83 replies to this topic

[General Solo]

Wraith of Shadow, on 07 June 2020 - 02:40 PM, said:

... because that's the actual goal of the mission you're on?

Goal is to win, each mode has several ways to achieve this.
Some ways are better than others.
The Goal is to WIN, what ever that takes

[RRAMIREZ]

Captain Caveman DE, on 07 June 2020 - 07:45 AM, said:

consider something like this (which is btw something 'zero sum'); just spitballed and an example.
it still pronounces a win over just matchscore-farming, but at least the people carrying the losing side still get something of it,
while the last 3 of the winning team still go down a little.

I'm sorry but i can't convince myself that there's always 3 people that didn't contribute in any way (and MS may be the problem) on winning side, whatever the match shape.
Or at least didn't contribute for their team to win more than the 3 "best" players on loosing team, always.


But as I already said, it won't prevent me from playing.

Edited by RRAMIREZ, 08 June 2020 - 01:27 AM.

[L1f3H4ck]

Ey, if the tweaks take time, let them take their time.

[Mochyn Pupur]

Three fantastic matches today - each one someone ran and managed to capture the base, no match scores over 80 on both sides - everyone loses on PSR (this on polar highlands), two other matches alpine and polar highlands lrm hell (system dropped into low teens fps because the sky was so heavily full of missiles), sixth match cored from behind with a weapons test fire, two more where everyone full on full nascar and got wolf packed in my assault. Yes, new PSR will work wonders; can I see my negative PSR score please . . . /sigh

[Snowhawk]

Captain Caveman DE, on 07 June 2020 - 07:45 AM, said:

consider something like this (which is btw something 'zero sum'); just spitballed and an example.
it still pronounces a win over just matchscore-farming, but at least the people carrying the losing side still get something of it,
while the last 3 of the winning team still go down a little.

oh and: people have to get rid of the idea that going-down is a punishment; elo/tier/whatever is just a number to sort performance, so a matchmaker can build a 'fair and balanced' match; that's all.
having a third of all games be stomps is an alarming number and it needs to go down;

Somehow I like this idea. Maybe the numbers need some fine tuning.... This System is "dynamic" and adapts to the current Team Performance and hardworking Players are rewarded even in a loss.

[Firefox54]

If they're sticking with the current (proposed) approach, it's a little disappointing that they didn't listen to the community (and people will state that's not a surprise) ... I was hoping that they would come up with some truly zero-sum ...

That being said, I kept track of the scores the last few times I played ... that was15 matches ... out of that, under the new (proposed) system, the difference between the sum of the points the winning team received, and the sum of the points the losing team received was, on average, -1.7 ... pretty close to zero with the spread ... or at least I can't say long term the difference is not going to be close to zero (the differences ranged from +16 to -29). On average the winning team gained 30.9 points and losing team lost 32.7 points (the median gain/loss was 32 for both teams).

So, while not zero-sum the system (albeit under a small sample size), seems to be close ... I'll probably keep track a little longer ... or maybe PGI will provide them.

[RRAMIREZ]

Snowhawk, on 08 June 2020 - 08:06 AM, said:

Somehow I like this idea. Maybe the numbers need some fine tuning.... This System is "dynamic" and adapts to the current Team Performance and hardworking Players are rewarded even in a loss.

It's not dynamic...
whatever the game shape or result, 3 on the winner team will go down (how well they did) and 3 on the loosing team will go up (how bad they did)

There's so much focus on "let's not punish the try harders" that the point is lost

Edited by RRAMIREZ, 08 June 2020 - 09:13 AM.

[Teknomancer]

Captain Caveman DE, on 07 June 2020 - 07:45 AM, said:

consider something like this (which is btw something 'zero sum'); just spitballed and an example.
it still pronounces a win over just matchscore-farming, but at least the people carrying the losing side still get something of it,
while the last 3 of the winning team still go down a little.

oh and: people have to get rid of the idea that going-down is a punishment; elo/tier/whatever is just a number to sort performance, so a matchmaker can build a 'fair and balanced' match; that's all.
having a third of all games be stomps is an alarming number and it needs to go down;

Now this is zero sum! A good suggestion...

