178 replies to this topic

[Nightbird]

OwO Kewensky, on 21 December 2020 - 11:16 PM, said:

Is it your assumption or it was documented somewhere?



Is it your another assumption? Right now I see no significant tonnage disparity but there were some extreme cases before.

Do you know the exact logic how the current matchmaker create and balance 2 teams and how much weight PSR has in that formula? Was it changed in any way after PSR reset, comparing to the MM we had before?
Knowing PGI designing/programming skills, I wouldn't assume anything about the MM.


When we had old PSR, matches were more tolerable comparing to what we have today. New PSR may be not good but it's better than the old one, so the only other significant contributing factor is groups.
High skill groups or low skill groups make soup queue equally unfun for solo players.

Even if I'm wrong, I'm not that far off because I was able to predict the MM result nearly half a year in advance. If the MM didn't minimize PSR difference the results would be worse that what I predicted. Tonnage controls doesn't matter if they are on or not.

[General Solo]

Egg Fu, on 22 December 2020 - 12:15 AM, said:

You say "The game has reached unplayable status as a solo player" yet your stats over this last month's December leaderboard 178 games are sitting well above any month of your entire MWO playing career. You're doing better than you ever have!

Rich!!!

Its easy to get better when your playing against garbage/skill gap
Still not what I call enjoyable
Gets boring quick

Where's the tension
Its just playing by auto pilot

[morosis]

Egg Fu, on 22 December 2020 - 12:15 AM, said:

You say "The game has reached unplayable status as a solo player" yet your stats over this last month's December leaderboard 178 games are sitting well above any month of your entire MWO playing career. You're doing better than you ever have!

Rich!!!

repeating myself, for the third time, on this: in the post PSR reset, stable soup era of MWO (roughly September to present) tracked stats are meaningless. so any argument you put forward based on those stats, is also meaningless.

I appear to be "doing better" only because i have increasingly refused to play as a solo, because the solo experience is truly unplayable.

Would it surprise you to learn that I can go 20-0 with a 5:1 KDR if i play in a group of people around my skill level, and also go 0-10 with a 1.5:1 if i play alone? That is, in fact the case. I wish it werent, but it is.

The entire point of this thread is to address the above problem in the current era of the game: playing solo is horrible, playing in a group, even a roughly average player like me can become Neo in the matrix.

it is completely broken, and I would like to see it fixed, not for my own benefit, but for everyone's. because as you just pointed out, if i just care about stats, clearly i am doing fine, right?

Edited by morosis, 22 December 2020 - 12:34 PM.

[morosis]

Monkey Lover, on 21 December 2020 - 02:34 PM, said:

What I see is people crying about getting killed by premade groups but then they show win ratios larger than premade groups lol

if that is all you see then i have no idea what you are looking at, and would question if your eyes are even open at all.

[morosis]

BackShot, on 21 December 2020 - 04:16 PM, said:

it is not even only that nightbird, a mediocre player can even have a same or better rating than a good player in the same amount of game, as long as the mediocre player put bigger raw damage number by using spread big damage weapon, while the good player kill effectively using pinpoint low damage weapon.

aka lrms, atms, streaks, mrms lbx etc... vs gauss, ppcs, acs..

the only meaningfull difference you will see if you check their stats, is the win loss ratio, wich is :

- not checked by " matchmaker"
- flawed by the grouping.

correcting for match score calculations / kill to damage efficiency etc is important, dont get me wrong, but out of scope for what i'm trying to hammer together here. what i want to do first, before anything else, is solve for the confounding factor that is group play in the solo queue.

groups break the current system and create most of the play problem issues, because they concentrate either skill or lack of skill on one or both teams. with the current implementation, high skill groups can also dodge each other, which ensures that consecutive or even 3-4 lobby strings will always have a different high skill group that cannot be matchmade against with an equivalent. this, according to both my own experience as well as the experience of many others, is the largest factor upsetting the quality of matches/ solo gameplay at present.

i think allowing a secondary balancing function to break up groups within an individual lobby and create more balanced solutions for players assigned to team 1 and team 2 can do nothing but help. i am pushing this option forward precisely because it only deals with this one aspect of the problem and avoids a scenario where PGI feels like we are asking for a large scale overhaul of the MM and PSR.

if large scale overhaul is on the table, then i think the bulk of us would agree that is the best option by far. under the assumption that PGI is unwilling to do that, then i think this proposal at least solves the worst aspects of the solo unplayablility issue the fastest.

