Jump to content

Psr Update And Hold On Patch.


713 replies to this topic

#101 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,262 posts

Posted 08 June 2020 - 10:36 PM

View PostKamikaze Viking, on 08 June 2020 - 10:22 PM, said:

Fear not, TOP MEN are working on the problem as we speak

What MM is? It's not absolute rating or leaderboard, like AvgMS is. It's very simple. MM gets players with closest possible MMRs for your match. If you're better, than them, then your rating should rise, so you'd get better enemies/teammates next time, so it would be harder challenge to stay competitive in this conditions, as better enemies are harder to kill and better teammates pull duvet on themselves more. And if you're worse, then opposite should happen. At the end you should get both enemies and teammates with as close skill level, as possible, as it results in the most fun games. As simple, as that. And AvgMS is bad for this purpose. As 500MS against complete noobs and 500MS against tryhards - are two completely different things.

Edited by MrMadguy, 08 June 2020 - 10:39 PM.


#102 Bistrorider

    Member

  • PipPipPipPipPipPip
  • The Howl
  • The Howl
  • 273 posts

Posted 08 June 2020 - 10:42 PM

View PostKamikaze Viking, on 08 June 2020 - 10:22 PM, said:

Fear not, TOP MEN are working on the problem as we speak


Any example how it should be done? Like with the numbers?

#103 Jochi Kondur

    Member

  • PipPipPip
  • Shredder
  • 66 posts

Posted 08 June 2020 - 10:48 PM

View PostZanotam, on 08 June 2020 - 05:46 PM, said:

Here is some relevant reading btw for those who think we shouldn't use mostly win/loss and insist on Match Score:

https://www.microsof...ranking-system/



This may not be a completely solved problem, but the fact it should be heavily based in Win/Loss isn't just a 'feeling' or something you can 'disagree with' with any real evidence to back that up.

Therefore I highly support the previously given W/L to determine positive/negative and matchscore to determine amount of positive/negative change in a player's skill ranking although I wouldn't mind if it changed to be similar to this one where say the bottom 2 players of the winning team gain nothing and the top 2 players of the losing team lose nothing and then scale up/down from there for the other players.


There is a revised version called TrueSkill 2 - TrueSkill2 predicts historical match outcomes with 68% accuracy, compared to 52% accuracy for TrueSkill.

#104 Kamikaze Viking

    Member

  • PipPipPipPipPipPip
  • 384 posts
  • LocationStay on Topic... STAY ON TOPIC!!!

Posted 08 June 2020 - 10:48 PM

Bistrorider we are working on a coordinated response - I guess we could be called 'Gulag II: Psr team'.
We are discussing PSR and MS separately.
People are working on simulators and experimenting with systems and trying to test using real world data examples.

More details once we come to a consensus.

#105 _Rorschach_

    Member

  • PipPipPipPipPip
  • Veteran Founder
  • Veteran Founder
  • 128 posts

Posted 08 June 2020 - 10:54 PM

View PostKamikaze Viking, on 08 June 2020 - 04:59 PM, said:

Thankyou Russ and Paul for the open communication on this topic.

I think I've hit my limit on how I can affect this discussion with my most recent proposal.
https://mwomercs.com...ost__p__6336540



In my opinion from what Ive seen in recent discussion here and on reddit, Dogmeat and Decency have the best grasp on Matchmaking in general AND the context of how it should work in MWO.

I have also seen very binary arguments about the importance of personal skill vs win loss. In my opinion they are BOTH a factor that must be considered.

And as Russ has said, all of us here are guessing at existing systems. I would really really like it if the Engineer involved would be allowed to come and discuss this with us directly.

To anyone proposing solutions. remember the K.I.S.S. Principle, and consider the work required to implement your solution. Perfect systems are great, but we WILL have to make compromises, its the reality of things.

And I would be fine with addressing PSR now and then letting it settle for a month before addressing Match Score calculations. We have seen in the past that too many simultaneous changes make it hard to understand the effects and what change caused what effect.


