Jump to content

Psr Community Version 1.0


379 replies to this topic

#21 Jimmy the Jim Jimmies

    Member

  • Pip
  • Littlest Helper
  • Littlest Helper
  • 11 posts

Posted 24 June 2020 - 12:00 AM

The limits on PSR rating may have a side-effect that hasn't been mentioned I think:

When the lowest ranked players play, with some having 0 rating, they can not go lower and therefor the total sum of the PSR changes in that match will be positive, causing an inflation. This isn't necessarily a bad thing but something to be aware of.
It will push lower skilled players into higher tiers.

The same goes for the top: top players can only push each other down.
The model that Nightbird linked shows this as instability, but I think it actually reveals that only players between 1000-4000 rating see the mobility of the new PSR rating.

It will be interesting to observe.

#22 Bowelhacker

    Member

  • PipPipPipPipPipPipPip
  • Hero of Marik
  • Hero of Marik
  • 922 posts
  • LocationKooken's Pleasure Pit

Posted 24 June 2020 - 12:08 AM

You know why win/loss should never be introduced into this sort of thing in quick play? Because of modes like conquest. When ******* just abjectly fail to give a **** about the objective and the other team does. Or when instead of just dipping a toe in the circle, your team pisses off to skirmish merrily three grid squares away from it.

Otherwise, do what you will.

#23 Sjorpha

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,478 posts
  • LocationSweden

Posted 24 June 2020 - 12:09 AM

View PostNightbird, on 23 June 2020 - 05:52 PM, said:

Well, we know the update will fail because:

1. the formula fails to stabilize people at any PSR value other than 0 or 5000 (forget that bell curve, 0 chance of that happening)
2. groups are given a large unfair advantage in the formula, solos get ready
3. number of stomps will not change (and may get worse for two months)
4. 60% of the player base will be sent - fast or slow - to Tier 5 for being below average
5. T5 players are 2.5x more likely to quit MWO

Because Math is not an Opinion.

So rather than waiting for 2 months to see proof of whether 1+1=5, I give you the result now:

Spoiler


P.S. The tier/population graph needs to be players*games to present the relative number of people in each tier as relevant for the queue. See 5 above.


I would very much like to see a response to this from PGI, they seem to be very valid concerns.

#24 Sjorpha

    Member

  • PipPipPipPipPipPipPipPipPip
  • Philanthropist
  • Philanthropist
  • 4,478 posts
  • LocationSweden

Posted 24 June 2020 - 12:18 AM

View PostBowelhacker, on 24 June 2020 - 12:08 AM, said:

You know why win/loss should never be introduced into this sort of thing in quick play? Because of modes like conquest. When ******* just abjectly fail to give a **** about the objective and the other team does. Or when instead of just dipping a toe in the circle, your team pisses off to skirmish merrily three grid squares away from it.


On the contrary, that's exactly the reason why w/l has to be the main factor.

Those players in your example that failed to play the game mode were objectively playing bad in those matches, and that lost them the match. But their avg matchscore might still be high, only their lower w/l reflects the fact that they are bad at conquest.

Edited by Sjorpha, 24 June 2020 - 12:19 AM.


#25 Gagis

    Member

  • PipPipPipPipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 1,731 posts

Posted 24 June 2020 - 12:22 AM

I am absolutely hyped for this!

#26 Nesutizale

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Privateer
  • The Privateer
  • 3,242 posts

Posted 24 June 2020 - 12:24 AM

I have to admit I don't get the math, I just hope it will work out.
What I like is to see that PGI actualy takes in stuff sugested by the community and tries to improve on that.

I hope to see more of that cooperation and maybe at some point a mapmaker for players?
I bet there will be people willing to work on those too.

#27 Nearly Dead

    Member

  • PipPipPipPipPipPip
  • Philanthropist
  • 274 posts

Posted 24 June 2020 - 01:29 AM

Everyone counted three times? Once per month? Plus every tier change counted, ie, PlayerX in May was in tier 4 for two weeks and tier 3 for two weeks so was counted twice that month?

Otherwise I don't get the player base numbers vs what we have heard previously.

