490 replies to this topic

[ShiverMeRivets]

Aidan Crenshaw, on 05 June 2020 - 06:58 AM, said:

The main fault I see in most post is that they concentrate on the outcome of a single match. This is wrong. The system has to account for all your matches together. That's where it is zero sum. That you don't go up on a loss is just to counter matchscore gaming strategies that don't contribute to wins.

One guy gets it.

Contributing to your side winning by any means which may not be factored into score is more important than farming match scores. So, to go up in PSR you need to WIN more than lose. If you somehow manage to *average* 400 score and still be under 1 W/L then you harvest garbage damage and not actually help your team - in practice this is very unlikely.

Good move PGI.
The refinement will come from tweaking the scoring system.

[Tesunie]

My take on the numbers as a rough guess for ranges. And this would be standard, win or lose.

Match Score: 0-100 goes down in PSR by -3
Match Score: 101-200 goes down in PSR by -1
Match Score: 201-300 does not move.
Match Score: 301-400goes up in PSR by +2
Match Score: 401+ goes up in PSR by +5

This is with an estimation that average match scores is probably around 250, so I gave a 50 point margin. Then, from there, I did roughly 100 point steps in either direction that give PSR changes. If you are performing below average for the tier you are in, then you will slowly drop until you hit the tier you are suppose to be in. If you are performing above average for your tier, you will go up until you start to perform more average numbers.

Of course, the ranges are a guess. I don't have the same data to make a more accurate number range for what might be considered "average".

[Dionnsai]

You do need to account for win/loss. And all of you need to be thinking in sets of hundreds to thousands of matches, not the last 10 matches you had. Players have agency and they do influence their team towards a win or a loss. If you play 1000 matches and you lose 600, YOU tend to influence your team towards a loss.

Edited by Dionnsai, 05 June 2020 - 07:22 AM.

[Ahh Screw it - WATCH THIS]

Tahawus, on 04 June 2020 - 11:59 AM, said:

Wow, PGI, you're not inspiring confidence here.

Number 1. The proposed algorithm is not zero sum.

Number 2. My professors would have failed my *** if I said a move from 29.91% to 33.99% was a 5% increase. It's a 5percentage point increase. It is more correctly a 13.6% increase.

Well, in their defense.....

Math is hard.

[Brauer]

Negat1ve Nancy, on 05 June 2020 - 06:48 AM, said:

Pretty sure I played this one skillfull



5 kill and four of them solo



Highest damage from both sides, Equal highest kills from both sides, 558 match score, second highest from both sides

MM currently gives me a small PSR boost which reflects the performance.
The new MM...smh

Yes this is my alt

TBF any PSR system should be more concerned about getting things right in aggregate than individual matches. So yes you may have deserved a slight increase here, but if the primary measure of pilot skill is wins and losses then PSR changes should be most heavily tied to W/L.

I have been hearing a lot of folks say that this system is not truly zero sum, and that strong performances in losses should also get some kind of PSR increase. I agree with both of those points. I just think it's important that people not lose sight of the forest for the trees like you are here. Individual matches really don't matter, what matters is if the system is doing a good job across hundreds of matches so that over time it does a better job.

[Negat1ve Nancy]

If the system is fundamentally flawed their is no forest
It got wiped out by longtom levels of incompetence.
Pilot Skill rating what a joke.

Poor performing winners will still, guess wot...
Create upward bias

Edited by Negat1ve Nancy, 05 June 2020 - 08:02 AM.

[Brauer]

Negat1ve Nancy, on 05 June 2020 - 07:55 AM, said:

If the system is fundamentally flawed their is no forest
It got wiped out by longtom levels of incompetence.
Pilot Skill rating what a joke.

Poor performing winners will still, guess wot...
Create upward bias

I mean yeah, a match score up to maybe 150 or 200 probably shouldn't see an increase no matter the case. I'd also argue though that if a player is consistently getting a match score of around 150 or 200 they probably are at best winning around half the time and at worst have a negative WLR. In both of those cases that player will actually be losing PSR on average, which is appropriate.

Who wins over half their games consistently with a 100-200 average match score???

[VonBruinwald]

Kill2Blit, on 05 June 2020 - 07:09 AM, said:

if you look at the distribution of match scores for the playerbase on jarl's, it looks like we'll have negative sum, or a downward bias. average players should be in tier 3, this system will push them towards tier 5. it's essentially the problem we have now in reverse. population will be skewed downward, t5 and t4 will be the mess that t1 is now, and the pop in t1 will be so low that they're pulled down into t5 with everyone else.

The problem with using the Jarl's data is it's from a broken system.

There are many players in T1 who shouldn't be there but are because of the upwards bias:

VonBruinwald, on 29 May 2020 - 04:43 AM, said:

Currently you're a T3 in T1 getting clubbed. Your current stats are what you managed to chip out between clubbings.

Post-PSR reset you will no longer be over-tiered. You'll be a T3 fighting T3's, putting you are equal footing. As a result your stats will rise because you're no longer the perpetual-underdog.