However, this has a potential problem. You can't always guarantee that the best on the losing team should go up, and the worst on the winning should go down. What if the best of the losers was still out-scored by the worst of the winners? Then he goes up despite losing and being outplayed, while there is a winner who outplayed every loser but still lost PSR.

Match score is already weighted to give a bonus to winning. So long as that is true, then there is no need for separate target numbers for winners and losers for their PSR adjustment. And so long as match score is variable by a lot of evolving calculations, then assigning target scores for PSR changes is also flawed. 500 match score in one game might have been achieved with worse performance than a 400 score in a different game, all depending on who values what, the overall match flow, the team dynamics, game mode, tweaks to the formula from one patch to another, etc.

A fair zero sum algorithm using the existing match score system needs to measure performance across all 24 players at once, not just your team. The score is already weighted to give winners a boost. It has to allow for a PSR gain if your team lost but you outplayed the field, and for a PSR drop if you were significantly outplayed even though your team won.

So to move this over from the original thread, here is a suggestion that is zero sum, accounts for match score being weighted for win/loss, and compares the quality of play across all 24 players:

• The top 8 in the match gain PSR (+5 for the top score, +2 for the rest)
  
• The middle 8 do not change
  
• The bottom 8 drop PSR (-5 for the bottom score, -2 for the rest)

Over time this will shuffle players into their appropriate Tiers. The more you play, the more accurate it gets, and the less any one stomp in either direction will matter. The whole point is to find that middle ground where you are well-balanced and that's why the middle 8 wouldn't change, they're already in the target play level for that match-up. The top 8 may be worthy of a higher Tier, so add to their PSR; the bottom 8 may be happier in a lower Tier, so move their PSR down. Winning teams will average more PSR gain, losing teams more PSR loss, but star or slug performers still get an appropriate individual adjustment.

It's not perfect, but it balances team win/loss and individual performance, and is zero sum, which is necessary to evenly distribute player skill across Tiers for the matchmaker (without an even spread, queue times could be awful for low population Tiers). Over time this should produce a fair spread to drop players with comparable opponents - and that is the goal, for the matchmaker to give us teams that turn into really fun scraps of equals!

Edited by Teknomancer, 08 June 2020 - 10:15 AM.

[Sniper09121986]

PeppaPig, on 08 June 2020 - 07:15 AM, said:

Three fantastic matches today - each one someone ran and managed to capture the base, no match scores over 80 on both sides - everyone loses on PSR (this on polar highlands), two other matches alpine and polar highlands lrm hell (system dropped into low teens fps because the sky was so heavily full of missiles), sixth match cored from behind with a weapons test fire, two more where everyone full on full nascar and got wolf packed in my assault. Yes, new PSR will work wonders; can I see my negative PSR score please . . . /sigh

The pain is real. The only consolation I have to offer is that you get a temporary seal hunting licence

[Knight Captain Morgan]

Sniper09121986, on 08 June 2020 - 10:13 AM, said:

The pain is real. The only consolation I have to offer is that you get a temporary seal hunting licence

Unless they bypass the match maker tier restrictions entirely by group dropping. Then it’s more of a permanent license.

[Tarl Cabot]

As previously mentioned, I do not see PGI completely reworking the Tier/PSR setup, simply modifying the current setup.

And people have different takes on zero sum, with PSR zero sum being static while others want dynamic zero sum/match, and PGI is using Tiers for the MM.

Is PGI's PSR zero static zero sum too extreme? I believe that is what many are trying to say, in different ways. Pic below adds a level, breaking up 201-400 sections. The break point now is 200-300 MS, combined level.

I modified my previous setup, again, adding an additional level by breaking up the 101-250 and 251-400. I also included the current values for a loss, in which the break point was the 101-250 MS range. This would help explain how someone w/ avg lvl 171, through brute force by playing 10s of thousands of games reached Tier 2 and Tier 1. Mind you, LOTS of games, and they would not be alone. And there has to be a break point, a spot where most people would float around

[Gagis]

The perfect PSR system would be Elo ratings. We had them in the past, and it was good.

