713 replies to this topic

[Nightbird]

The Teddy Bear, on 12 June 2020 - 12:20 PM, said:

Well, the variance between solo and group W/L, compared to the variance in solo and group MS? Did you ever investigate that?

No because PGI never released data separating solos and groups. According to verbatim reports, WLR in groups increase greatly, think 2-5x higher WLR. avgMS on the other hand doesn't change much at all, people say it remains unchanged.

[The Teddy Bear]

Nightbird, on 12 June 2020 - 12:24 PM, said:

No because PGI never released data separating solos and groups. According to verbatim reports, WLR in groups increase greatly, think 2-5x higher WLR. avgMS on the other hand doesn't change much at all, people say it remains unchanged.

Thanks this is in line with my stomach feeling.

[VonBruinwald]

OneTeamPlayer, on 12 June 2020 - 07:14 AM, said:

How are we intended to come to a community consensus from a 20+ page thread which over the course of a week has had zero developer input on any of the suggestions?

Looks like the "community consensus" is coming from an off-site discord which isn't being advertised:

https://twitter.com/...845618194280448

[Dauntless Blint]

So Player win loss, and global chassis MS or K/D. Because MS is an end of match stat trying to do the poor mans version.

That bit about groups being 2-5x better, quite crucial I mean how can we even begin to have a conversation without stuff like that.

So player W/L is huge. Chassis W/L or K/D or MS is the next. <<( we do have to pick the chassis before the match surely they have global chassis average MS,K/D.)

Because every time you change chassis you corrupt the match maker for several games, so everyone corrupting the mm for several games MM becomes broken.

If you want free consultants we need more information on what data is logged.

Edited by Dauntless Blint, 12 June 2020 - 01:24 PM.

[Z Paradox]

VonBruinwald, on 12 June 2020 - 01:04 PM, said:

Looks like the "community consensus" is coming from an off-site discord which isn't being advertised:

https://twitter.com/...845618194280448

I like idea from that post for MS ratings...














codiene42

@Codiene42
·Jun 11

Replying to
@Paul_Inouye
@NoGutsNoGalaxyand
@russ_bullock
Well, for one thing, the top 3, in order, things that should be accounted for are 1. Solo kills, 2. kmdds, and the 3. kills.

[OneTeamPlayer]

VonBruinwald, on 12 June 2020 - 01:04 PM, said:

Looks like the "community consensus" is coming from an off-site discord which isn't being advertised:

https://twitter.com/...845618194280448

So they're responding more on external media than they are in a thread on their own official forums.

Yeah, that really says it all about this thread and this overhaul, doesn't it?

[Nightbird]

Please like my tweets if you agree and maybe Paul will read them.

Edited by Nightbird, 12 June 2020 - 01:58 PM.

[VonBruinwald]

Nightbird, on 12 June 2020 - 01:57 PM, said:

Please like my tweets if you agree and maybe Paul will read them.

Sorry, I don't have a twitter, if I did I'd be commenting on the "community consensus" being limited to a private group.

[Nightbird]

VonBruinwald, on 12 June 2020 - 02:00 PM, said:

Sorry, I don't have a twitter, if I did I'd be commenting on the "community consensus" being limited to a private group.

Don't worry about it. I was in that group but my idea was rejected because of the reason... and I paraphrase "The community is too stupid to understand it". I have a screenshot lol

[Capt Deadpool]

Nightbird, on 12 June 2020 - 12:09 PM, said:

The reason that average MS is bad is that it doesn't recognize groups at all when creating teams. Playing in a group is a major advantage to a player. You want a MM that recognizes that a grouped player is more dangerous than a solo player. By saying avgMS is stable, you're basically saying it cannot recognize a player becoming more effective and dangerous.

WL seesawing is good because the contribution of the player to the team is seesawing as he changes between playing solo and playing in a group. WL reflects that change, avgMS doesn't.

OR, (assuming individual players cannot be assigned different PSRs tied to each mech they use and whether or not they are playing in a group or solo, based on W/L history rolling avg./post-reset values) utilizing Dauntless's idea you apply some kind of modifier if someone is grouped such as 1.25x multiplier in 2-man, 2x in 3man, 2.5xin 4-man or whatever the data shows so that MM is at least taking into account a PSR that effectively increases as players group up. Multiple PSRs seem ideal so if someone has a 7-1 W/L in a 3-man and 2-1 W/L playing solo MM knows exactly how to rank them...

I am no programmer, however...

Edited by Capt Deadpool, 12 June 2020 - 02:24 PM.

[Nightbird]

Capt Deadpool, on 12 June 2020 - 02:22 PM, said:

OR

If making things more complicated is on the table, there are many ways to improve the MM further. PSR=WLR is simply the K.I.S.S. solution.

[OneTeamPlayer]

Nightbird, on 12 June 2020 - 02:09 PM, said:

Don't worry about it. I was in that group but my idea was rejected because of the reason... and I paraphrase "The community is too stupid to understand it". I have a screenshot lol

Post the screenshot.

[Nightbird]

OneTeamPlayer, on 12 June 2020 - 02:30 PM, said:

Post the screenshot.

Can't, it insults the community. I PMed it to you.

[Poor-Life-Choices]

Nightbird, on 10 June 2020 - 08:29 PM, said:

How would you handle teams of 2, 3, 4?

