Jump to content

Psr Update And Changes - Jun 2020


490 replies to this topic

#261 MeterH1

    Member

  • Pip
  • Mercenary Rank 2
  • 10 posts

Posted 05 June 2020 - 10:59 AM

View PostPaul Inouye, on 03 June 2020 - 12:09 PM, said:



Well it seems that the overwhelming majority is fine with a Tier/PSR reset. And with that reset, we will make all future calculations zero sum in order to have PSR more accurately reflect player skill rather than experience playing the game.




I am all in favor of fixing the PSR issues, but as you said, these changes are to reflect player skill... That said, even on a loss a player should be able to move up. A player that plays extremely well, but is placed on a team that refuses to work together (which is often today) should not be punished... That system does not base movement on player skill, but bases movement on the skill of the other 11 players. I disagree with this.

I also disagree with resetting the current Teirs... I started with a potato for a computer and had to grind my way to where I am now. Nonetheless, I can regrind my way back here and take most offense at punishing those who perform well because others could not.

#262 OmgKllL

    Member

  • PipPip
  • The Aggressor
  • 34 posts
  • Twitch: Link
  • LocationMilan - Italy

Posted 05 June 2020 - 11:02 AM

just hide the tier bar! :D

#263 SoccerMomsAtNight

    Member

  • Pip
  • Shredder
  • 15 posts

Posted 05 June 2020 - 11:03 AM

It seems super dependent on winning and losing. What if you did something like this instead:

MATCH SCORE
Match Score: 0-99 goes down in PSR by -2
Match Score: 100-199 goes down in PSR by -1
Match Score: 200-299 PSR does not change
Match Score: 300-399 goes up in PSR by +1
Match Score: 401+ goes up in PSR by +2

WIN/LOSE
Win: goes up in PSR by +1
Lose: goes down in PSR by -1



Or you could mirror the old losing table instead of mirroring the old winning table and end up with something like this

Player LOSES:
Match Score: 0-100 goes down in PSR by -2
Match Score: 101-250 goes down in PSR by -1
Match Score: 251-400 does not move.
Match Score: 401+ goes up in PSR by +1

Player WINS:
Match Score: 0-100 goes down in PSR by -1
Match Score: 101-250 does not move
Match Score: 251-400 goes up in PSR by +1
Match Score: 401+ goes up in PSR by +2


Just my two cents worth.

Edited by SoccerMomsAtNight, 05 June 2020 - 11:04 AM.


#264 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 05 June 2020 - 11:08 AM

View PostBrom96, on 05 June 2020 - 09:10 AM, said:


My issue is this - a single player can have limited to no effect on the match. Again, it is team that can drag a single player down through its incompetence. I will be honest, while I do see some games where there is coordination in randomly created team, as the teams are in this game, in most cases there is zero or almost zero communication and coordination is mostly instinctive -based on the experience and expectation that other will have the same experience, i.e. they played before and have seen what goes and what doesn't on each of the maps...

Well, good luck with that. No matter the number of matches, you will see players who have good personal score being damaged by teams they play with and the only way to ensure you have a good score will be to team up with someone, creating groups and clans. In the end, the game will strongly favor playing in groups, not leaving the place for single players, which will, I fear, make game more unfriendly for new players. Past month had shown me that group people usually are not better then the average single player, do not communicate over game voip, or do it rarely, and have poor understanding of the tactics required for the maps. Or in general. On the other end of the spectrum are the groups that leave the rest of the team to die and do not interact with them in any manner, playing their own game, whether the eventual outcome is a loss or a win.

So, we end with the team ratings, not personal ones, because the personal achievement in the match is extremely outweighed by playing of the others. Do not call it player skill rating because it ain't.


Missing the forest for the trees. One match is irrelevant. What matters is whether or not in the long term a player is driving their team to win. So in a solo QP environment having WLR have a huge influence on PSR is fine.

In the current environment with groups in the queue this is cancelled out somewhat. Over time whether an individual's teams win more than they lose should still come through, but having groups in the queue adds a whole lot of noise to that signal.

Look, long-term if you have a WLR around 1 you are not statistically doing anything above and beyond an average player to help your team win. If you have a sub 1 WLR then you are on average a detriment to your team. If you have a WLR above 1 then on average you are doing more to drive wins than the average player. It's really simple.

