357 replies to this topic

[Tarl Cabot]

Question then.

As long as MM is being seeded by a player(or grouped players) then MM pulls people together who are available to form teams. And if a person is a low scorer in drops with higher end personnel, and is moved down, then it is feasible his MS increases so he is no longer the low scorer against other low scorers who have also moved down, seeing fewer high end matches.

I have two players whose current avg MS is 171 MS, one bounces back/forth between Tier 2 and Tier 1 while the other is in Tier 3.

Player 1 - avg 171 MS / 0.83. W/L ratio / 23% survival rate / 28k games Tier 2/1 (bounces back forward)
This player's avg MS exceeded 220 MS for at least 2 months (1600 games) and 190 -199 MS for a few months (2K games) and primarily plays heavies/assaults.

Player 2 - avg 172 MS / 0.86 W/L ratio / 30% survival rate / 26k games Tier 3
This player's monthly avg never exceeded 190 MS and primarily plays lights/meds

Now lets put the PSR values being closer in value for Win/Loss. Wouldnt the above players reside more so in Tier 4/5 while most of the population would reside in Tier 3, floating back/forward between Tier 2 or Tier 4?

Most of the charts and others assumptions is based on the MM being opened to Tiers majority of the time, yes? But using Jarls though is also based on a fallacy since the current PSR values have an upward movement, provided a player hits the necessary threshold and plays enough games, meaning said person on the lower end of the bell curve would be grouped and facing tougher opponents, dying quicker (more pinpoint/concentrated damage) while targets also die quicker, thus not allowing him to hit targets more often.

How would said players, with the time and experience they have put in, as well as any shortcomings they may have to play which affects their play, fare in a less crushing environment more often than not?

Take myself. Piloting same mechs 99% of the time between both accounts, my main's stats this month and previous recorded months is worse while in Tier 1 than my alt in Tier 2 by a huge margin. Just on avg MS, 253 MS vs alt 305 MS. Part may likely be accounted for having Tier 3 more so than Tier 1 if MM opens up. That 52 MS spread is a pretty difference, not as huge as 100 MS spread since it is an overall avg on both Wins/Losses.

[Too Much Love]

I'm in favor of 1a option.

It's the most clear, precise and at the same time understandable and simple method. Moreover, it could be easily verified, contrary to the rest options, based on dynamic variables.

The win/loss metric is the worst.

It keeps player hostage to the arbitrary conditions of win/loss, so your individual input doesn't matter. What would matter is the random flow of the events and random matchmaking. It also produce a vicious circle: your ELO depends on the performance of random team and the comparison of the next team is also based on the previous random team etc. In other words, there would be no constant parameter to base calculation on.

So I would prefer the metric that a) depends on my own performance is clear and coherent.

[Cluster Fox]

Jay Z, on 17 June 2020 - 04:38 PM, said:

I know I know I know WLR is a great metric but with 0 access to global or stored statistics it is outside of PGI's scope unless they change their scope, at which point hey WLR becomes a thing.

True, no direct access to WLR. But the higher the PSR gain splits between any Win and Loss, the closer we are to factoring in WLR. It's like saving the WLR data into the PSR system.

Edited by Cluster Fox, 17 June 2020 - 08:38 PM.

[Nightbird]

For those who don't want Win/Loss, keep in mind... what is it... 55, 60% of the player base? Are below average in MS. Tier 5 will be welcoming you with open arms

This applies to all choices of course.

[MrMadguy]

My opinion stays the same. MS shows personal skill. W/L shows teamwork skill. What skill do we need to measure? Personal or teamwork? Or both? And also. Currently MS is CORRELATED with W/L. Winning team usually gets higher MS due to overall longer life and having some victory bonuses:


Implementing W/L-based MM would cause double-dependance between W/L and PSR. Should we really do it?

Why DMG modifier should be decreased? Because pure DMG =/= skill. Kills are. Some players (including me) use some completely ineffective passive-DMG weapons just for lulz. LRMs, MRMs, etc. They aren't lethal due to too much spreading. And due to too high DMG modifier our PSR becomes overinflated. This just isn't right.

Edited by MrMadguy, 17 June 2020 - 08:54 PM.

[Sitzheizung]

Keldomet, on 17 June 2020 - 02:56 AM, said:

TLDR: vote fore JAYZ variant 2B
rest is not viable

i actually switched to variant JayZ 2C
it looks just the best.

[Fenrir-1]

Nightbird, on 17 June 2020 - 08:51 PM, said:

For those who don't want Win/Loss, keep in mind... what is it... 55, 60% of the player base? Are below average in MS. Tier 5 will be welcoming you with open arms

This applies to all choices of course.