I think with the right weight for the overall win and "complete objectives" for MS it'll be possible to get the best of both worlds. I can see the argument that winning needs to be incentivized, so most plus PSR should end up on the winning team size. I can also see the argument that PSR should be based on individual rather than team performance, as we want individual players to move to their correct tiers, so they get the correct matchups.
By ranking all 24 players regardless of win/loss it at first glance would seem like prioritizing objectives/the win is disincentivized. But since both have a positive influence on match score we actually can get exactly that effect, and as a community we just have to agree how big of a factor it should be.
Personally I think that with fair matchups win/loss should be close to 1:1 for everyone but the few players that are just so good that they carry even against the 90 to 99%ers, but you can't do anything about that anyway. So win/loss shouldn't be the overwhelming factor as it will balance out over enough matches anyway, just enough to encourage going for the win over exclusively trying to dish out damage.

Overall, even if we get the factors imperfect, the zero sum nature of the system will be a great improvement over old PSR wrt to getting each tier populated correctly; and hopefully both lead to the MM needing the release valves far less often and casual/new players to get more enjoyable matches and keep playing the game. If the new player experience is improved as well (you completed your first/second/third/fourth/fifth/... cadet match, here's a free light mech/equipment for a good build on that light mech/cbills/GSP/MC/medium mech/equipment for it/... rather than just cbills) we may get to a point where we have sustained player number growth. *optimist-mode engaged*

#106 OmgKllL

    Member

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

Posted 08 June 2020 - 11:04 PM

Good job PGI for the community feedback, you are doing a great job with the psr adjustment.

Maybe we'll see this game rise again in the near future, it's what I hope for.

View Post_Rorschach_, on 08 June 2020 - 10:54 PM, said:

...
Overall, even if we get the factors imperfect, the zero sum nature of the system will be a great improvement over old PSR wrt to getting each tier populated correctly; and hopefully both lead to the MM needing the release valves far less often and casual/new players to get more enjoyable matches and keep playing the game. If the new player experience is improved as well (you completed your first/second/third/fourth/fifth/... cadet match, here's a free light mech/equipment for a good build on that light mech/cbills/GSP/MC/medium mech/equipment for it/... rather than just cbills) we may get to a point where we have sustained player number growth. *optimist-mode engaged*

I totally agree, maybe one from the meta considering there are a lot of meta loadouts which are very easy to play.

I've just brought my brother and my nephew in the game and they are very thriled with it but if they hadn't have me (who's experience is very relative) they would had quitted the game after 2/3 matches.

Reducing the cadet cbills, deleting the trial mech system and adding a pool of 2/3 mechs per category (you can choose only one of them at the beginning and one at the end of the cadet period), fully equipped and skilled, will improve significantly the new player experience imho.

Edited by OmgKllL, 08 June 2020 - 11:07 PM.


#107 Jochi Kondur

    Member

  • PipPipPip
  • Shredder
  • 66 posts

Posted 08 June 2020 - 11:24 PM

View PostKamikaze Viking, on 08 June 2020 - 10:22 PM, said:


Fear not, TOP MEN are working on the problem as we speak


Please help to remind those TOP MEN that the often cited Jarl's charts on Average Match Scores and W/L Ratio are both positively skewed distributions but were being interpreted as though they were symmetrically distributed (mode=median=mean) when in fact the median and mean are pushed to the right (i.e. mode < median < mean). The cited charts only give the meaningful fact that they are positively skewed distributions with a readily apparent mode value. However, critical information on the median, means (arithmetic mean, geometric mean and harmonic mean), standard deviation, coefficients of skewness, kurtosis, etc were missing.

#108 cougurt

    Member

  • PipPipPipPipPipPipPip
  • CS 2023 Silver Champ
  • CS 2023 Silver Champ
  • 691 posts

Posted 08 June 2020 - 11:36 PM

View PostTeknomancer, on 08 June 2020 - 10:36 PM, said:


A role warfare player who won the match by a base cap would say they absolutely contributed to the win.

sure, but you aren't going to win matches with any regularity by base capping. the vast majority of matches are decided almost entirely by damage and kills. either way, if what you're doing actually works then it will be reflected in your WLR.