#265 MeterH1

    Member

  • Pip
  • Mercenary Rank 2
  • 10 posts

Posted 05 June 2020 - 11:11 AM

View PostSlothasaurus, on 05 June 2020 - 09:49 AM, said:

I don't think W/L outcome should be weighted so heavily vs one's performance. Yes winning or losing should still count and have value but just not as much. Putting skill and all other things aside I have seen quite a few matches where there might be 1-2 AFK players and a disconnect. That puts one team in quite a hole to start the match at. A pilot on that team could go out and play really good in the loss, but under the new system proposed they will only end up ranking-wise right where they are at. They should not be penalized for things out of their control like other players being AFK or disconnects.


I absolutely agree with this.. If the point is to tier people based on skill, then this needs to be based on skill, not the luck of landing a team that has afks and disconnects.

View PostSoccerMomsAtNight, on 05 June 2020 - 11:03 AM, said:

It seems super dependent on winning and losing. What if you did something like this instead:

MATCH SCORE
Match Score: 0-99 goes down in PSR by -2
Match Score: 100-199 goes down in PSR by -1
Match Score: 200-299 PSR does not change
Match Score: 300-399 goes up in PSR by +1
Match Score: 401+ goes up in PSR by +2

WIN/LOSE
Win: goes up in PSR by +1
Lose: goes down in PSR by -1



Or you could mirror the old losing table instead of mirroring the old winning table and end up with something like this

Player LOSES:
Match Score: 0-100 goes down in PSR by -2
Match Score: 101-250 goes down in PSR by -1
Match Score: 251-400 does not move.
Match Score: 401+ goes up in PSR by +1

Player WINS:
Match Score: 0-100 goes down in PSR by -1
Match Score: 101-250 does not move
Match Score: 251-400 goes up in PSR by +1
Match Score: 401+ goes up in PSR by +2


Just my two cents worth.


I'd be happy with either of those suggestions -- The trick is getting PGI to meet us somewhere in the middle rather then going from one bad extreme (the current system) to another horrible extreme (the proposed changes for Jun 9th)

#266 ERSmurf

    Member

  • PipPip
  • Knight Errant
  • 20 posts

Posted 05 June 2020 - 11:12 AM

View PostVonBruinwald, on 05 June 2020 - 10:08 AM, said:


You're looking at the wrong numbers, most players will have outliers. What you need to look as is the standard deviations.

~70% of your games will occur within that first deviation.


@VonBruinwald: I must have bothched the math, because in a sample of 23, this IS not true. If I did not make errors, the match average is roughly 165 (actually 164 and very high decimals)... and the first deviation is about 77,79... out of the samples matches, only SIX fall within that range, nowhere near 70%... *puzzled*

#267 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 05 June 2020 - 11:12 AM

View PostSnowhawk, on 05 June 2020 - 09:43 AM, said:


Yes... that's true for games like Tennis (1 v1 ) where your personal Performance is indeed crucial. Here in Mwo you are one of 12 Players.... and unfortunately there are Epic-fail-Teams which can not even be saved by Proton or Bowser. This Teams will soon decide your psr.


Sure, a single match can be turbo-thrown by your team. But if after 1,000 matches your WLR is around 1.00 that is a fairly accurate signal about your contributions to matches. Again, any PSR should not be primarily concerned about getting every single player's performance perfectly assessed for each individual match (such a measure would require an extensive and likely pointless analysis). WLR is a decent quick and simple way to get an overall picture of someone's contributions.

#268 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 05 June 2020 - 11:16 AM

View PostERSmurf, on 05 June 2020 - 11:12 AM, said:

@VonBruinwald: I must have bothched the math, because in a sample of 23, this IS not true. If I did not make errors, the match average is roughly 165 (actually 164 and very high decimals)... and the first deviation is about 77,79... out of the samples matches, only SIX fall within that range, nowhere near 70%... *puzzled*


You need a bigger n. 23 matches is simply too small of a number to do much with.

#269 Negat1ve Nancy

    Member

  • PipPip
  • 32 posts

Posted 05 June 2020 - 11:19 AM

Of course a group player would say that win/loss is the way to measure skill in a queue with solos and groups.

smh


Old Match Maker and the new Match maker

Posted Image

Edited by Negat1ve Nancy, 05 June 2020 - 11:20 AM.


#270 MeterH1

    Member

  • Pip
  • Mercenary Rank 2
  • 10 posts

Posted 05 June 2020 - 11:19 AM

On another though... There should be some consideration for a light pilot vs assault pilot.. Im largely an assault pilot, you will rarely see me in a light... But Lights should move up with less of a match score just because A 200 Match score in a light is truly the equivalent of a 400 in a assault.