[Nightbird]

morosis, on 22 December 2020 - 12:52 PM, said:

i think allowing a secondary balancing function to break up groups within an individual lobby and create more balanced solutions for players assigned to team 1 and team 2 can do nothing but help. i am pushing this option forward precisely because it only deals with this one aspect of the problem and avoids a scenario where PGI feels like we are asking for a large scale overhaul of the MM and PSR.

Your option isn't really an option though, you haven't even describe how it works.

To use an analogy... you want a car to be faster. Great. How do you achieve this? What is secondary balancing and how does it work? Do you have examples of lobbies with good and bad groups and show what the MM does in those cases? How do you even identify good and bad groups?

[Odins Fist]

morosis, on 16 December 2020 - 02:19 AM, said:

THE GAME HAS REACHED UNPLAYABLE STATUS AS A SOLO

I disagree, I all but quit the game but I comeback during Holiday events and have ZERO trouble going solo for old time's sake.

[morosis]

Nightbird, on 22 December 2020 - 01:04 PM, said:

Your option isn't really an option though, you haven't even describe how it works.

To use an analogy... you want a car to be faster. Great. How do you achieve this? What is secondary balancing and how does it work? Do you have examples of lobbies with good and bad groups and show what the MM does in those cases? How do you even identify good and bad groups?

as you more than most know, it is easy to spend way too much time attempting to do PGI's job for them, to no avail.

but honestly, even though this is climbing into the rathole a bit, i dont think it requires too much explanation, so i'll describe it as follows:

the mm is tasked with going out to the queue and sourcing 24 players it feels like "match" for a game lobby. it divides those players among the two teams as best it can via Tier and PSR, leaving premade groups intact. that is what we have today, and it is not good.

the concept of secondary lobby balancing is to have another round of logic that takes the 24 players already chosen, ignores any premade grouping that players have requested, and sees if the MM solution still makes sense. i.e. If team 1 has 4 99% players that were selected in a premade, and the highest rated player on team 2 is at 95%, then the secondary lobby balancer would "choose" to break up the team 1 premade, and assign two of the 99% players to team 2. similarly, if team 2 has a premade of 4 75% players and the lowest rated player on team 1 is at 80%, then the secondary lobby balancer would "choose" to break up the team 2 premade and assign two of them to team 1.

so for such an example, instead of launching a game where the 4 best players are guaranteed to be on team 1, now we have a game where at least two of them are guaranteed to be on each side.

again, there are still the confounding factors of PSR accuracy and match score problems to contend with, but just having the above corrected in the soup queue would be a godsend for the competitiveness of matches overall, imo.

Edited by morosis, 22 December 2020 - 01:48 PM.

[Nightbird]

morosis, on 22 December 2020 - 01:47 PM, said:

snip

Thank you for taking the time to describe your idea, I think I understand what you are asking for. The main issue I have complained (and provided evidence) about the current PSR is that two people with the same PSR number, or the same Tier, do not have the same skill level. Using your wording, assigning 99%, 95%, 75% PSR players evenly to two teams still requires those ratings to be accurate for it to work. What if the ratings today are inaccurate? A 90% rated player could be 70% in reality, and a 80% player could be 95%. Your MM wouldn't function in that case right?

All the evidence I have points to the the current PSR numbers being bad, and a Matchmaker with inaccurate PSR ratings can't create balanced teams and good matches as a result. Therefore I don't think changing the Matchmaker, while leaving the PSR, would make much of a difference.

P.S. I think you edited in a note about PSR at the end, but yeah, basically from my analysis PSR is 90% of the reason matches are bad.

Edited by Nightbird, 22 December 2020 - 02:14 PM.

[Anomalocaris]

Nightbird, on 22 December 2020 - 02:12 PM, said:

Thank you for taking the time to describe your idea, I think I understand what you are asking for. The main issue I have complained (and provided evidence) about the current PSR is that two people with the same PSR number, or the same Tier, do not have the same skill level. Using your wording, assigning 99%, 95%, 75% PSR players evenly to two teams still requires those ratings to be accurate for it to work. What if the ratings today are inaccurate? A 90% rated player could be 70% in reality, and a 80% player could be 95%. Your MM wouldn't function in that case right?

