Jump to content

Psr Update And Hold On Patch.


713 replies to this topic

#541 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 14 June 2020 - 03:41 AM

Also PGi guys dont forget the actual construction and applying the skill pool to both teams equally.

Do Part A right but forget the part B

Even if A is good (PSR reflects player skill more closely), the result will be (sheet) aka less than optimum unless both parts are looked at as a whole, as part of the MM process.

TLDR: Dont forget PSR gold is useless unless it spreads the skill pool to both sides to minimize any skill gap.
Edit: Especially important when population levels are low!

Deep breath

Edited by OZHomerOZ, 14 June 2020 - 03:50 AM.


#542 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 14 June 2020 - 03:54 AM

I wrote something about that topic earlier, its not expert but maybe a start or food for thought for some maths experts

View PostOZHomerOZ, on 08 June 2020 - 05:18 PM, said:


This is what imo may help with what you described. Posted Image

I just want to add that another thing that must be considered is how teams are constructed to avoid stacked teams, let me clarify. Example one side has a bunch of 99%ters and the other side none, we seen it before the merge too.

Lets say MM has available a list of 24 or more players ranked by PSR (using the new system) in the player pool. How MM constructs the two teams using the available players forms an integral part of the MM process and if ignored can un do all the good work being proposed.

So don't forget about that as MM is a system and all the parts gotta work together to give the intended result. Low skill gap games where each player is matched with a counter player as close as possible.

The below is my thoughts

Kinda like school yard games (maybe old school I dunno havn't been to school long time) where you have two captains in our case that would be the two highest PSR players in the available playa pool Player Ranked 1st and Player Ranked 2nd.

Lets call these teams, Team One and Team Two.
Now imo cause Player Ranked 1st is more skilled than Player Ranked 2nd MM places the next highest PSR player from the pool on Team Two and then repeats the process for Team One.

As Team Two got the higher PSR player last time, MM will swap and the next highest PSR player from the pool will be placed onTeam one first and then repeats the process for Team Two.

etc etc till you made two teams.

For groups match them with other similar groups or equal PSR solos, player for player.
Sure groups may still have a coms and coordination advantage, but similarly skilled solo'es put in what is effectively an ad hoc group should hold their own unlike the current situation.

TLDR: How teams are constructed is as important as zero sum and matchscore calculations.


All I can add is monitor the results and adjust settings until the requirements are met.
Logic is moi forte, maths I leave to calculators or really smart at maths people.

2nd Edit: This could be due not to match maker but the fact that Tier 1 at present consist of players from the 99.999% percentile down to less than 50% percentile ( I heard as low as the high thirties percentiles.)

Which the new psr proposal is meant to fix. Hi 5 for that!

Edited by OZHomerOZ, 14 June 2020 - 06:57 AM.


#543 Dogmeat1

    Member

  • PipPipPipPipPip
  • CS 2020 Gold Champ
  • CS 2020 Gold Champ
  • 123 posts
  • LocationButte Hold

Posted 14 June 2020 - 04:47 AM

Before I start I should say I understand the limitations of what PGI has asked from us in this thread. This thread is primarily about deciding the PSR gain/loss values for the new zero-sum system and for suggesting changes to the matchscore formula. However while things are still up in the air there are a some ideas I would like to bounce out there. This is mostly based on a idea I wrote up in 2017 but never bothered to post.


The problem of tiers and zero-sum

Tiers, in theory, seem like they would work well to enable fast and efficient matchmaking. The problem is that tiers quickly become massively unbalanced both in terms of population size and the skill gap between stronger and weaker players within the same tier. The upcoming PSR changes will reduce the skill gap somewhat but it will not solve the problem entirely. Furthermore with a zero-sum PSR system, because it’s unrealistic to expect a 1.0 WLR across the whole population, some players are going to infinitely gain or lose PSR. Therefore artificial caps or constant rebalancing of tiers would be needed to keep the populations balanced.


How about a pure WLR focused system?

There have been a number of good proposals recently and in past years, that have suggested simply using WLR as PSR (perhaps limited to their recent history). Many competitive games take this approach and it does actually work reasonably well. How well you play (i.e. your skill) will affect how often you win, so given enough games, if you put winning players against winning players and losing players vs losing players, eventually all will filter to a level that is appropriate for their skill level. Simply put if a system like this was implemented, it would improve matchmaking quality.

