357 replies to this topic

[OneTeamPlayer]

Kamikaze Viking, on 22 June 2020 - 05:02 PM, said:

Ok People, Time to stop arguing.

https://twitter.com/...221935631118336

"@Paul_Inouye
Full update tomorrow... spoiler.. 2C is up to the plate.
9:49 AM · Jun 23, 2020"


Thankyou Paul. I eagerly await this update, and am ready to test.

Why is this being posted by a forum user from a third party website rather on this site from the developer who started the thread?

Is anyone at the helm of this ship, lol?

Out of 4 options presented, one that wasn't on the list posted by the developers won.

Inherently amusing.

Who is going to take the time to tally up the votes and see what the final tallies were?

Edited by OneTeamPlayer, 22 June 2020 - 05:43 PM.

[Nightbird]

Spoiler, new PSR MM doesn't do diddly

[OneTeamPlayer]

Glancing over the spreadsheet for 2C it looks like this system involves competing against both one's own team and the enemy team to improve PSR?

If that's the case, good to know, looking forward to playing accordingly starting tomorrow.

[ESC 907]

It seems to me that the people that are arguing against a WLR-based ranking are far too short-sighted. SURE, you may get off matches, but eventually everyone should be playing at a tier where they can manage a WLR of 1.0 or as close to it as possible! WLR is a longer-term solution, and easier since PSR has been difficult to balance properly.

[Ridir Semii]

Nightbird, on 22 June 2020 - 05:43 PM, said:

Spoiler, new PSR MM doesn't do diddly

it goes live after tomorrow's maintenance, I bet

[Nightbird]

Ridir Semii, on 22 June 2020 - 06:50 PM, said:

it goes live after tomorrow's maintenance, I bet

Will take at least 2 months to show it doesn't work, and PGI will be like, we did what the community asked for

[Kamikaze Viking]

Nightbird, on 22 June 2020 - 06:52 PM, said:

Will take at least 2 months to show it doesn't work, and PGI will be like, we did what the community asked for

*within the restrictions given
and then advised we can do better IF other things can be changed.
Primarily Team balancing (currently off the table) and MS Kicker adjustments (to be addressed next)

Baby steps and we'll get there eventually. I wonder if you could start running simulations on Match score data for the next phase. (if only there was an easy way to get a dump of example matches)

[Nightbird]

Kamikaze Viking, on 22 June 2020 - 06:56 PM, said:

*within the restrictions given
and then advised we can do better IF other things can be changed.
Primarily Team balancing (currently off the table) and MS Kicker adjustments (to be addressed next)

Baby steps and we'll get there eventually. I wonder if you could start running simulations on Match score data for the next phase. (if only there was an easy way to get a dump of example matches)

Picking the right kickers for each action is entirely a data-driven maximization problem. Easy to solve.. with data. Impossible otherwise.

[Kamikaze Viking]

Nightbird, on 22 June 2020 - 07:04 PM, said:

Picking the right kickers for each action is entirely a data-driven maximization problem. Easy to solve.. with data. Impossible otherwise.

@Paul

Please Sir can we have a dump of match data please? even just a few hours worth?

Obviously we understand your concerns regarding exploitation. but if you could send it to Key trusted members of the community involved in this it would be greatly appreciated.

Edited by Kamikaze Viking, 22 June 2020 - 07:17 PM.

[Nightbird]

Kamikaze Viking, on 22 June 2020 - 07:15 PM, said:

@Paul

Please Sir can we have a dump of match data please? even just a few hours worth?

Obviously we understand your concerns regarding exploitation. but if you could send it to Key trusted members of the community involved in this it would be greatly appreciated.

I don't need the current kickers (what Paul said was exploitable), they're actually meaningless. I need data of matches showing the final score result in kills for each team, and the number of actions taken by each player on both teams. From that I can calculate the kicker for each action that best predicts the final score.