#28 ghost1e

    Member

  • PipPipPipPipPipPip
  • CS 2023 Gold Champ
  • CS 2023 Gold Champ
  • 403 posts
  • Twitch: Link
  • Location2023 World Champion

Posted 24 June 2020 - 01:39 AM

View PostNightbird, on 23 June 2020 - 05:52 PM, said:

1. the formula fails to stabilize people at any PSR value other than 0 or 5000 (forget that bell curve, 0 chance of that happening)


well, and what if the current avg matchscore is 222, right in that zone that doesn't go to extremes?
(pulled from https://leaderboard.isengrim.org/stats)

#29 ghost1e

    Member

  • PipPipPipPipPipPip
  • CS 2023 Gold Champ
  • CS 2023 Gold Champ
  • 403 posts
  • Twitch: Link
  • Location2023 World Champion

Posted 24 June 2020 - 01:42 AM

View PostOneTeamPlayer, on 23 June 2020 - 06:06 PM, said:

Given that quote was given to us a few weeks ago by Paul Inoyue, how does the current matchmaker implementation work within the current playerbase limitations, rather than in a massive hypothetical playerbase that does not log in simultaneously or much at all, apparently.


well, it'll definitely be better than it is currently. it'll help MM a lot knowing to separate a multiple-times world champion from a 30th percentile player, which might balance matches more, even with the current MM, as it now knows what players NOT to put on one team

#30 Blechreiz

    Member

  • Pip
  • The Nocturnal
  • The Nocturnal
  • 13 posts

Posted 24 June 2020 - 01:57 AM

View PostNightbird, on 23 June 2020 - 05:52 PM, said:

Well, we know the update will fail because:

1. the formula fails to stabilize people at any PSR value other than 0 or 5000 (forget that bell curve, 0 chance of that happening)
2. groups are given a large unfair advantage in the formula, solos get ready
3. number of stomps will not change (and may get worse for two months)
4. 60% of the player base will be sent - fast or slow - to Tier 5 for being below average
5. T5 players are 2.5x more likely to quit MWO

Because Math is not an Opinion.

So rather than waiting for 2 months to see proof of whether 1+1=5, I give you the result now:

Spoiler


P.S. The tier/population graph needs to be players*games to present the relative number of people in each tier as relevant for the queue. See 5 above.


Sorry, but I think your models aren't entirely correct since they seem to be based on the assumption that the average players performance will be static across all tiers. This is not the case.

Apart from the extremes at the top and bottom of the MS/performance scale, there's a strong indication that people will get higher AMS in the lower tiers 4 & 5 and will get lower scores in tier 1 & 2. This, imo, will prevent the envisioned scenario where you will end up with 60% of active players in T5 and will lead ultimately to a normal distribution.

I think, even though I voted for a different approach, the proposed changes would be a good change. Maybe the implementation and reset could be combined with some longer lootbag-type event to encourage people to play more?

One think that's not really clear to me is the situation of people starting the game; would they start in T3 or T5? Remembering the rather "brutal" start and steep learning curve for me, I think it wouldn't be a good idea to let them start in T3 where they could end up facing T1-players.

Edited by Blechreiz, 24 June 2020 - 01:59 AM.


#31 General Solo

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • 3,625 posts

Posted 24 June 2020 - 02:35 AM

smh give groups even more advantage.
Communication pff
Posted Image

#32 Bud Crue

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Rage
  • Rage
  • 9,943 posts
  • LocationOn the farm in central Minnesota

Posted 24 June 2020 - 02:49 AM

It may have taken a couple of years, but right now, I'd like to think that the blatant absurdity and irony of Russ's quote in my signature block is now just a bit less so.

I for one am pleasantly surprised that this effort appears headed to actual implementation. I still don't think it will change much at this point, but I have to applaud PGI for making the effort to engage the community, and give kudos and respect to those community members who put effort into trying to provide real, workable, solutions. Truly uplifting.

#33 Bistrorider

    Member

  • PipPipPipPipPipPip
  • The Howl
  • The Howl
  • 273 posts

Posted 24 June 2020 - 03:47 AM

Cool, now it's time to check i out. Props for PGI and Jay Z and all... There is nothing to assume right now, just to check it. Some values may be changed. Interesting thing about T-4 to T-1 players.

I'm looking forward for more Posted Image

EDIT: And it sounds sooo good. PSR Community Version 1.0

Edited by Bistrorider, 24 June 2020 - 03:49 AM.