I tend to score anywhere between 350-500 in MS, thank you very much. Keep your blanket statements to yourself, ForumWarrior.

[AdmiralAmazing]

I'd vote for:

2C (JZ) [with a factor C > 0.5. I'd prefer around 0.65.]

Pretty fine for me are in order from most acceptable to least:

2B (JZ)
Kami (JZ)
Kami

Unhappy for the outcome of this great opportunity would make me these:

1B (JZ)
Russ Post
Somaru
1A

Edited by AdmiralAmazing, 17 June 2020 - 09:56 PM.

[Kamikaze Viking]

AdmiralAmazing, on 17 June 2020 - 09:49 PM, said:

I'd vote for:

2C (JZ) [with a factor C > 0.5. I'd prefer around 0.65.]

Pretty fine for me are in order from most acceptable to least:

2B (JZ)
Kami (JZ)
Kami

Unhappy for the outcome of this great opportunity would make me these:

1B (JZ)
Russ Post
Somaru
1A

yeah thats pretty much my order of preference too.

Jay Z's proposal is really the result of input from a lot of good people. With some amazing insight on his part on HOW to implement things. (eg when i was explaining mine, he took a bit and then something clicked after I re-stated it in a different way, 2 hours after he got home from work, he had mine working in his demo, and had already applied the tweak to it. Freaking amazing work Jay.).

Edited by Kamikaze Viking, 17 June 2020 - 10:15 PM.

[General Solo]

Sjorpha, on 17 June 2020 - 07:27 AM, said:

I appreciate that you are trying to solve the divergence issue, but it seems to me that a psr formula that needs periodical adjustment of this kind has some fundamental issues.

Agreed!
So PGI need to reset as needed (semi-regular) to keep PSR accurate within margins, they can make an event of it.
Touch in cheek "Corgi Invasion, put my PSR down NAOW!"

Due to limited scope, we just need something that biases slowly so the new system don't need to be non divergent to infinitly and beyond, just long enough that you don't have to reset too often that its a pain.

As a semi active dude, my game count is 50K games since closed beta. So if a reset is needed every 15K is that so bad?

Reset Necessary medicine for good MM health for a system which simply needs resets for further development vs another needing more work.

My big fear is this.
You maths guys are great, but before you came up with your works, your obviously come up with a requirements spec about what the MM should do before you started building.

What were those requirements?

Did those requirements take into account for a mixed environment, with soloes and groups?

Did they take into account both the solo and group demographic perspective? Is it biased.

See your number could be right for your requirement spec but is your requirement spec right?

Edited by OZHomerOZ, 17 June 2020 - 10:55 PM.

[Magic Pain Glove]

I like both 1A( Somaru) and 2B(JayZ) . 2B is probably better.

[General Solo]

MisterSomaru, on 16 June 2020 - 10:37 PM, said:

Why peeps pickin' 1A? That one's mine, and I'm dumb.

I like yours, its the same as mine.
Its simple and treats pilots the same whether they are solo or grouped.
After all its a pilot skill thing not a group thing.

If I were to build the match maker I would do 1A with resets and then using the Solaris MM system using PSR instead of Tier when constructing teams.

But that's another story.

[Jay Z]

MrMadguy, on 17 June 2020 - 08:53 PM, said:

Hey mate, I threw your screenshots into the sheet so you could see what the PSR shifts would be in the various systems. You are pilot 12 in both screenshots. The sheet is available for anyone to put their own match data into for their own viewing.

[General Solo]

Nesutizale, on 17 June 2020 - 12:47 AM, said:

Haven't time to go into the depth but from my gut feeling Core 1A: Global Compare is the way to go.

To me it seams to be the most fair system and the one that is mostly working for "messuring" and not in the way of "rewarding" someone.



Wouldn't this just be the case when you give out extra points as MS Kicker at the end for winning or loosing?
Like this

350 point player on loosing team gets +50 MS for loosing = 400 MS final result
200 point player on winning team gets +250 MS for winning = 450 MS final result
Second player would move up further even though his team lost because of the winning bonus.

If this is the case you (Paul) are talking about....kill the bonus !
When PSR is about rating, placeing and matching individual people then there shouldn't be a winning or loosing bonus as this is a bonus that is team dependend and you would mix a team value into a single persons skill value. Something he can neither prevent or force to change....well at least in 99% of the cases. There is allways the one guy that seams to carry a team but that guy will allready have a very high matchscore anyway.

Maybe change the name to Personal Score Rateing

In my mind pilot is personal, unlike group.

PSR

Not GSR

Edited by OZHomerOZ, 17 June 2020 - 11:24 PM.