Edited by Nightbird, 22 June 2020 - 07:25 PM.

[Davegt27]

Kamikaze Viking, on 22 June 2020 - 05:02 PM, said:

Ok People, Time to stop arguing.

https://twitter.com/...221935631118336

"@Paul_Inouye
Full update tomorrow... spoiler.. 2C is up to the plate.
9:49 AM · Jun 23, 2020"


Thankyou Paul. I eagerly await this update, and am ready to test.

cool lets get this show on the road

can we recap and explain what 2C is ??

[General Solo]

Nightbird, on 22 June 2020 - 06:09 AM, said:

Look at the first post in the last thread. The problem is reducing stomps. Not one solution has a mathematical proof showing stomps going down.

Forget 1+1=5, we're dealing with 1+1=Ohio level of thinking here.

Now we are getting somewhere!

The problem is not stomps the problem is skillgap.
So reducing the skillgap is what we need to do.

Stomps are just a symptom of the skillgap problem.

Stomps can occur in this game even when the skill gap is low, due to the snowballing nature of kills once a team has a kill lead.

We have to minimize skillgap; player for player, group for group, team for team.
That will lower stomps, but stomps cannot and will not ever be eliminated.
Its just the nature of the game.

This updates only purpose is to endure that the PSR data for a player more accurately reflects the players actually skill.

This one PSR number will describe the players skill so that MM can construct teams with minimized skillgap.

Which will lead to less stomps.

[General Solo]

Vindicated, on 22 June 2020 - 10:45 AM, said:

You might want to explain why you consider match score to be a fair (unbiased). It obviously favors certain type of mechs.

Easy.

Win/Loss is Biased in favor of groups and bias against non grouped players. Groups are simply a large portion of a team and so have a bigger influence. Is join a group skill, if no what are we measuring?

Matchscore is available to all players equally, grouped or non grouped players have equal access to MS.
Matchscore biases based on player contributions which some call skill. Since we measuring skill in a PSR its a good metric.

Mechs is available to all players equally, grouped or non grouped players have equal access to Mechs.
Mechs building biases based on player mech building skill contributions which some call skill. Since we measuring skill in a PSR its reflected in the matchscore and winning more, but thats relative if grouped winning much much much more.

When solo, MS reflects skill of 500MS performing pilot, when 500MS performing pilot is on losing team better than w/l.
False Negative or something isn't it, when that happens to Mr 500 MS?

[Cluster Fox]

Well, Round 2 it is then...

Here's something I brewed over the weekend. Feel free to download a copy, tweak the values, try things out,
etc.



https://docs.google....dit?usp=sharing

This spreadsheet allows you to rate kickers by contribution to teamplay, then assign multipliers to each group. On page two, you can enter scoring screen data from actual matches and see the comparison.

DISCLAIMER: I do not claim anything with the kicker values included. I entered some values off the top of my head which ended up ballpark similar. Data or proof of concept is necessary to get the kickers to work.

Without data, all those kickers are meaningless, therefore the tuning should be done by those in the know. I strongly hope PGI will release some data on the frequency of each kicker vs match results. I hope this spreadsheet can help a bit.

Edited by Cluster Fox, 22 June 2020 - 09:16 PM.

[General Solo]

Nightbird, on 22 June 2020 - 10:51 AM, said:

Noticed some people bringing up teams in QP. Sorry, you have no choice to vote for that will keep groups from beating down on solo players, since all choices are biased towards groups.

All choices not not biased toward group, some are though.

Choice 1A allocates PSR movement based on player MS performance solely.
Get 500MS or top MS win or lose you gain PSR
Get lowest match score win or lose you drop PSR.
Thats the logic, and it only some what favors groups, just a little do to their overall possible increase in effectiveness compared to soloes.

Sounds fair.