View PostTeknomancer, on 08 June 2020 - 10:36 PM, said:

And that's the thing, you are right that it's wins and match scores, but your match score is determined by a whole lot of stats, including role warfare and objectives, not to mention damage and kills. Match score also has your bonus for winning. That's the question here: what about match score shows your performance? Which scoring factors are most important, of those many things listed at the start of this thread that go into calculating your match score? Does the current algorithm work or does it need to be tweaked?

the most important factors are damage, kills, and winning, hence why they're weighted the most heavily. there may be some improvements that could be made to the match score formula, but overall i'd say it does its job well enough.

#109 rascje

    Member

  • PipPipPipPipPipPip
  • The Privateer
  • The Privateer
  • 455 posts

Posted 08 June 2020 - 11:48 PM

Thanks to all the tips and suggestions from the players.

#110 Skipmagnet

    Member

  • PipPipPipPipPipPip
  • Pack Leader
  • Pack Leader
  • 230 posts

Posted 08 June 2020 - 11:56 PM

View PostKamikaze Viking, on 08 June 2020 - 10:22 PM, said:


Fear not, TOP MEN are working on the problem as we speak


I'd feel a lot better about this if said TOP MEN did not hold the bottom 95% of the player base in absolute contempt. I mean, why should I trust that they're going to come up with something equitable when they have said, at length, that they would rather players like me would go play a different game?

Edited by nocomfort, 09 June 2020 - 12:00 AM.


#111 Kamikaze Viking

    Member

  • PipPipPipPipPipPip
  • 384 posts
  • LocationStay on Topic... STAY ON TOPIC!!!

Posted 09 June 2020 - 12:07 AM

Im sorry you feel like that @nocomfort.

Hence we are even discussing the correct way to present our plan so as to openly show you that our intent is to measure people correctly and hence create a best effort scenario where ALL players should get the most enjoyable and balanced games.

#112 Jochi Kondur

    Member

  • PipPipPip
  • Shredder
  • 66 posts

Posted 09 June 2020 - 12:21 AM

View PostKamikaze Viking, on 09 June 2020 - 12:07 AM, said:

Im sorry you feel like that @nocomfort.

Hence we are even discussing the correct way to present our plan so as to openly show you that our intent is to measure people correctly and hence create a best effort scenario where ALL players should get the most enjoyable and balanced games.


May I also suggest that the TOP MEN team work with Scurro to log transform or square-root transform all asymmetrical charts (i.e. charts displaying skewness) so that any public facing charts in the said presentation are the more easily understood symmetrically distributed "bell curve" charts. Thanks!

#113 ZortPointNarf

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 261 posts
  • LocationIsle of Man

Posted 09 June 2020 - 12:23 AM

Just a very quick one, would it not make sense to have a PSR per weight class, maybe even split between Clan and Sphere as well.
I believe that may alleviate some of the problems of assaults dropping into a match that really shouldn't be there and almost immediately you are down 200 tonnes of fire power and armour.

Final note, I am very much in favour of a reset irrespective of the changes made. The sense of progression from the start was fun, but being capped at T1, jsut through sheer volume of play is no fun.

Final final note, good luck :)

#114 Akenatum Malthus

    Member

  • PipPipPip
  • Legendary Founder
  • Legendary Founder
  • 65 posts
  • LocationSydney, Australia

Posted 09 June 2020 - 12:40 AM

Long time without posting... and Have only read page1.

I would pair off the players in a match highest match score wining team to lowest match score loosing team, 2nd highest winning to 2nd lowest loosing, and so on until you get to lowest on the winning team and highest on the loosing team.. Then I would compare match scores and make PSR Movement based on the difference in rating, because being able to Win or lose a match should also have a relevant Value then add a +1 or -1 modify for someone weather or not someone won or loss the match. (exact values to be determined)

Example:
0-50 match score difference between players 0PSR Change base to both players (+1PSR to player who's team won, -1Psr to Player who's team lost)
51-100 match score difference between players +1psr to player with higher Score/-1 tp player with Lower score Base (+1PSR to player who's team won, -1Psr to Player who's team lost)
101-200 match score Difference +2/-2 between players (+1PSR to player who's team won, -1Psr to Player who's team lost)
201+ match score Difference +3/-3 between players (+1PSR to player who's team won, -1Psr to Player who's team lost)

What these values should ultimately be I'm not sure... but in games with blow outs this should make bigger movements between players, and tight close games have less movement. This should allow better players to climb faster with worse players to drop faster have the reset to everyone at the midpoint of Tier 3 and players will settle into their respective tiers quickly.