[Negat1ve Nancy]

I have no faith in the guy and company making the changes to match making

[Snowhawk]

Tesunie, on 05 June 2020 - 06:55 AM, said:

PSR still linked heavily to W/L... means PSR will still be a problem.

Remove the W/L calculation from your numbers completely. Players who perform within an average range of match score should not move. Those who perform under the average range bracket should decrease. Those who perform above the average should increase.

This makes it Pilot related, not "how did the team do" related.

The more I think about it the sooner I come to the same result…
Winning and losing a match is overweightet in this new psr matchmaker System. Your personel Performance should be more crucial.

[Brauer]

Negat1ve Nancy, on 05 June 2020 - 08:17 AM, said:

I have no faith in the guy and company making the changes to match making

Well you're not exactly alone in being skeptical of PGI's ability to make a good change to the game.

[kapusta11]

According to Jarl's list 251-400 match score bracket will contain almost half of the population. The thing is top 1% or even top 0.5% players have a much greater capability to swing games than top 5 or 10% for example. PSR should be heavily biased towards higher match scores. Similarly the difference between players and therefore their impact on the game's outcome within the bottom 40-50% isn't enough to warrant their separation.

Win
0-200 (approximately bottom 40%) -1
201-275 - No change
276-350 +1
351-400 (top 5-10%) +2
401-450 (top 1%) +3
451+ (top 0.5%) +5

Loss
0-200 (approximately bottom 40%) -5
201-275 -3
276-350 -2
351-400 (top 5-10%) -1
401-450 (top 1%) No change
451+ (top 0.5%) +1

Edited by kapusta11, 05 June 2020 - 08:34 AM.

[Brauer]

Snowhawk, on 05 June 2020 - 08:18 AM, said:

The more I think about it the sooner I come to the same result…
Winning and losing a match is overweightet in this new psr matchmaker System. Your personel Performance should be more crucial.

The purest measure of personal performance is whether or not you are winning or losing. Anything else is subject to being gamed or could improperly weight different actions in game.

[VonBruinwald]

Snowhawk, on 05 June 2020 - 08:18 AM, said:

The more I think about it the sooner I come to the same result…
Winning and losing a match is overweightet in this new psr matchmaker System. Your personel Performance should be more crucial.

The problem with using personal performance is people can cheese it at the expense of their team.

This is a team game. You're meant to play for the team's win.

Solaris was always intended to be the mode for personal glory.

[SirHavan]

I'm good with it - don't think its going to matter much with the current player base - my question is when will it be implemented (did I miss that part somewhere?

[VonBruinwald]

SirHavan, on 05 June 2020 - 08:27 AM, said:

I'm good with it - don't think its going to matter much with the current player base - my question is when will it be implemented (did I miss that part somewhere?

Last line.

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

[color=orange]Target Eta: June 9th, 2020 [/color]

[Dionnsai]

Lets all line up against the wall, 2 team captains and we'll pick teams at the start of every match.

[Cluster Fox]

Unpopular opinion.

I was reluctant at first, but the more I think about it, the more I think Win up/Loss down is fine. With one caveat : it doesn't work with absolute values.
Because:
- This is an average over time. Don't think of it as single match outcome.
- As a secondary objective to get better teamplay, this is done correctly.

Caveat. This system only works if the middle value (250) is exactly the average matchscore for the whole population. Even then, using absolute matchscore is flawed since someone with constant (say) 255 match score will always trend upwards, while 245 would trend downwards - this pushes people to the extremes.
The system would be fair if it was using a factor of the average score of the team from a match, rather than a fixed value.

Example:
WIN
0.0-0.4 Average team score : no change
0.4-0.8 Average team score : + 1
0.8-1.2 Average team score : + 2 (Zero sum here)
1.2-1.6 Average team score : + 3
1.6+ Average team score : + 5

LOSS
0.0-0.4 Average team score : - 5
0.4-0.8 Average team score : - 3
0.8-1.2 Average team score : - 2 (Zero sum here)
1.2-1.6 Average team score : - 1
1.6+ Average team score : no change

This results in a relative movement of all players based on team performance, effectively sorting out players inside the match, and teams against one another rather than moving them based on fixed values based on outside factors.

As for why I think not going up on loss / down on wins is fine:

Ability to go up on losses would skew some builds / drivers downwards or upwards.
What is usually the last mech standing on a loss? Usually a light, other times a LRM boat that's been ignored or an ECM poker. What is usually the last mech standing on the winning side? I honestly can't see that big a trend.

Example with the two extremes:
Having a high match score in a light for a loss is relatively easy, since lights arguably depend less on teamplay, however they can influence the outcome of a match heavily when used right.
On the flip side, assaults are the most dependent on teamplay. They usually have to use the team around them and sometimes follow the general trend of the team. Their score is more affected by W/L.

Edited by Cluster Fox, 06 June 2020 - 09:34 AM.

[SirHavan]

VonBruinwald, on 05 June 2020 - 08:30 AM, said:

Last line.

thanks, the sooner the better so we can all move on and just play the game

[MC Biber]

SirHavan, on 05 June 2020 - 08:38 AM, said:

... so we can all move on and just play the game

WORD