OZHomerOZ, on 17 June 2020 - 02:38 AM, said:

My thoughts are theirs the real possiblity we won't get other chances so this change has to have a big impact in leveling the battlefield in order to firstly retain and maybe grow players

Firstly in my experience if groups are the tier 1 and dominate the queue then it dies like every other group dominated que.

I have two points why I disagree with using win/loss, all this assumes a feasable population or expert team construction with low populations.

Two points

1) By using Win/Loss you are effect causing a downward PSR bias simply because a player is solo, not in a group.
Its like before with poor performing on a win bias but a diferent biased, but still biased!

You may say its a team game and your right, but listen. MATCHMAKER's don't care about team game its goal is produce a metric called PILOT SKILL RATING that ranks a pilot against all other pilots, groups are not mentioned.
Its logic not maths.

We never had a proper functioning group MM other wise it wouldn't be a dead queue. So we are using and modifiying the solo one. But any way biased against soloes when its supposed to be a pilot skill metric, groups had another MM that was crap.

Anyway point two

And if a solo pilots PSR does not reflect their skill simply because they are not in a group, IMO by using win loss you are in fact causing MM to forgo using the very soloes that can best counter groups as far as soloes can counter them, when a suitably high PSR group is not available. Giving groups a further advantage.

imo

Edit: In my experience being in a group don't magically make you a better player. You still the same player.

I think the game should track how many group games a player does, so they can be ignored when they solo and then complain on the forums about matchmaker and losing PSR.

Tell me my two points are wrong, cause I can't see it.

Top level soloes are a menace to groups and if concentrated are a pita, using w/l seeks to prevent that due to bias against solo PSR movement when using match results w/l as PSR metric

People may say Oh Homer just wants his T1 badge?
Your wrong, I just like playing against good players but I ain't going be no cannon folder grouped with some green troops.
But allocated teammates by MM with similar skills to myself, peers I can take on alot. Maybe not win the Galaxy but I don't care, they speak my name that enough.

This is more about for those less able, will they have a good experience or will they leave?
Cause I can't have a game if the servers are off.

When I hear some of the group pre made community I wonder!
After all they killed every queue they every been in. And now they helping design a part of the match maker.
Rah roh

[General Solo]

Laser Kiwi, on 22 June 2020 - 04:16 PM, said:

The issue is, this is a total strawman. Except in exceptionally rare circumstances these mechwarriors do not exist. People with a match score average 340 virtually all have a 1+ win loss, people with 170 match score are all less than 1. You can't argue from a position that doesn't even exist in 99.9% of the population. Maybe you can find one, maybe 5, but in a population of several thousand the vast majority have a positive kdr when the match score is high and vice versa and are by every metric a better player than the person with half their match score.

edit: win loss, not kdr, nobody cares about kdr

You said it your self, high MS corresponds with winning and skill, unlike winning alone where a pilot can get carried which is where the group bias comes from.

MS is not biased apply to all pilots equally.
W/L is biased toward groups as groups are a larger portion of their team (1/6-1/4) and so have a greater influence on the outcome of a match compared to a solo (1/12).

So MS > W/L metric for a PSR.

[Capt Deadpool]

^FYI none of my more recent posts were directed at you, Homer

[General Solo]

Nightbird, on 22 June 2020 - 03:00 PM, said:

On average. If you don't believe me, flip a coin a million times, and you'll get awfully close to 50% heads and 50% tails, despite the possibility of there being much more or less it just doesn't happen.

Did you account for whether the flipper is a T5 flipper or T1 flipper. Will this help our situation? And How?

Edited by OZHomerOZ, 22 June 2020 - 09:45 PM.

[Horseman]

Davegt27, on 22 June 2020 - 08:06 PM, said:

cool lets get this show on the road

can we recap and explain what 2C is ??

It's 2B (JayZ version)

[General Solo]

Capt Deadpool, on 22 June 2020 - 09:43 PM, said:

^FYI none of my more recent posts were directed at you, Homer

Thats noice, but this is bigger than both of us