849 replies to this topic

[Paladin IIC]

PSR resets are welcome, so are the delicious tears of the stat ****** it will create :3

[Bennesto]

FearThePaladin, on 26 May 2020 - 05:21 AM, said:

I say get rid of the teir system and just drop folks.

No, I am tired of farming potats.
Potats are tired of getting farmed.
If you honestly think that balance doesn't matter in a competitive team vs team setting, I have bad news for you...

[Dogmeat1]

This is going to be long but please try to bear with me.

Zero-sum sounds great in theory but zero-sum systems like ELO fail in team based games like this because they only usually only take into account wins and losses. For matchmaking quality to improve, it is critical that a player ranking system is implemented that actually assesses how much an individual player contributed towards a win or loss. PSR could actually do this if scaling matchscore targets and a dynamic tier system was introduced. To my knowledge the current PSR system works as follows;

Match score target	Win result	Loss result
300	large increase	small increase
250	small increase	no change
200	no change	small decrease
150 and below	small decrease	large decrease

Now this doesn’t work because the targets are too easy to achieve and given enough games, nearly anyone will max out their PSR and be matched against players well above their skill level. However simply removing the PSR cap and adding a formula like this would solve this issue;

scaledMatchscoreTarget = MatchscoreTarget * (PlayerPSR / MedianPSR)

And here's how it looks if we decide the server median (50%) PSR is 1600;

Player with 1600 PSR	Player with 2400 PSR	Player with 1200 PSR
300	450	225
250	375	187.5
200	300	150
150 and below	225	112.5

Given enough games, eventually every player will reach a point where their matchscore targets are going to be roughly on par with their average game results thus resulting in minimal PSR changes. Of course in general heavier weight classes produce higher average matchscores than lower ones, and certain mechs are much better than others. To reflect this we can simply use the average matchscore (based on server stats) of each variant as additional variables in the formula (we can default to weight class averages for new or rarely played variants). The new formula might look like this;


scaledMatchscoreTarget = MatchscoreTarget * ( (PlayerPSR / MedianPSR * 0.8) + (playerMechAverageMatchscore / medianMechMatchscore * 0.2) )

In this example I assigned a 80% weight to the player’s PSR and a 20% one to the mech’s average matchscore but these could easily be tweaked. You could even add further variables so that the strength of the player’s team compared to the opposition is also taken into effect.


So how would the tier buckets work?

To make a system like this work the static PSR targets, currently used to determine which tier an individual player belongs to, need to be removed. Instead a dynamic system that is regularly updated and that uses the median values of the active playerbase should be implemented. This would mean, regardless of how big or small the active playerbase becomes, each bucket would maintain the same percentage of players. For example;

Tier 5 = 1-25%; Tier 4 = 26-50%; Tier 3 = 51-75%; Tier 2 = 76-90%; Tier 1 = 91-100%.

As long as the targets are being updated on a regular basis (monthly, weekly, daily, etc) then each bucket would always roughly maintain that percentage of players.


What about the matchmaker?

To make this really work, the matchmaker needs to assign players to lobbies in pairs. What I mean if a tier 1 assault player jumps into queue, the MM waits until another tier 1 assault players enters before assigning them to a lobby (on opposite teams). This would ensure that even if there is a mix of tiers on any given team, the opposition would have the same mix of tiers and weight classes. If after a certain period no equal tier player can be found, then the MM can start breaking the MM rules. I would like to see some testing to see which produces better results; break the tier rules or breaking the weight class ones first.

For groups the same pairing system should apply. If a tier 1 4-man group jumps into queue then they should be only matched against a tier 1 group of equal size. If the MM can’t find another tier 1 4-man then it should look for (at tier 1) 2 2-mans, a 3-man + 1 tsolo player and finally a 2-man + 2 solo players in that order. Only after waiting a certain period should it start looking for lower tiers and a group should never be matched against a complete lance of solo players.

Now these PSR, tier and MM changes are far from perfect, but I think they would result in far better quality matches and a better separation between new, casual and try-hard players that would result in a happier situation for everyone involved. Furthermore I think I don’t think this would require extensive modification of existing code (other than possibly the MM) and should be realistic to implement.