#115 Robinson Crusher

    Member

  • PipPipPipPipPip
  • Bad Company
  • Bad Company
  • 129 posts

Posted 09 June 2020 - 12:52 AM

View PostTesunie, on 08 June 2020 - 04:55 PM, said:

It would currently be an easy experiment to track... Play matches, screenshot, see who would move up and down by side... Does it look reasonable or not?

I would like to see more "average performance" be "no gain" personally. But as long as W/L doesn't become the main focus of the new system, I think I could be happier with it. I currently don't like how much W/L impacts your PSR changes at the moment.


Posted Image First, my compliments to the devs for putting this forum issue where I would see it logging into the game, without intending to go to the forums. I suspect some of the game's past problems could have been avoided with similar treatment and look forward to seeing this sort of thing continue in the future...

Tetsunie, I haven't been following this issue, so I'm curious as to why you want less impact by Win/Loss. My first reaction was that eliminating the impact of W/L would punish people (like me) who often choose to die for the team. W/L is the only stat I've ever cared about. You play for fun, sure, but winning fun is more fun than losing fun. Personally I get more fun from a well played match that sees my team win even if it means sacrificing my other stats, and I often do as a result. Why should that be discouraged?


- Two points about PSR:

1) I'm very good with some mechs and totally suck with others. My PSR goes up and down depending on whether I am taking it easy or trying to get good. It would be very interesting if the match maker could keep PSR stats for each of a players mechs.

2) If the problem is that imbalanced matches drive away newer players, is not the real problem that good players are not evenly distributed between the teams? My impression is that a match is filled out one team at a time. So for a tier one match one team gets most of the available tier one players, and the other team gets the rest.

Anyone know if this is accurate, or an illusion, because if true then it's the root problem.

Edited by Robinson Crusher, 09 June 2020 - 12:54 AM.


#116 Voxsera Hazen

    Member

  • PipPip
  • Legendary Founder
  • Legendary Founder
  • 25 posts
  • LocationSingapore

Posted 09 June 2020 - 12:53 AM

Good to know feedback was actually read and considered. The original idea was kinda crap, to say the least. Opinion on the PSR system is simple ;

1: Ignore win/loss in distributiong PSR changes

2: Match score ranked in the match from top to bottom, with the top gaining most PSR, bottom losing the most. If wanted, leave some room in the middle for players that don't move at all, being in the middle 2/4/6 players.

3: Adjust numbers for PSR rise/fall as whatever, top can be +45 or +20, doesn't matter, since there will be a mirror on the bottom half getting a PSR of -45 or -20.

4: ???

5: Everyone pews.


It's already pretty much what was proposed in the new post, with the example being 156 points as the total number for PSR changes. it's simple and elegant. Match score is already a pretty good indicator of how well one plays, for now at least.

I'm absolutely AGAINST tying PSR changes to win/loss, as that's just dumb, sort it on PURELY match score. So that's a no from me and anyone who tries to say so otherwise. Winning should NEVER automatically mean a PSR gain.

Here, have an example ;

Do 30 match score,47 damage, 0 kills, 1 assist, be on winning team and pretty obvious your are getting carried? Your PSR should go down as you did jack **** and you will rank in the bottom when it comes to all of the 24 player's match score.

Do 717 match score, 1339 damage, 5 skills, 5KMDDs, 4 assists, match ends in 11/12. Hey pilot tried tried to carry but didn't quite make it. PSR should go up - since the match score will indicate you did more than your team anyway, even with the win providing a bump in match score for the other side.

That last example? it was me, and I pulled out a screenshot for it too. Highest match score is me at 717, next being 578 on the winning team. The annies on the other side carried them to victory while the annie on my team and me tried, but just didn't quite make it. Both sides had ultra mechdads who couldn't do jack ****, on either the team who won or lost - in the new proposed system, will be bottom ranked and have their PSR go down.