[The Teddy Bear]

One aspect of this that I dont know if it has been discussed too much is the impact for the solo players that, when having teams involved in the matches, they are not going to be able to have the same impact on a game. This is at least my past experience.

If this is correct. We should by now be able to look into jarls list, or better yet PGI look in their data, and see that a "heavy impact solo player" to a match will have seen a good reduction in their W/L in May and vice versa a "low impact solo player" to a match can very well have increased W/L in May due to neither of them having the same impact in the matches as they could have in the old solo que.

Putting this aspect, if true, in the proposed PSR change from PGI. I think the "good" team players will skyrocket themselves into tier1 while the "good" solo players will get the more in line with a slow climbing gasbaloon. I the same way the "bad" team players will sink like a big stone to a low tier while the "bad" solo player will take considerable longer time on them to get there.

[Lionheart2012]

OneTeamPlayer, on 06 June 2020 - 08:25 AM, said:

1 in 3, folks, 1 in 3.

It was 1 in 3 prior to the merging of the queues.

[Turlo]

Teknomancer, on 08 June 2020 - 10:09 AM, said:

So to move this over from the original thread, here is a suggestion that is zero sum, accounts for match score being weighted for win/loss, and compares the quality of play across all 24 players:

• The top 8 in the match gain PSR (+5 for the top score, +2 for the rest)
  
• The middle 8 do not change
  
• The bottom 8 drop PSR (-5 for the bottom score, -2 for the rest)

Over time this will shuffle players into their appropriate Tiers. The more you play, the more accurate it gets, and the less any one stomp in either direction will matter. The whole point is to find that middle ground where you are well-balanced and that's why the middle 8 wouldn't change, they're already in the target play level for that match-up. The top 8 may be worthy of a higher Tier, so add to their PSR; the bottom 8 may be happier in a lower Tier, so move their PSR down. Winning teams will average more PSR gain, losing teams more PSR loss, but star or slug performers still get an appropriate individual adjustment.

I like this idea as it addresses the fact our match scores are already influenced by winning and losing.

The elephant in the room is the calculation of the match scores. There is no strong mechanism that I am aware of that rewards the dedicated team players (which PGI wants to promote). If I am in an ECM light that stays with the assaults to provide ECM cover, I get absolutely zero match score for that. For the scouts and skirmishers that distract or push the LRM boats so the rest of the team can push, there is zero match score reward for that. That Atlas that charges in, taking all the fire from the opposition to the glass cannons can charge in and avenge the fallen Atlas, gets zero match score.

As such, I stopped providing dedicated ECM coverage, despite the requests and insults of the team. I am less eager to distract the grouped LRM boats over chasing lone mechs behind the conflict line because that way, I get points and can move into a more challenging Tier. When I play assaults (a rarity) I don't charge in, soaking up damage and dying fast for the team to be able to charge in behind me, but instead wait and pick my moments. Sometimes that means we lose initiative, but hey, I was going to get a low match score anyway. However, I LOVE providing AMS cover for my team mates, because it has a nice impact on match score.

Once I made this change in game philosophy, I moved from Tier 4 to Tier 1 in 18 months, after years of not moving out of Tier 4.

That said, I am looking forward to the PSR reset and a new way to calculate Tier leveling. I changed my game once, I can change my game again.

[Xaat Xuun]

guess with the zero sum, possibility of no more objective rushes, everyone would be -5 win or lose

[Captain Caveman DE]

RRAMIREZ, on 08 June 2020 - 12:53 AM, said:

I'm sorry but i can't convince myself that there's always 3 people that didn't contribute in any way (and MS may be the problem) on winning side, whatever the match shape.
Or at least didn't contribute for their team to win more than the 3 "best" players on loosing team, always.