However is WLR the most accurate and effective way of judging player skill? Well if we analyse the stats of high level players on the leaderboards and Jarl’s list we can see a strong correlation between kills damage (as represented by matchscore) and high WLRs. The best players are always strong in all 3 areas. But its important to note that the players with the highest WLRs on each given team are not always the best players on those teams. A player with a good WLR but relatively poor kill and damage stats is far less competitive against a good player than someone who has an ok WLR but is strong in the other two indicators. Organized groups in group and soup queue are a heavy influence on winrates. WLR can also fluctuate a lot month-to-month, more so for many players than the other indicators and there are also many random factors in games beyond an individual player's control that affect it. Given enough games these factors will correct for themselves, but in the short-to-mid term, the process can be more frustrating and rage inducing than it needs to be.

So while I do believe a pure WLR system would work, I think there are other options worth considering too.


Average matchscore - a simple (non zero-sum) fix

One of the simplest fixes to the whole PSR issue would be to throw out the current zero-sum PSR plan and instead just make each player's recent average matchscore (originally I thought the past 3 months of data but from reading other people's posts probably the past 150-300 games would work better) their PSR. If matchscore was fixed so that it reflected wins, kills and damage appropriately, then right away you would have a PSR that would represent a given player’s ability to influence games more so than just their WLR.

Because the system only uses recent history, it would be quite fluid and players would quickly be matched in games as appropriate for their ability. It’s simple and efficient because if you play well you go up, if you play badly you go down. There is no need to mess around with how much PSR to increase/descrease or to worry about making sure other players move down. As long as the matchscore formula works, then average matchscore as PSR would work too.

How your team performs has a lesser impact on how you will be matched in the future, although wins still need to represented appropriately with the final score. No one likes to lose and its far less frustrating in a 12v12 game to be judged on how well you play as an individual, rather than purely just whether you won or not. Finally for most people, competing for points and the win is always going to be more fun than just going for the win alone. A win of course is a good thing, but over the years I see a lot more people getting excited about the kills or damage they have done, than just the win itself.

This of course all relies on fixing matchscore and that’s easier said than done. Currently there are 44 separate matchscore triggering conditions within the formula so it is far more complicated than it needs to be. However that doesn’t mean it can’t be done and right now matchscore isn’t a completely terrible indicator of skill. With sufficient tweaking and testing I think we could get it to a stage where the formula would produce acceptable results from a both matchmaking and gameplay perspectives.

But like I said at the beginning, completely replacing the PSR system (and the MM too) is beyond the scope of what has been asked by PGI here so I'll end my wall of text on this point.

#544 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 14 June 2020 - 05:35 AM

Edit: Win loss is good in a homogeneous environment, but we have soles and groups mixed! So I disagree.

What about using average match score in a given match as the anchor point for no change to PSR.
Players lower than the average in a given match recieve a reduction to PSR and those with a higher MS than the Average MS in that match receive a gain in PSR.

That way when a good team rolls and all the winning players gain a small gain PSR gain, its still relative to other players in that match.

Top 3 highest PSR players (Pick a number) gain PSR and the bottom 3 PSR players lose PSR (again pick a number, dont have to be 3).
This way players are ranked against players in the match.

Or use the global Average match score to rate players against the entire player pool.

Edited by OZHomerOZ, 14 June 2020 - 06:30 AM.


#545 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 14 June 2020 - 05:39 AM

View PostDogmeat1, on 14 June 2020 - 04:47 AM, said:


Finally for most people, competing for points and the win is always going to be more fun than just going for the win alone. A win of course is a good thing, but over the years I see a lot more people getting excited about the kills or damage they have done, than just the win itself.



Then their are others richer than Blake Posted Image that only care for the win and don't care for the space bucks etc.
Is caring for space bucks skill or is winning more representative of skill?

Edit: Also the as you say the so called "competing for points and the win is always going to be more fun than just going for the win alone "doesn't this competing for points" thing is easier when skill gap is greater.
Thus it will not be popular with those types of players. Who IMO had it too good for too long at the expense of others who left the game entirely and silently imo.
2nd Edit: which in imo is why we ain't World of Mechs material yet! Gotta stop the leaks, Wanna bay a mech pack gives PGi bucks short term, but leaky bath tub will kill its growth long term.

If a team mates loses a game for space bucks Im gonna report them for assisting the enemy cause space bucka were more important than winning TOS and stuff.

Besides you win more you get more space bucks and stuff.

2nd Edit: There's winning and farming which should be rewarded more?
imo its clear!

Its harder to get High MS on a loss consistently unless your good any way, just saying.
3rd Edit: Less hard if your grouped.

Edited by OZHomerOZ, 14 June 2020 - 06:36 AM.