#271 Firefox54

    Member

  • PipPip
  • 21 posts

Posted 05 June 2020 - 11:22 AM

View PostKamikaze Viking, on 04 June 2020 - 08:43 PM, said:


Posted Image


^^^ this!

This would be a fine solution ... consistent with PGI ... except based on rank not match score, and in this model you can increase your PSR if you lose and decrease if you win (which some seem to like ... as I do) ... if you made it wins are + or =, and loss are - or =, I'd like it less, but it would still be ***zero-sum***

For those who think people will farm to go up in rank i say: 1) that's probably a small portion of the players; 2) those players that farm just to go up are going to run into the better players that will keep them progressing further ... in the end, they'll be capped from moving any higher because they haven't gotten any better, and 3) it doesn't seem like most of those posting care about their Tier ... I think the farming concern is a small issue ... I think we all want a better match making system based on a system that won't artificially move players up into Tiers they don't belong.

The proposed system can still cause that to happen ... this approach (or something like it) is truly zero-sum.

Also ... if the system is truly zero-sum, use the PSR scores to do the matchmaking and forget using the Tiers altogether ... they're meaningless anyway ...

[... and let's not forget that W/L is in the Match score already ...]

Edited by Firefox54, 05 June 2020 - 11:25 AM.


#272 Major0103

    Member

  • Pip
  • 14 posts

Posted 05 June 2020 - 11:22 AM

I've read only through a few first pages of this thread, so maybe what I want to say has been already stated. However, I recon, there are two options for PSR:
  • prize win-lose ratio (even good performance in a losing team makes you go down, poor performance on winning team elevates you anyway)
  • prize individual contribution (good performance makes you go up no matter the lose, poor performance is punished even if your team wins)
Both of them are correlated with the player skill and over a large pool of matches they will drive the player into his adequate tier. However, there is a possibly huge difference in fluidity and how fast it will happen.


Option 1 is the slower, more inert measure. From your individual perspective, the performance of your teammates is some random variable. Your record of wins and loses reflects mostly this randomness, not your skill, as the performance of the others is usually decisive for win/lose. To extract your individual performance, you need to average it out, which kills randomness. But the rule of thumb is the more randomness in a sample (higher variation) the more data (in this case: matches) you need to calculate the average accurately. Otherwise it gets erratic. Once you establish it, it will also take many future matches to change this average.
On the other hand, option 2 is less affected by the performance of your teammates, so you can average out fewer matches (possibly 10x less?) to get the desired accuracy of the average. Thus, in option 2 your PSR could be more fluid and reflect your current rather than general performance.

Edited by Major0103, 05 June 2020 - 11:26 AM.


#273 VonBruinwald

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Undisputed
  • The Undisputed
  • 3,460 posts
  • LocationRandis IV

Posted 05 June 2020 - 11:28 AM

View PostERSmurf, on 05 June 2020 - 11:12 AM, said:

@VonBruinwald: I must have bothched the math, because in a sample of 23, this IS not true. If I did not make errors, the match average is roughly 165 (actually 164 and very high decimals)... and the first deviation is about 77,79... out of the samples matches, only SIX fall within that range, nowhere near 70%... *puzzled*


Post all the individual match scores. Either you're doing something wrong or you're a really oddball player.

Brauer is right in that you've got a small sample size. But still.

Edited by VonBruinwald, 05 June 2020 - 11:29 AM.


#274 OneTeamPlayer

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • 399 posts

Posted 05 June 2020 - 11:31 AM

View PostNegat1ve Nancy, on 05 June 2020 - 11:19 AM, said:

Of course a group player would say that win/loss is the way to measure skill in a queue with solos and groups.

smh


Old Match Maker and the new Match maker

Posted Image


I'm honored as mimicry is the height of flattery, but in an internet literally full of options are you going to keep just literally copying every image i post but with a different message attached?

Posted Image




I'm going to go ahead and assume you're far more creative than that, don't sell yourself short.

Also, i'm not only not a group player i'm an exclusively solo player (with the stats to match). I'm not top tier anything and I just got a decent computer, though my internet is shared and thus hits me with 300 ping at random times of the day.

If i drop in tier, good that's where i belong. If my tier is raised, so be it.

You're not responding to any of my points, by the way.

