data, because that's where it's from.

Notice it's only QP, if you play a lot of FP too then your totals will look different. As well as the data is based from season 1 onwards so that was early 2016?? Or something close when the leaderboards game, about the same time as I started playing. First season was longer, later seasons were all one month.

#37 EnochsBook

    Member

  • PipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 163 posts

Posted 23 March 2018 - 05:14 AM

View PostCatra Lanis, on 23 March 2018 - 04:11 AM, said:

That would be true if everyone played the same mech and did so all the time. I've had 30 losses in a row and the majority of those games I was in the upper 70% damagewise. I've also had a lot of matches where one or even two players have been absent and we still won by a big margin. Yes you affect w/l but I want to see proof that you do it to such a high degree as stated.

Last season I played 186 matches, all of them in solo queue. I have a WLR of 1.42. Is that because I am luckier than someone whose winrate over a similar number of matches is lower, or do you think there could be any other reason?

Edited by EnochsBook, 23 March 2018 - 05:15 AM.


#38 Krivvan

    Member

  • PipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 4,318 posts
  • LocationUSA/Canada

Posted 23 March 2018 - 06:05 AM

View PostBombast, on 22 March 2018 - 01:31 PM, said:

There's inherit problems in the data Jarl's uses that means it shouldn't be used as an absolute metric, but generally speaking, 1.2 or better WLR Good, 0.8 or worse WLR Bad. And so on.

There are a ton of inherent problems with the data it uses, but it serves as a useful very, very general look at a player's performance. The difference between a 95% player and 99% player is minimal (or rather the variance can be large). But the difference between 50% and 90% is usually quite apparent.

#39 Krivvan

    Member

  • PipPipPipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 4,318 posts
  • LocationUSA/Canada

Posted 23 March 2018 - 06:09 AM

View PostCatra Lanis, on 23 March 2018 - 04:11 AM, said:

Yes you affect w/l but I want to see proof that you do it to such a high degree as stated.

I've done this before. Take several nights of publicly streamed solo queue matches while playing a bunch of different mechs and looking at the W/L. It always was noticeably higher than 50/50.

#40 Asym

    Member

  • PipPipPipPipPipPipPipPipPip
  • Nova Captain
  • 2,186 posts

Posted 23 March 2018 - 06:16 AM

Ah, guys.....what a silly discussion. The data is nice and Jarl's does a good job using.....un-verified and wildly skewed data.

I went to Jarl's just to "see for myself" what this tool produces. I am missing seasons 4 through 7, The data in season 20 is not correct. And, after retiring from corporations that have extreme reliability metrics, ah, MWO could never produce statistics that would ever make it to any of the corporate board rooms I've served in....never. It is that invalid as many in this article and a dozen other articles have laid out... A good try at something that has zero real value.....

Look at the enclosed link to a good article:

http://www.breitbart...ic-meritocracy/

Enclosed is a good article in a growing academic movement. I've alluded to before several times and many scoffed at the idea that video games are being "studied" in depth and many doctoral projects are looking at why players play and what players see as value-added in the games they play ! As I have said before, there are human factors teams across the country looking at what is the link between technologies and value-added video-game practices. What drives a million player every hour to play game "x"??? What is that link. Why?

MWO isn't as toxic as some but, look at this thread and ask yourself: what are we discussing? "How good am I": as if that means something in a game where where less than 5% of the players are actually competitive players? Or, are metrics needed in a game at all if all you want to have is fun? And, if what you do having fun are the metrics others use to devalue your experiences having fun, what is the point of having statistics in the first place??? I.e. "Get GuD" noob, Potato trash, your stats prove you are a potato and what you say is stupid, etc.. There are some "very smart" and highly experienced "potatoes" out there......

Data is only usable if it can be "validated" against metrics that are rock solid standards.... everything else is "Meritocratic" and that is a slippery slope.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users