[Mr.Wrong]

reset the tiers on a regular base so they actualy mean something

[Horseman]

Belorion, on 26 May 2020 - 05:28 AM, said:

I kind of like the way PSR works now. Why does it have to be a zero sum game?

Because right now over 60% of the playerbase has floated up to Tier 1 and the matchmaker can't really tell apart an 80-percenter from a 99-percenter. Or even a 38-percenter who ground so many matches that he maxed out T1 anyway. As a result the matchmaking quality is a dumpster fire.

[Sniper09121986]

Kurlon, on 26 May 2020 - 05:08 AM, said:

So, it appears years of the current malfunctioning 'tier' system have trained some of the population to think tiers are unlocks like classes of ships in World of Warships, an achievement that once garnered can never go away. Additionally they now "deserve" to play in top skill level matches. But PGI was already aware of this, which is why they didn't just say they're resetting things but instead asked what the community feels.

So, as noted before, the 'tier' system right now in combination with the matchmaker's doors being not just open but ripped clean off the hinges means you could literally just randomize people's PSR before each round of matchmaker action and not radically alter the overall gameplay at the moment. My updated proposal then is not only a PSR reset, but a UI change, get rid of Tier in the UI. Show the PSR/ELO number, or come up with a new name for the brackets, or just not show it at all, but don't continue to call it Tier. That allows the portion of the population who wishes for better balance in matches to see some light at the end of the tunnel, while also allowing those that think of Tier as an achievement bar to hold on to the memory of their old 'rank' without the UI trying to tell them otherwise after the change.

I think you care too much about what people think about "tier entitlement". Nothing wrong with playing in bronze league in SC2, as long as nobody is stomping anybody. Tired of being in bronze? Just play better. Gradual increase of difficulty makes for better learning experience too.

jonomy, on 26 May 2020 - 05:41 AM, said:

great everyone is talking about this and thanks for a chance to chime in with insight, concern and questions. I am curious if resetting PSR will become a regular or periodic task going forward, or will this become a discussion again in a few months?

Not directly stated in Paul's post, but SC2 and QC reset PSR across the board on a regular basis, so does our own S7. Player skill also does not stand idle. Personally I used to do well in the past, but recently fell into a slump for reasons I cannot fully understand, and as much as it would irk me it would be better for me and everyone I play with for me to spend some time in lower leagues until things get better. Periodic PSR reset would speed up people's travel up and down significantly and thus go a long way in determining actual player skill at any given point in time.

FlynnTheAvatar, on 26 May 2020 - 05:57 AM, said:

- Dying early does not mean you played badly (you led the push that won the game).

Not seeing how the system would determine that particular situation. Agree on the other points though.

Paladin IIC, on 26 May 2020 - 06:05 AM, said:

PSR resets are welcome, so are the delicious tears of the stat ****** it will create :3

Eh, after reading this thread I say that particular salt mine is already starting up

[Jesslp]

Belorion, on 26 May 2020 - 05:30 AM, said:

I agree, if you are going to make it zero sum, you can still do that without placing it all on the win loss. Base it on the performance for the individual. First people to get killed should get a lower PSR and the ones that do the most should go up.

I completely disagree with that. So lone lurmers, or lone gauss snipers, if they die last, they get even an extra for staying away from team and hiding, apart from the damage they do, at the same time they completely disregard teamplay?

N-O W-A-Y !

[Ebelon]

Dakkalistic, on 26 May 2020 - 04:49 AM, said:

But they're absolutely right. I do not belong in Tier 1, the XP bar pushed me there.

With all respect, Gentleman, they are not!

Especially the "git gud" crowd is not right.
Since all they prove is they mastered to gain the max of the actual PSR calculation, which means:

- do as much damage as possible, leading to shoot torsos, instead the already yellow / red CT
- stay behind, use meat shields with your heavy and or assault mechs to be able to do as much damage as possible
- steal kills wherever possible

Getting to tier 1 very fast requires not being a good team player, but to misuse the PSR calculation which leads to playing selfishly

Others, who try to be a bonus to the team do not get the rewards, they earn because they usually die earlier (Assault, pulling the fire on a rush and so on).