#546 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 14 June 2020 - 06:45 AM

View PostDogmeat1, on 14 June 2020 - 04:47 AM, said:


This of course all relies on fixing matchscore and that’s easier said than done. Currently there are 44 separate matchscore triggering conditions within the formula so it is far more complicated than it needs to be. However that doesn’t mean it can’t be done and right now matchscore isn’t a completely terrible indicator of skill. With sufficient tweaking and testing I think we could get it to a stage where the formula would produce acceptable results from a both matchmaking and gameplay perspectives.



I thought the MS part is OK, after all, Jarls list is acceptably accurate (not perfect) and is derived from the existing data. And didn't Russ or Paul say to the effect: we just talking about a PSR rejig and not MS rejig but they gave us some data about MS triggers to help make the PSR better. In other words MS twerking is out of scope!

(Otherwise they would of given us the formula not just the triggers and in fact stated they won't give the formula due to concerns about gaming/exploiting/abusing it etc)

Please correct me if I am wrong!

Edited by OZHomerOZ, 14 June 2020 - 07:02 AM.


#547 Dionnsai

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 469 posts

Posted 14 June 2020 - 07:04 AM

View PostKurlon, on 08 June 2020 - 04:30 PM, said:

The community driven PSR project

In response to PGI’s announcement on this thread, the discord and reddit communities have decided to work together to try to find the best solution to this issue. At first this was a JGx led initiative but quickly expanded to include dozens of representatives from most of the major units, as well as a large number of voices from the solo community. After many hours exchanging and refining ideas, we believe we have come up with a solution that both meets PGI’s requirements and that will improve the gaming experience of the community as a whole.

Arbitrary fixed PSR gains/losses present a number of problems in terms of balance and fairness to all players involved in an individual match. Therefore our solution is to use scaling PSR gains/losses based on an individual player’s match score matchscore as compared to the team's average. The system uses the Win/Loss status to compute a 0 sum PSR shift that is scaled fairly to the pilot’s contribution in that match. The maths involved is no more complex than the current XP, C-bill and Matchscore calculations requiring minimal development time.

This system was developed by Jay Z with heavy input from the veteran community. A link to a full explanation is below;

https://docs.google....3LgMBQJ4hE/edit

And a spreadsheet simulating the system can be found here;

https://docs.google....dit?usp=sharing

In summary, this proposal uses simple arithmetic similar to existing calculations to assign fair and reasonable PSR shifts in accordance to pilot contribution to the match result. The aim of PSR is to segregate the playerbase on their respective match impact to give the Matchmaker the best possible information to create fair and interesting matches.

The version of this proposal you see presented is simplified from a more complicated model we have been testing which has more adjusting and weighting factors. Within the parameters set for this task, we believe this to be a straightforward implementation which will have community support. Finally, given there are only two values that control the maths, the long term management is simple and easy to hand over. If you have any questions, or concerns, please let us know as we have investigated this and numerous other solutions in great detail.

The matchscore formula itself does need quite a bit of adjusting. We plan on posting an update regarding that soon.

Finally during our discussions we also came up with 2 other proposals we strongly believe would be worth considering;

Kamikazi viking’s PSR lookup table https://mwomercs.com...05#entry6338105

Gagis’ PSR gains based on matchscore ratios https://mwomercs.com...st__p__6338091


do this stuff

#548 OneTeamPlayer

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • 399 posts

Posted 14 June 2020 - 07:09 AM

Does anyone honestly believe that in this stage of the game at this age "cbills or xp" is an incentive for more than 20% (generously) of the population?

Most people have all the mechs they want, money to buy consumables forever, money to impulse buy a mech, enough spare GXP or free Skill Points to instantly level that mech, and cosmetics out the wazoo from events and loot boxes.

There's not a just not a large population that is actively leveling at this point in the lifecycle.

#549 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 14 June 2020 - 07:18 AM

So for those 20% match making is vitally important for their game experience if PGI intend to get some coins from them considering the other 80% are EXPERIENCED players.
Coz upward biased Tier 1 system said so.

Edited by OZHomerOZ, 14 June 2020 - 07:20 AM.


#550 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 14 June 2020 - 07:30 AM

View PostHorseman, on 14 June 2020 - 01:26 AM, said:

I would call it an improvement rather than a cure. Anything better than what we have now will benefit the playerbase, the only question is by how much.


I think Jay Z gave an honest reply to my question. He thinks it will make a small improvement, but if people's life were at stake he's not confident in that assessment.