All the evidence I have points to the the current PSR numbers being bad, and a Matchmaker with inaccurate PSR ratings can't create balanced teams and good matches as a result. Therefore I don't think changing the Matchmaker, while leaving the PSR, would make much of a difference.

P.S. I think you edited in a note about PSR at the end, but yeah, basically from my analysis PSR is 90% of the reason matches are bad.

I don't think anyone looking at the current PSR system can argue it doesn't do a good job of measuring skill. It might be somewhat better than it was before the rework, but it isn't zero sum and still has all sorts of issues.

That said, there are plenty of matches people post that show vastly higher PSR ratings on one team than another. Even if current PSR is a poor analog for skill level, it still has _some_ correlation. While a WLR based skill rating is better, secondary lobby balancing with current ratings would improve some portion of matches. What percentage? I don't know. But we certainly don't look like we're going to get a PSR/MM rework that we should.

Frankly, as you know, I'd settle for random matchmaking without groups. At least then it's a roll of the dice rather than certain types of players getting special treatment by the MM.

[Nightbird]

Anomalocaris, on 22 December 2020 - 02:41 PM, said:

That said, there are plenty of matches people post that show vastly higher PSR ratings on one team than another.

There isn't. If you're talking about matches that look up players on Jarl's rankings, Jarl's is based on average Match Score and very different from PSR ratings.

Edited by Nightbird, 22 December 2020 - 03:52 PM.

[Anomalocaris]

Nightbird, on 22 December 2020 - 02:49 PM, said:

There's isn't. If you're talking about matches that look up players on Jarl's rankings, Jarl's is based on average Match Score and very different from PSR ratings.

One is an analog for another over a short time horizon, are they not? I know PSR can by skewed games/time, but hell, once you pick the 24 players, then just secondary balance using AMS if you like (although we know WLR is best).

[Nightbird]

Anomalocaris, on 22 December 2020 - 03:50 PM, said:

One is an analog for another over a short time horizon, are they not?

Not really. For example, everyone over 300avgMS on Jarl's and over 500 games over the past 6 months probably has the maximum 5000 skill rating in Tier 1. Are they all equally skilled?

Edited by Nightbird, 22 December 2020 - 03:59 PM.

[Anomalocaris]

Nightbird, on 22 December 2020 - 03:58 PM, said:

Not really. For example, everyone over 300avgMS on Jarl's and over 500 games over the past 6 months probably has the maximum 5000 skill rating in Tier 1. Are they all equally skilled?

Good point. I'd forgotten that PGI capped the min and max for PSR. Which makes it useless at some point. Oh well, back to a rolling 3 season avg of AMS or just use WLR!

[crazytimes]

Nightbird, on 22 December 2020 - 03:58 PM, said:

Not really. For example, everyone over 300avgMS on Jarl's and over 500 games over the past 6 months probably has the maximum 5000 skill rating in Tier 1. Are they all equally skilled?

I'd say the benchmark is a fair bit lower than that, I've been pegged to 5000 PSR for a while now and I'm only averaging about 250.

I'd also concur that I am not even remotely in the same skill league as actually good players. I can't even get close to a 1:1 wl or kd.

[morosis]

Nightbird, on 22 December 2020 - 02:12 PM, said:

Thank you for taking the time to describe your idea, I think I understand what you are asking for. The main issue I have complained (and provided evidence) about the current PSR is that two people with the same PSR number, or the same Tier, do not have the same skill level. Using your wording, assigning 99%, 95%, 75% PSR players evenly to two teams still requires those ratings to be accurate for it to work. What if the ratings today are inaccurate? A 90% rated player could be 70% in reality, and a 80% player could be 95%. Your MM wouldn't function in that case right?

All the evidence I have points to the the current PSR numbers being bad, and a Matchmaker with inaccurate PSR ratings can't create balanced teams and good matches as a result. Therefore I don't think changing the Matchmaker, while leaving the PSR, would make much of a difference.