Everybody, who feels "top tier" by getting there in no time since his skill is so good, is not a worthy lancemate.

Like I said before:
- Remove tier indicator in the client, so players do not know in which tier they are in
- change tier calculation to reward team players more than selfish players who just abuse the system (of course thats the tricky part, to determine, who has contributed to the team and who has not)

[Znoop]

Please reset the pilot skill rating Mr. Inouye.

[Jesslp]

One really big difference that i find anytime I play, its wether there is or not, a "drop leader" that uses in-game comms to call targets and command the drop.

If there IS a "drop leader" doing that -> big chances of winning. If theres not, and the other team has one, big chances of being stomped.

Leading a drop, currently is not rewarded (apart from the fact that your team might win). In fact there is not a way to tell the game who makes the effort to try to call the drop, so those players get rewarded, even if only for giving it a try.

Game should then, implement a way for players of a team to "vote", who among them, at least tried to call the drop. And reward that. And probably, for drop callers, to also indicate to the game, which players did actually respond to "team orders" and deserve some extra match score etc, for doing teamplay.

Best way to even teams, is to make sure you get both "drop callers" ("drop leaders"... whatever you want to call them), and some team-playing players (:-) in both teams.

Edited by Jesslp, 26 May 2020 - 07:07 AM.

[Curccu]

reset!

[Znoop]

Ebelon, on 26 May 2020 - 06:57 AM, said:

With all respect, Gentleman, they are not!

Especially the "git gud" crowd is not right.
Since all they prove is they mastered to gain the max of the actual PSR calculation, which means:

- do as much damage as possible, leading to shoot torsos, instead the already yellow / red CT
- stay behind, use meat shields with your heavy and or assault mechs to be able to do as much damage as possible
- steal kills wherever possible

Getting to tier 1 very fast requires not being a good team player, but to misuse the PSR calculation which leads to playing selfishly

Others, who try to be a bonus to the team do not get the rewards, they earn because they usually die earlier (Assault, pulling the fire on a rush and so on).

Everybody, who feels "top tier" by getting there in no time since his skill is so good, is not a worthy lancemate.

Like I said before:
- Remove tier indicator in the client, so players do not know in which tier they are in
- change tier calculation to reward team players more than selfish players who just abuse the system (of course thats the tricky part, to determine, who has contributed to the team and who has not)

Sir, it's really sad you so broadly attack the "git gud crowd", I personally try to better myself in-game and get the best outcome out of the matches I participate in, it can happen they players farm damage to increase their matchscore (for example to get an event done) but in most close or even matchups players try to dispose of enemy Mechs in the fastest possible manner to preserve their team. It's also natural to use the armor of your team mates, in general Assault Mechs have the most firepower in a team, the longer they live and can apply damage to the enemy team the better it is for the outcome of the match (Edit: Assaults just mentioned as example). I also believe ( I don't know though) that simple kills are not worth as much as KMDDs. I personally don't care about the kill, if I damage an enemy so bad that a teammate can finish him easily it's totally fine by me, a destroyed Mech cannot hurt you anymore.

Edited by Znoop, 26 May 2020 - 07:10 AM.

[Xeroize]

I love the idea of resetting the tier ranking and making it zero-sum (as this would make the ratings far more meaningful); however, I think it needs to be accompanied by a revised/improved new player introduction/tutorial.

I myself pushed hard to get to tier 1 not because I care about what my rating is for some kindof flex, but rather to get away from players in the lower tiers many of whom I feel create a much more passive game.

I think all new players have that experience where they push out by themselves and get slapped hard for doing so (either by LRM fire, or by crashing into an enemy firing line) creating an instant strong feedback loop against overly aggressive behavior. The problem is that many players swing far to hard in the opposite direction and become afraid to push out from cover under any circumstance. It is frustrating to have a short-range brawler build and be stuck where your team is simply trading at mid-long range with no opportunity to contribute to the fight until the game is already decided. I get that you can simply change to a longer range trading mech, but it feels like the lower tiers force you into a particular type of play style until you are past the lower tiers and MWO comes alive. I've known several individuals who gave up on the game as being "not fun" or "too slow and passive" only to later watch me playing tier 1 games and go "hey! that actually looks like a blast, that wasn't at all what I experienced" and I have to explain to them that the game becomes far better once you get out of the lower tiers.