But as I already said, it won't prevent me from playing.

reality is: there's a lot more people who 'don't contribute' when I look over recent match-screenshots in this forum - but you can't punish everybody
joking aside: matchscore sorts things fine (enough), and if you have the 10th worst matchscore on your team, chances are very high *cough* that there's 9 people performing better than you on your team;
"real" performance aside, if you drop with 11 "gods amongst men" who do a 12:0 in under 2 minutes, and you didn't get a chance to contribute, you'll go down only a little. and only for that specific match. it's not the end of the world.



oh, and the 3 best on the losing side are (nearly, and the chances are so slim I have to type it tiny as this) ALWAYS gonna carry more than the 3 worst performers on the winning side. but test it yourself; screenshot a few or even a lot of matches and post a picture that proves this 'guesstimate' wrong. pls

Snowhawk, on 08 June 2020 - 08:06 AM, said:

Somehow I like this idea. Maybe the numbers need some fine tuning.... This System is "dynamic" and adapts to the current Team Performance and hardworking Players are rewarded even in a loss.

thx mate; the numbers where just casually spitballing, they surely need adjustment; I was just trying to show what zero-sum looks like and how it can be implemented;
you could for example stretch out the zero-points-zone in both teams, give or take another place that goes up and down, put bigger differences in numbers in there - just anything, really.
*was just really trying to show how easily you can replace the current system with a totally different one, that probably works better - and is adjustable by itself in many ways*

Teknomancer, on 08 June 2020 - 10:09 AM, said:

Now this is zero sum! A good suggestion...

However, this has a potential problem. You can't always guarantee that the best on the losing team should go up, and the worst on the winning should go down. What if the best of the losers was still out-scored by the worst of the winners? Then he goes up despite losing and being outplayed, while there is a winner who outplayed every loser but still lost PSR.
Match score is already weighted to give a bonus to winning. So long as that is true, then there is no need for separate target numbers for winners and losers for their PSR adjustment. And so long as match score is variable by a lot of evolving calculations, then assigning target scores for PSR changes is also flawed. 500 match score in one game might have been achieved with worse performance than a 400 score in a different game, all depending on who values what, the overall match flow, the team dynamics, game mode, tweaks to the formula from one patch to another, etc.

A fair zero sum algorithm using the existing match score system needs to measure performance across all 24 players at once, not just your team. The score is already weighted to give winners a boost. It has to allow for a PSR gain if your team lost but you outplayed the field, and for a PSR drop if you were significantly outplayed even though your team won.

Cheers mate.
-you're right in that the win-bonus has to go @MS

-you're wrong on the "look at 24, not at 2x12" thing, imo. let me explain by example;
you got a 12:3 stomp. they happen a lot.
now, players on the winning team have, after taking the kill-lead, a relatively easy time on the battlefield; they will roll out more damage, have more bodies to rely on, to distribute the other teams damage on etc, etc.
so it's an environment where you can get a good MS easy.

the losing side has it harder to get meaningful MS, since they have all the above things going AGAINST them.

you gotta look at 2 sets of 12, to keep things fair; if you don't, you always drag a lot of people upwards, just because they live in an environment that gives them more MS (and more points if you do it your way).

hope it does make sense

Edited by Captain Caveman DE, 08 June 2020 - 01:01 PM.

[RRAMIREZ]

Captain Caveman DE, on 08 June 2020 - 12:49 PM, said:

if you drop with 11 "gods amongst men" who do a 12:0 in under 2 minutes, and you didn't get a chance to contribute, you'll go down only a little. and only for that specific match. it's not the end of the world.

no.. 2 "gods amongst men" will also go down.
Not sure that they will appreciate and not complain here ^^ (more than I do)

I'm not afraid to go down.
I'm an average player.

But you still didn't convinced me that this methode is (edit) better more fair than looking at MS (as long as the threshold are based on global stats) to get relative + - than positionning in one match whatever the result or scenario.

ps:
One the other hand, i don't want to spend time trying to convince you or anyone any longer.
Hope you'll play again.

Edited by RRAMIREZ, 08 June 2020 - 01:05 PM.