P.S. I think you edited in a note about PSR at the end, but yeah, basically from my analysis PSR is 90% of the reason matches are bad.

i truly do understand where you are coming from. I am by no means a fan of the current PSR system. However, it strikes me that PGI is either unwilling or unable to do a ground up redesign of the system. If that is the case, then we must operate with the constraints of the KPI we have, and unfortunately that leaves precious few options.

to your question: "A 90% rated player could be 70% in reality, and a 80% player could be 95%. Your MM wouldn't function in that case right?" It should, after an adjustment period. In fact I believe this is one of the advantages of a system such as the one I described. At present there are a lot of group-only players who have had a massive advantage in the current MM/PSR system, and generated inflated stats as a result. Many, but not all, of those players would be the former case you mentioned, and they would likely have a much harder time achieving the same stats in the new system. Over a period of a few hundred games, I think their inflated rating would normalize to something much more akin to their skill level. OTOH, solo-only players who have been put at a tremendous disadvantage in the current system, should see the opposite happen, and their rating could rise as the fights get fairer and teams are more balanced.

In the end there are two results I am seeking: that the solo experience feels better, and that matches get more balanced. i think that as people are forced to carry the weight of their PSR as individuals, their PSR rating, for all its flaws, should more likely represent their ability to contribute to the team as an individual. If we are able to achieve that, again, it can do nothing but help.

[Egg Fu]

morosis, on 22 December 2020 - 12:28 PM, said:

I appear to be "doing better" only because i have increasingly refused to play as a solo, because the solo experience is truly unplayable.

Would it surprise you to learn that I can go 20-0 with a 5:1 KDR if i play in a group of people around my skill level, and also go 0-10 with a 1.5:1 if i play alone? That is, in fact the case. I wish it werent, but it is.

Ok I get you now. Disregard my above post.

[Nightbird]

morosis, on 23 December 2020 - 06:51 AM, said:

"A 90% rated player could be 70% in reality, and a 80% player could be 95%. Your MM wouldn't function in that case right?" It should, after an adjustment period. In fact I believe this is one of the advantages of a system such as the one I described. At present there are a lot of group-only players who have had a massive advantage in the current MM/PSR system, and generated inflated stats as a result. Many, but not all, of those players would be the former case you mentioned, and they would likely have a much harder time achieving the same stats in the new system. Over a period of a few hundred games, I think their inflated rating would normalize to something much more akin to their skill level. OTOH, solo-only players who have been put at a tremendous disadvantage in the current system, should see the opposite happen, and their rating could rise as the fights get fairer and teams are more balanced.

In the end there are two results I am seeking: that the solo experience feels better, and that matches get more balanced. i think that as people are forced to carry the weight of their PSR as individuals, their PSR rating, for all its flaws, should more likely represent their ability to contribute to the team as an individual. If we are able to achieve that, again, it can do nothing but help.

I can only say that this is wishful thinking... without accurate PSR, your system would not create better matches over any amount of time.

Edited by Nightbird, 23 December 2020 - 08:33 AM.

[morosis]

Nightbird, on 23 December 2020 - 08:32 AM, said:

I can only say that this is wishful thinking... without accurate PSR, your system would not create better matches over any amount of time.

i know you've looked at the PSR problem more than just about anyone, so i dont want to suggest that you're wrong in your mistrust of PSR.

all i know is that over time, if people were forced to carry the weight of their PSR (or better yet, a KPI derived from actual recent performance that can function instead of PSR for matchmaking purposes), without the ability to hide in premade groups, it will separate those that are carried to their current standing through the enormous advantage of group play from those who are legitimately that skilled as individuals.

in a world where PSR values are bounded at the top end, there is no way for the top tier to clearly separate itself from the pack. throw in that many of the top tier players also self balance by playing underperforming chassis and meme builds, and it makes the situation even more difficult to quantify.

where the system would (hypothetically) add value though, is that something like jarls is just a simple calculation based on KPI that the current MM already has access to. with a little code, you could duplicate a calculation like jarls (or another scaled calc based on KDR, WLR, avg MS) on the fly, for the 24 players in a lobby, for a relatively recent time horizon (say rolling last season + current season) and produce a more or less reliable recent evaluation for the secondary lobby balancer to take in for its decision making. the better news is that over time, these derived values would essentially replace PSR/Tier for matchmaking purposes. and while that is not ideal, i cant see any way that it wouldnt produce better matches than current. frankly, zero matchmaker would be better than current.

Edited by morosis, 23 December 2020 - 12:08 PM.