#551 Gozuri

    Member

  • PipPip
  • 25 posts

Posted 14 June 2020 - 09:15 AM

Can someone tell me whats wrong with only taking W/L into consideration and foregoing MS altogether? I mean, yeah, its nice to have a matchscore to show your contribution and all, but when it comes to Ratings, W/L will suffice. I know many say its player SKILL rating, but there is no objective way to measure a player's skill. Besides, what is the working definition of skill? From what I understand, at least in this context, it is a person’s ability to perform actions that increases the odds of winning. If this is what we’re going by, actually winning would be a good indication of skill and thus W/L should work as a pretty close approximation. More skilled = more wins.

If we were to take matchscore into consideration, we'd be competing against our teammates as much as our opponents, something I don't think should be encouraged in a game that is supposed to include teamwork.

If you need a system, just get the average ratings of both teams and determine potential gain/losses in ratings based on the difference.
E.g.

Red Avg psr: 1000 1010 990
Blue Avg psr: 1000 990 1010
R Potential chg: +10/-10 +8/-12 +12/-8
B Potential chg: +10/-10 +12/-8 +8/-12


Everyone on the same team will have the same potential gain/loss.
Winning against a higher skilled team is more rewarding and losing to them is less punishing.
Show team avg psr and potential gain/loss in the match lobby so players don't feel cheated.
If you go by the ‘strength of the team is its weakest link’ ideology, adjust it so that those with higher psr will experience smaller changes and vice versa. I don’t think this is necessary though.

Isn't this how psr normally works? Whats wrong with applying that here? Am I missing something?

If anyone feels bad about being singled out for not contributing due to low dmg or match score, well just remove that from the end game scoreboard and replace it with a badge/tag/symbol/whatever indicating those with top dmg, top ams missile destroy, top spotting assist, or whatever other category you like. It may not mean anything, but it will let players feel good about themselves. That's what games are about after all.

Edited by Gozuri, 14 June 2020 - 09:29 AM.


#552 East Indy

    Member

  • PipPipPipPipPipPipPipPip
  • The Hammer
  • The Hammer
  • 1,244 posts
  • LocationPacifica Training School, waiting for BakPhar shares to rise

Posted 14 June 2020 - 09:22 AM

View PostXiphias, on 13 June 2020 - 03:03 PM, said:

8) I ran a bunch of matches to get the final results of the system

How many is "a bunch"? I'm curious how long it takes to get that divergence.

Also — while I acknowledge that a feature of Jay Z's proposal is a "fair" approach to contribution, what happens to long-term simulations when a coefficient (or whatever) is added to make significant PSR shifts for, say, the top 4 and bottom 4 only, and very tiny shifts to the other 16 players? In other words, can the system assume that two-thirds of the players it's being fed average 175-250 matchscore and work to keep them in the middle, sending only the really good and really subpar folks to the extremes?

Edited by East Indy, 14 June 2020 - 09:22 AM.


#553 Gagis

    Member

  • PipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 1,731 posts

Posted 14 June 2020 - 09:40 AM

View PostGozuri, on 14 June 2020 - 09:15 AM, said:

Can someone tell me whats wrong with only taking W/L into consideration and foregoing MS altogether?

It makes people feel bad. For the sake of matchmaking there is absolutely nothing wrong with it.

But we are here exactly because the players, or at least those who are crazy enough to use these forums, loudly opposed any system that does not reward you for losing with a high match score, so PGI, and many of us working on these proposals, work on the assumption that MS has to be included and PSR has to reward losers who tried the hardest, even if it makes matchmaking less accurate.

#554 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 14 June 2020 - 09:49 AM

View PostGozuri, on 14 June 2020 - 09:15 AM, said:

Can someone tell me whats wrong with only taking W/L into consideration and foregoing MS altogether?


Players that play in groups prefer MS because it is blind to the advantages of playing in a group.

A 400 average match score SOLO player has around 2 WLR. A 400 average match score GROUP player can have 5-10 WLR. A MS based system can't tell the difference between the two, a WLR based system can.

#555 spannerturner

    Member

  • PipPip
  • The Spear
  • The Spear
  • 48 posts

Posted 14 June 2020 - 10:02 AM

View PostGagis, on 14 June 2020 - 09:40 AM, said:

It makes people feel bad. For the sake of matchmaking there is absolutely nothing wrong with it.

But we are here exactly because the players, or at least those who are crazy enough to use these forums, loudly opposed any system that does not reward you for losing with a high match score, so PGI, and many of us working on these proposals, work on the assumption that MS has to be included and PSR has to reward losers who tried the hardest, even if it makes matchmaking less accurate.


Pisses people off is what it does. Frustrates those who lose a match but had the highest MS of the match only to see that they lost PSR. Especially when they have no input as to the team they are randomly paired up with.