If the game had a tutorial that went over some of the tactical basics (Trading, Nascar, Pushing, Brawling, Etc) as well as some of the essentials (armor sharing, torso twisting, and heat management) we would have a player base that understands the game far better coming in, and hopefully have more engaging fights in the lower tiers. Maybe set up a series of challenges with some C-bill rewards or something to encourage folks to learn the basics.

Just my 2 cents on the subject.

[LordSkippy]

PSR should look at your match score, unweighted by win/loss. Someone that does 500 damage on a loss was a better pilot than someone who did nothing, but got carried to a win. The major problem with the current implementation was pointed out when it first went live, there is no way to go down on a win. Effectively making it skew PSR up overtime. If you want to make it harder to gain on a loss, fine. But make it possible to go down on a win to balance it out.

If you're going to make it impossible to gain PSR on a loss, then it needs to be harder to gain PSR on a win. For most games, win or lose, your average pilot shouldn't gain or lose much, or any, PSR. The top players on a win go up, while the bottom players on the loss go down would be okay. Except, so much of the match is determined by so much out of a single player's control, that group win/loss is almost completely meaningless.

Edited by LordSkippy, 26 May 2020 - 07:27 AM.

[Gagis]

If you are a good team player, you win more often than you lose, because you bring something to the table and luck evens out over a large number of matches.

If you are a bad team player, you lose more often than you win. If you try to contribute to team play but still lose more, whatever it is you were trying to do, does not actually work, and should not be encouraged.

The details of individual games do not really need to be taken into account with minute detail, since they all even out over time.

Edited by Gagis, 26 May 2020 - 07:27 AM.

[Jesslp]

denAirwalkerrr, on 26 May 2020 - 07:21 AM, said:

---(BS here)---
Teamplay is not a metric you can measure but constantly doing horrible damage is.

There is no "absolutely perfect teamplay metric", but it indeed can exist. Even better, several metrics for teamplay can be mixed and weighted. A metric for perfomance purely based on "horrible damage" it isnt a good one, too.

Read this:

Jesslp, on 26 May 2020 - 07:05 AM, said:

(drop leaders and teamplay...)

Also, a good punishment for shelfish players would be... to team them together, no matter how good or bad they are, against players who do good teamplay, and WANT to play as a team. Even if just for seeing what comes out of that. It probably wont cure selfishness, but... hey! Worth giving it a try,

Edited by Jesslp, 26 May 2020 - 07:33 AM.

[Spare Knight]

I will start by saying that this is my favorite game of all time. I hate playing against A.I. and I love big stompy robots.

I had said that I would quit when the merger was made permanent, but this news changes things. :-)

Right now, the Tier level means doodly squat. Let's do this. Those who deserve to be Tier 1 will climb right back up there and those who do not will settle into the Tier where they belong. The Tiers will finally mean something. This is a win for everyone!

Do it!

[denAirwalkerrr]

Jesslp, on 26 May 2020 - 07:31 AM, said:

There is no "absolutely perfect teamplay metric", but it indeed can exist. Even better, several metrics for teamplay can be mixed and weighted. A metric for perfomance purely based on "horrible damage" it isnt a good one, too.

Also, a good punishment for shelfish players would be... to team them together, no matter how good or bad they are, against players who do good teamplay, and WANT to play as a team. Even if just for seeing what comes out of that. It probably wont cure selfishness, but... hey! Worth giving it a try,

Bruh do you understand that about 90% of games are won with whole enemy team dead so damage is the most important thing you do in the game no matter what because it's how you win. And the best teamplay is killing half of enemy team on your own.

Edited by denAirwalkerrr, 26 May 2020 - 07:37 AM.

[Diagonalcircle]

Reset please, thanks for sharing the data.

[ENOVA]

R E S E T P S R

Edited by ENOVA, 26 May 2020 - 07:37 AM.