As far as how the MM organizes players from tiers, for all solo players matches this is already how it does it. The reason it doesn't work is because past MS is a poor predictor of performance on the team. (For the people that haven't hit the T1 cap, total past MS earned is their PSR bar location today.) It doesn't go by the 5 tiers but rather by a more granular number in the background. (The EXP bar line)

This is also why all the proposals based on MS won't work. Past MS is only loosely related to odds of winning future matches, whereas past WLR is strongly related. Picked any match score on Jarl's list (not adjusted) and look up and down the list at the player's WLR. You'll see a huge variance. This is clearer when you take all the data and graph it, showing a clear pattern that MS is inferior to WLR.

I analyzed Nightbirds Jarls list data vs my data for the most recent 6 months and vs the data from my first 6 months.
Nightbirds WLR varies 40% vs his MS which only varies 10%

My current average WLR is less than it was when I first started, but my match score is higher. Hmm, presumably I'm a better player now. In all three cases WLR varied much more over the period vs AVG match score. So explain to me how WLR will do a better job of identifying pilot skill?
https://www.flickr.c...2791@N05/N26583

Edited by Socal Bronco, 12 June 2020 - 03:15 PM.

[JoeCold]

It occurs to me that if the population forces the matchmaking system to mix tiers to any degree, shouldn't the PSR adjustments for the match handicap above and below the average of the match. That is, shouldn't players above the average tier in that match have a penalty modifier to the rest of their PSR calculation and vice versa? Otherwise mixing tiers drives everyone to the extremes, albeit at different rates depending on ones distance from average.

[Nightbird]

Socal Bronco, on 12 June 2020 - 02:57 PM, said:

I analyzed Nightbirds Jarls list data vs my data for the most recent 6 months and vs the data from my first 6 months.
Nightbirds WLR varies 40% vs his MS which only varies 10%

You have to analyze the data for all players to see trends. A single person's data isn't useful.

[Horseman]

VonBruinwald, on 12 June 2020 - 01:04 PM, said:

Looks like the "community consensus" is coming from an off-site discord which isn't being advertised:
https://twitter.com/...845618194280448

It has been advertised... you just missed when and where.
They're working on their proposals. In an environment that's more conducive to actual work as opposed to getting constantly interrupted by armchair mechwarriors who think they have a constitutional right to be in Tier 1.
Don't worry, even if you missed the bus you'll get to see the proposals when they're presented.

Edited by Horseman, 12 June 2020 - 03:08 PM.

[Eatit]

We need more information on how the current MM works in order to make any meaningful evaluation.

My imagining of how it works currently.

One Pool of players each with a PSR value.

MM asks pool to provide 24 players that have been in que longest in chassis broken down by size. 3A-3H-3M-3L per team.
Limited by current tier valve setting.

If the MM can't provide 24 broken down by size it opens up the size limitations to allow variation on what's available. e.g. 4A-4H-2M-2L per team.

If the pool is still too small it opens the tier valve.

My understanding is that the current MM is allowing Tier 1-3, 2-4, or 3-5 normally and opens to tier 1-4 or 2-5 and in extreme cases 1-5.

Once it has opened the size and tier valves and two teams are created it sorts those teams based on Size. It only uses the tiers to create the pool to pull from. It uses size to create the 2 teams. This will create randomly skilled teams solely based on luck of the draw.



Most of the playerbase is in T1 currently and the MM still has to combine 1-3 at best and 1-5 at worst. What is the importance of moving players from T1 to T4 if the MM is going to open the valve and allow them all to be in the same match anyway? And then sort the teams randomly based on size. This still has the possibility of 12 tier one players vs. 12 tier 4 players.



ALL MATCHES ARE UNBALANCED. SHUFFLING THE PLAYER BASE AROUND IN THE TIERS WILL CHANGE NOTHING IF MM IS PULLING FROM ALL TIERS.

Sorting the teams by adding equal numbers of players from each tier on each team will help with this issue. Fixing the PSR rating will help to sort the players but they both need to be done to create balance not just one or the other.

PGI/Paul please correct me if I'm wrong.

[Nightbird]

Horseman, on 12 June 2020 - 03:07 PM, said:

It has been advertised... you just missed when and where.
They're working on their proposals. In an environment that's more conducive to actual work as opposed to getting constantly interrupted by armchair mechwarriors who think they have a constitutional right to be in Tier 1.
Don't worry, even if you missed the bus you'll get to see the proposals when they're presented.

I'm eagerly awaiting the proof behind the proposals

[Zanotam]

To everyone complaining about groups:

A group can generally expected to be maybe 4 slots, 1/3 of your team. In League of Legends however the 'solo queue' lets people queue in pairs which makes them 2/5ths of the team (~6.6666% repeating more!) and yet nobody is bitchign about how it's impossible to accurately rank people in LoL.... in fact, matchmaking in LoL is pretty damn good.

There are exactly 3 meaningful things for the future MM:

1. player population,
2. using something meaningful to predict winrate
3. Effectively composing teams in the most fair fashion to adjust for top-tier and bottom-tier players skewing things.


GROUP QUEUE IS NOT RUINING THINGS AND SMALL GROUPS AS PART OF A LARGER TEAM ARE NOT AN ISSUE FOR A MATCHMAKER!