[Teknomancer]

Gagis, on 08 June 2020 - 11:44 AM, said:

The perfect PSR system would be Elo ratings. We had them in the past, and it was good.

Elo was designed for 1v1 games, specifically chess originally. It has the advantage of a rating change that accounts for the rating of your opponent. A low-rating player who upsets a high-rated player can get a significant boost, but that high-rated player is not dropped an equivalent amount because it accounts for having a bad day.

But it doesn't scale as well to coordinated teams, and that's one reason I think it was decided not to use Elo for QuickPlay. There are team sports that use it, yes, but they aren't randomly assigning players to teams the way QuickPlay in MWO has to. Then there is the "Elo hell" problem, where a good player can be held back to a low rating unfairly if they are consistently matched with players who have inflated ratings. Your Elo rating over time can't adapt accurately if the pool of opponents is either really strong or really weak.

[Teknomancer]

Captain Caveman DE, on 08 June 2020 - 12:49 PM, said:

reality is: there's a lot more people who 'don't contribute' when I look over recent match-screenshots in this forum - but you can't punish everybody
joking aside: matchscore sorts things fine (enough), and if you have the 10th worst matchscore on your team, chances are very high *cough* that there's 9 people performing better than you on your team;
"real" performance aside, if you drop with 11 "gods amongst men" who do a 12:0 in under 2 minutes, and you didn't get a chance to contribute, you'll go down only a little. and only for that specific match. it's not the end of the world.



oh, and the 3 best on the losing side are (nearly, and the chances are so slim I have to type it tiny as this) ALWAYS gonna carry more than the 3 worst performers on the winning side. but test it yourself; screenshot a few or even a lot of matches and post a picture that proves this 'guesstimate' wrong. pls





thx mate; the numbers where just casually spitballing, they surely need adjustment; I was just trying to show what zero-sum looks like and how it can be implemented;
you could for example stretch out the zero-points-zone in both teams, give or take another place that goes up and down, put bigger differences in numbers in there - just anything, really.
*was just really trying to show how easily you can replace the current system with a totally different one, that probably works better - and is adjustable by itself in many ways*



Cheers mate.
-you're right in that the win-bonus has to go @MS

-you're wrong on the "look at 24, not at 2x12" thing, imo. let me explain by example;
you got a 12:3 stomp. they happen a lot.
now, players on the winning team have, after taking the kill-lead, a relatively easy time on the battlefield; they will roll out more damage, have more bodies to rely on, to distribute the other teams damage on etc, etc.
so it's an environment where you can get a good MS easy.

the losing side has it harder to get meaningful MS, since they have all the above things going AGAINST them.

you gotta look at 2 sets of 12, to keep things fair; if you don't, you always drag a lot of people upwards, just because they live in an environment that gives them more MS (and more points if you do it your way).

hope it does make sense

It does make sense, and I can absolutely see the problem of stomps. Any effective rating has to handle stomps, and stomps will be a point of debate through all this, heh. Your scoring definitely prevents a stomp from unfairly moving a PSR score because of a tough break. There is much to like with your numbers.

However, I do believe you need to compare all 24 players, because the more narrow the comparison, the less it helps balance over time. The more matches it will take to find your appropriate Tier, which doesn't help more casual players. Those 24 players could be allies or enemies on any given drop, or mixed up. Over multiple games, the matchmaker needs to account for your performance compared to as many players as possible. You're not ranking against your teammates, you're ranking against the whole player pool to find where you best fit. Adjusting PSR solely within your team's performance doesn't account for how well or poorly you did relative to the opposition.

Now yes, a stomp throws any scoring out of whack a bit. But there can be reasons for that stomp. If you all sat around and did nothing, or squabbled with no coordination, a stomp may be justified. Should your best players be guaranteed to rise if they still were awful compared to the opposition? Then your PSR is artificially inflated. You could have decent players on the winning team forced to lose PSR needlessly. You need to compare all 24 to avoid this.

Edited by Teknomancer, 08 June 2020 - 01:43 PM.