Posted Image
*screenshot of the GREAT MEOWRAUDER 5D - where D stands for damage.
** PSR actually went up in that match too, which is based on the current system.

All in all, sums up with a quote from Ash


"A player that contributes nothing in a win should never, ever, be going up. They should go down. A player that tries to carry his teams *** off, 4-5 kills / 5-6KMDDs and 1400dmg match - should never be PSR neutral, ever."

#117 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,262 posts

Posted 09 June 2020 - 12:54 AM

View PostKamikaze Viking, on 08 June 2020 - 10:22 PM, said:

Fear not, TOP MEN are working on the problem as we speak

And also. Who are that TOP MEN? Tweaking game by top players for top players has always been bad idea. Because they don't even understand, how 99% of playerbase plays this game.

#118 Kakia

    Member

  • PipPip
  • Ace Of Spades
  • Ace Of Spades
  • 23 posts

Posted 09 June 2020 - 12:57 AM

I think the hold on the update is wise while we try to seek solutions on tweaking the PSR a little bit more. But honestly I think it's about as good as it's going to get...

The real problem is the low player base. I mean it's a great game. Nothing else like it. But I don't see ads in print, commercials on TV, or promotions online like for other games. I rarely see MWO mentioned except in a game review once in a blue moon. I see some advertisements for MW5 on Epic.

Why can't MWO be promoted or advertised openly. I see world of tanks, world of war ships, war thunder, and league of legends as targeted ads online or on occasion as a real TV commercial. Gees, make a movie out of this already like so many other popular franchises...but I think there in lies the rub about why it can't be done due to legalese agreements. Find another way. It's do or die.

A lot of people don't know about this game. If possible use the power of advertising. And if that isn't possible consider preloading agreements with PC OEMs to put on new computers. Give away the launcher with a powerful visual trailer so that the buyer can op in on installing the whole game. MW5 has a nice trailer and so did MW4 Vengance. You get the idea. This might be a way to get around a legalese dilemma. It's giving the game away, not advertising. Can work for both MW5 and MWO. Kill two birds with one stone!

Edited by Kakia, 09 June 2020 - 01:25 AM.


#119 Vellron2005

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The Blood-Eye
  • The Blood-Eye
  • 5,444 posts
  • LocationIn the mechbay, telling the techs to put extra LRM ammo on.

Posted 09 June 2020 - 01:05 AM

@Russ and Paul

Guys, I appreciate you trying and putting dev hours into this game, but I think that your effort right now is pointless.

What ever you do to the matchmaker, there simply isn't enough players still playing this game for your changes to make a meaningful impact on matchmaking.

You said that 50-100 people plays the game at best at any one time, with 1500 active players total? I can tell you from personal experience that when I log in and hit that QP button, I can play all night and see mostly the same player names.. If you get one set of players, you're pretty much stuck with 90% of them for the night.

So your MM changes won't have much effect. The population is too low to provide any significant match making where good players aren't gonna roflstomp bad players.

I know some of the veteran players still playing want you to do X or Y or Z to the MM.. That's just their own personal desires aimed to show e-peen and stroke their egos when players drop into a lower tier and they are confirmed as "Tier 1 gods".

But the truth is.. it's not gonna change anything. Not really.

So save your Dev hours and go take a good hard look at your finances, and start thinking about MWO2.

#120 -P U R E-

    Member

  • PipPipPip
  • The Predator
  • The Predator
  • 71 posts
  • Twitch: Link
  • LocationWarsaw

Posted 09 June 2020 - 01:19 AM

View PostMrMadguy, on 09 June 2020 - 12:54 AM, said:

And also. Who are that TOP MEN? Tweaking game by top players for top players has always been bad idea. Because they don't even understand, how 99% of playerbase plays this game.


PGI will tweak it - not players.
It is help in tweaking by ppl who not only play but also understand this game. And there is a difference.
Paraphrasing your post:
''You don't even understand, how 10% of playerbase plays this game''





5 user(s) are reading this topic

0 members, 5 guests, 0 anonymous users