Do you actually believe the conversation on here is going to cause PGI to revert the changes they've decided upon and go with one of the formulas written here that go directly contrary to what they believe i the problem with the system?

Cause what i've been saying is that that is not my experience with PGI's decision making process in any way.

Edited by OneTeamPlayer, 05 June 2020 - 11:33 AM.


#275 Thorn Hallis

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • 5,902 posts
  • LocationUnited States of Paranoia

Posted 05 June 2020 - 11:31 AM

So when will the reset happen?

#276 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 05 June 2020 - 11:32 AM

View PostNegat1ve Nancy, on 05 June 2020 - 11:19 AM, said:

Of course a group player would say that win/loss is the way to measure skill in a queue with solos and groups.

smh


Old Match Maker and the new Match maker

Posted Image


Well the new meta is to find two to three strong players to group up with.

Also, over time WLR should tell all anyway. My WLR does emphatically say that I am playing in groups, that's obvious. Even as a solo having a WLR above 1 means you are helping your team win more than the average player.

#277 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 05 June 2020 - 11:33 AM

View PostBrauer, on 05 June 2020 - 11:08 AM, said:


Missing the forest for the trees. One match is irrelevant. What matters is whether or not in the long term a player is driving their team to win. So in a solo QP environment having WLR have a huge influence on PSR is fine.

In the current environment with groups in the queue this is cancelled out somewhat. Over time whether an individual's teams win more than they lose should still come through, but having groups in the queue adds a whole lot of noise to that signal.



Not sure you can say the signal should still come through with groups in the queue. Precisely because they are a non-random skill grouping. This should have been done long before groups and solos were merged. I'm not sure it can be done accurately now.

Regardless, the system that Paul proposes is still broken in that it will drive players into a Tier 1 or Tier 5 grouping with little in between.

#278 ERSmurf

    Member

  • PipPip
  • Knight Errant
  • 20 posts

Posted 05 June 2020 - 11:35 AM

View PostVonBruinwald, on 05 June 2020 - 11:28 AM, said:


Post all the individual match scores. Either you're doing something wrong or you're a really oddball player.

Brauer is right in that you've got a small sample size. But still.


@Brauer, VonBruinwald: Thank you very much both... and here is the sequence: 94, 198, 189, 77, 178, 152, 72 129, 164, 127, 237, 117, 346, 269, 87, 200, 269, 307, 95, 94, 92, 81 and 214.

#279 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 05 June 2020 - 11:38 AM

View PostAnomalocaris, on 05 June 2020 - 11:33 AM, said:


Not sure you can say the signal should still come through with groups in the queue. Precisely because they are a non-random skill grouping. This should have been done long before groups and solos were merged. I'm not sure it can be done accurately now.

Regardless, the system that Paul proposes is still broken in that it will drive players into a Tier 1 or Tier 5 grouping with little in between.


I think with enough drops the signal will come through. But it is certainly an issue, and I don't disagree with the concerns you're raising. From what I can see people worried about WLR being involved are mostly either 1) coming up with one match where they got high damage and kills and lost, or 2) have not been driving wins and thus feel WLR doesn't recognize their contributions although objectively if you aren't driving wins over a large number of matches you aren't playing that well.

As you know I think having groups in QP is stupid, but while they're here I'll farm away and have some laughs with friends.

#280 Negat1ve Nancy

    Member

  • PipPip
  • 32 posts

Posted 05 June 2020 - 11:41 AM

View PostOneTeamPlayer, on 05 June 2020 - 11:31 AM, said:


I'm honored as mimicry is the height of flattery, but in an internet literally full of options are you going to keep just literally copying every image i post but with a different message attached?

Posted Image





I'm going to go ahead and assume you're far more creative than that, don't sell yourself short.

Also, i'm not only not a group player i'm an exclusively solo player (with the stats to match). I'm not top tier anything and I just got a decent computer, though my internet is shared and thus hits me with 300 ping at random times of the day.

If i drop in tier, good that's where i belong. If my tier is raised, so be it.

You're not responding to any of my points, by the way.

Do you actually believe the conversation on here is going to cause PGI to revert the changes they've decided upon and go with one of the formulas written here that go directly contrary to what they believe i the problem with the system?

Cause what i've been saying is that that is not my experience with PGI's decision making process in any way.


I like the game, have to saved it from stupeed where possible.





7 user(s) are reading this topic

0 members, 7 guests, 0 anonymous users