Base it solely on W/L and here is a trend I can see happen... "Uh oh, we're down by 4, might as well quit this match now and cue up for a new match since I am not going to get any reward for sticking it out to the end, even if I am carrying my team damage-wise..."

#556 Gozuri

    Member

  • PipPip
  • 25 posts

Posted 14 June 2020 - 10:04 AM

View PostGagis, on 14 June 2020 - 09:40 AM, said:

It makes people feel bad. For the sake of matchmaking there is absolutely nothing wrong with it.

But we are here exactly because the players, or at least those who are crazy enough to use these forums, loudly opposed any system that does not reward you for losing with a high match score, so PGI, and many of us working on these proposals, work on the assumption that MS has to be included and PSR has to reward losers who tried the hardest, even if it makes matchmaking less accurate.


That would work in a more causal PvE game where you are rewarded for trying. In a competitive PvP, a loss is a loss, doesn't matter how hard you tried. In a game of chess losing in 10 moves is no different from losing in 100.
If you wanna reward players for scoring high, give them cbills or something.

View PostNightbird, on 14 June 2020 - 09:49 AM, said:

Players that play in groups prefer MS because it is blind to the advantages of playing in a group.

A 400 average match score SOLO player has around 2 WLR. A 400 average match score GROUP player can have 5-10 WLR. A MS based system can't tell the difference between the two, a WLR based system can.


Simplify that so that a pleb like me can understand because I'm not seeing the problem.

#557 Gozuri

    Member

  • PipPip
  • 25 posts

Posted 14 June 2020 - 10:08 AM

View Postspannerturner, on 14 June 2020 - 10:02 AM, said:


Pisses people off is what it does. Frustrates those who lose a match but had the highest MS of the match only to see that they lost PSR. Especially when they have no input as to the team they are randomly paired up with.

Base it solely on W/L and here is a trend I can see happen... "Uh oh, we're down by 4, might as well quit this match now and cue up for a new match since I am not going to get any reward for sticking it out to the end, even if I am carrying my team damage-wise..."


If you're paired with a squad of weaker players, that also means the opponent will have a bunch of deadweight. Carry the game like you're meant to and carry it hard.

#558 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 14 June 2020 - 10:14 AM

View PostGozuri, on 14 June 2020 - 10:04 AM, said:

Simplify that so that a pleb like me can understand because I'm not seeing the problem.


Read it slowly and I promise you that you will understand it.

Edit.. sorry if I wasn't clear on what I was implying. The community proposal is put together by GROUPs, they're not going to propose a system that takes away their advantage.

Edited by Nightbird, 14 June 2020 - 10:24 AM.


#559 spannerturner

    Member

  • PipPip
  • The Spear
  • The Spear
  • 48 posts

Posted 14 June 2020 - 10:17 AM

Here's a real world example for you...

Let's take Mike Trout. Arguably one of the best baseball players currently. He's been on the LAA Angels since 2011. Let's now look at their record since then. Only one Divisional Title and a .508 W/L%. So, if we tie his "performance" to the Angels W/L%, is Trout worthy of being considered one of the best players of all time? Or is that argument made solely on his individual performance as a player?

If you want to tie Tier ranking to W/L, then fine. Call it WLR (Win/Loss Rating). Not PSR since it's not really based on skill.
PSR (Pilot Skill Rating) suggest that your skill as a player is what constitutes the make up of this stat.

#560 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 14 June 2020 - 10:27 AM

View Postspannerturner, on 14 June 2020 - 10:17 AM, said:

Here's a real world example for you...

Let's take Mike Trout. Arguably one of the best baseball players currently. He's been on the LAA Angels since 2011. Let's now look at their record since then. Only one Divisional Title and a .508 W/L%. So, if we tie his "performance" to the Angels W/L%, is Trout worthy of being considered one of the best players of all time? Or is that argument made solely on his individual performance as a player?

If you want to tie Tier ranking to W/L, then fine. Call it WLR (Win/Loss Rating). Not PSR since it's not really based on skill.
PSR (Pilot Skill Rating) suggest that your skill as a player is what constitutes the make up of this stat.


If Mike Trout played on a randomly generated team for each game, he would have a WLR over 3. The reason professional games have such strict team making rules is to create teams that are as evenly matched as possible i.e. to force WLR down to as close to 1 as possible to make the matches fun for the fans (and the players).

Who would go to a game if you can see beforehand one team has all the best players, and the other all the worst right? That's the problem with the MM, it is blind to the skill of players because Tiers doesn't predict skill, so we get really bad matches made.

Edited by Nightbird, 14 June 2020 - 10:30 AM.






2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users