[Bistrorider]

Nightbird, on 17 June 2020 - 08:51 PM, said:

For those who don't want Win/Loss, keep in mind... what is it... 55, 60% of the player base? Are below average in MS. Tier 5 will be welcoming you with open arms

This applies to all choices of course.

Really? Please look at this
https://docs.google....#gid=1739121335

My theory is: if you're gonna choose something like 2A team compare or 2B team scalar - and remember what you said: 55-60% players are below average - we all, or almost all gonna end in tier 5. And thats gonna be freaking awesome. One dc here, one afk there. Well some will be lucky.

[General Solo]

Capt Deadpool, on 17 June 2020 - 10:48 AM, said:

Well, groups (or more specifically, some groups) are going to be dominant no matter what tier they are placed in, it's just they will likely be matched more often against other groups and less often against full teams of solos if 2A/2B is used.

That's a bit inaccurate, the top whatever percent of groups will dominate, Totally!!!, DoH their are plenty of non dominant groups playing at the moment.

Got 9 kills yesterday or so first time in ages, some are just not that dominant.
A little secret, only good groups play like a group and they are a percentage, not the majority, others play as a collective of individuals very very often.

Sure I'm losing more but I'm still doing decent, when I'm not blind with RAGE with groups in the queue!

Capt Deadpool, on 17 June 2020 - 10:48 AM, said:

Being placed in almost the top tier but not the very top doesn't mean good solo players have doward bias, and I think many good solo players will end up there in 2A/2B, and there will definitely be good solo players in top tier in 2A/2B, but it will be a different level of skill than we see in current Tier 1.

Why not have solos at the top, why bias against them?

Firstly top level solo's are more competition than non dominant groups or in effect Seals of Soup Queue.
I have seen group players in SQ before than they are just players, variable.

Edit: It makes being in a group less meta and necessary, if you are a 90% Jarls list solo, you have other 90% Jarls list solo team mates to help you counter strong groups.
Rather than non dominant groups and the seals of sop queue as is currently the case when a suitable counter group is not available. /Edit

A high PSR soloes easily make their own ad-hoc groups on the fly, they don't always win, but they often do unlike the non dominant groups or in effect Seals of Soup Queue.

We know in a mixed environment Win/Loss favors groups as a metric?

So working from that is illogical in a mixed environment due to bias of solo's vs groups to PSR movement simply by virtue of being .solo's or grouped, and totally not game play or skill related

Edited by OZHomerOZ, 18 June 2020 - 12:43 AM.

[Mat Sorkas]

I vote JayZ's 2C.

@Paul Inouye, have you and your team addressed the divergence of all the proposed systems? It will not be bad initially but over time will present itself like more an more of an issue.

[General Solo]

Anomalocaris, on 17 June 2020 - 12:33 PM, said:

Still won't help. The population is too small to prevent the top half the of the population from facing upper crust players in most matches. And its getting smaller by the day.

You are right yet imo something can still be done.

I'm thinking that this PSR update is just the first step.

It's purpose is to ensure PSR more accurately gauges a players skill.

But that's just one part of the Match Maker pipeline, so the next step would be to see how MM actually makes the teams.

Does it distribute the skill evenly between teams. With a player pool of 24 players how well could it construct two teams.

Should scale up



Ulitmately those PSR are converted to Tier which is a pretty coarse metric.
Maybe programmer dude can make it switch to sort by PSR when numbers are low.
And as the population is low and not many games are running the overhead could be acceptable.

Edited by OZHomerOZ, 18 June 2020 - 12:45 AM.

[General Solo]

Nightbird, on 17 June 2020 - 08:51 PM, said:

For those who don't want Win/Loss, keep in mind... what is it... 55, 60% of the player base? Are below average in MS. Tier 5 will be welcoming you with open arms

This applies to all choices of course.

Yeah but you have a secret skill
Smells like Comstar
Edit : T5 Ha Sure

Edit the 2nd: Come to think of it.
Maybe those Tier 5's will be having a lot of fun never having to meet me or worse.
Could they even create Tier 5 groups, their own sub-cultre unknown to us decents and Lords.

Wanna buy a mech pack?

Edited by OZHomerOZ, 18 June 2020 - 01:30 AM.

[Nesutizale]

Question, do we have a list of all the MS Kickers and what score they give?

Just asking because someone brought up the "we need to reward teamplay" part over pure damage more.

So I thought what MS Kickers do we have that are more team orianted?
I thought of stuff like, Scouting, Flanking, UAV (starting and shooting down), AMS/ECM cover.

Also had an idea about two that I would like to see, don't know if that could be done though.

"Focused fire"
Fire at the same target, withing 1-2 seconds, as a teammate rewards higher points then doing damage alone.

"Takeing the hit"
Walk into the line of fire of an enemy to protect a teammate gets you pionts.