#34 Belorion

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,469 posts
  • LocationEast Coast

Posted 24 June 2020 - 04:45 AM

In my opinion the weights are too heavily counting in the W/L and not enough on the match score, but I am glad you are putting in the new system where the values can be tweaked a little.

Setting the W/L factor to 1 generates a more ideal outcome... again in my opinion.

#35 Belorion

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 5,469 posts
  • LocationEast Coast

Posted 24 June 2020 - 04:51 AM

View PostSjorpha, on 24 June 2020 - 12:18 AM, said:

On the contrary, that's exactly the reason why w/l has to be the main factor.

Those players in your example that failed to play the game mode were objectively playing bad in those matches, and that lost them the match. But their avg matchscore might still be high, only their lower w/l reflects the fact that they are bad at conquest.


A strictly W/L rating will simply drive everyone to the middle and leave very little to the curve. It will favor groups even more than 2C since people that drop together are more likely to affect the outcome of the game. It basically is a faster way of crumpling your metaphorical bridge.

#36 OneTeamPlayer

    Member

  • PipPipPipPipPipPip
  • Ace Of Spades
  • 399 posts

Posted 24 June 2020 - 04:56 AM

View PostBud Crue, on 24 June 2020 - 02:49 AM, said:

It may have taken a couple of years, but right now, I'd like to think that the blatant absurdity and irony of Russ's quote in my signature block is now just a bit less so.

I for one am pleasantly surprised that this effort appears headed to actual implementation. I still don't think it will change much at this point, but I have to applaud PGI for making the effort to engage the community, and give kudos and respect to those community members who put effort into trying to provide real, workable, solutions. Truly uplifting.


A grand total of 6-ish responses in 6 weeks (most of them omni-directional announcements) is an improvement over previous radio silence, but let's not go crazy here.

The community did a great job doing consultant level work at the best possible rate for PGI, though.

#37 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 24 June 2020 - 04:58 AM

View PostTheUltimateGhost, on 24 June 2020 - 01:39 AM, said:

well, and what if the current avg matchscore is 222, right in that zone that doesn't go to extremes?
(pulled from https://leaderboard.isengrim.org/stats)


If the PSR doesn't assign higher static values to people with higher skill, and lower static values to people with lower skill, then it is blind when making teams when trying to put an equal amount of talent on both teams.

View PostBlechreiz, on 24 June 2020 - 01:57 AM, said:


Sorry, but I think your models aren't entirely correct since they seem to be based on the assumption that the average players performance will be static across all tiers. This is not the case.

Apart from the extremes at the top and bottom of the MS/performance scale, there's a strong indication that people will get higher AMS in the lower tiers 4 & 5 and will get lower scores in tier 1 & 2. This, imo, will prevent the envisioned scenario where you will end up with 60% of active players in T5 and will lead ultimately to a normal distribution.

I think, even though I voted for a different approach, the proposed changes would be a good change. Maybe the implementation and reset could be combined with some longer lootbag-type event to encourage people to play more?

One think that's not really clear to me is the situation of people starting the game; would they start in T3 or T5? Remembering the rather "brutal" start and steep learning curve for me, I think it wouldn't be a good idea to let them start in T3 where they could end up facing T1-players.


Past MS performance has a weak correlation to winning, like mech tonnage and KDR. WLR has strong correlation.
https://mwomercs.com...and-suggestion/
Sim tend not to be perfect, but give you an idea of what's to come. The better the sim the more accurate. There is no sim that suggests the idea in the first post works and one that shows it doesn't

#38 Blechreiz

    Member

  • Pip
  • The Nocturnal
  • The Nocturnal
  • 13 posts

Posted 24 June 2020 - 05:35 AM

View PostNightbird, on 24 June 2020 - 04:58 AM, said:


There is no sim that suggests the idea in the first post works and one that shows it doesn't



Well, that's where we have to agree to disagree Posted Image

Problem is, and the reason why I doubt that I would be able to change your opinion, that, in order to come up with a comprehensive simulation of the future distribution of players across the different tiers, I think we simply do not have the necessary data Posted Image

Hence I don't think that the posted analysis can be used to reliably predict the player distribution in, let's say, 3 to 6 months after the tier reset and the implementation of the new PSR calculation.

#39 Nightbird

    Member

  • PipPipPipPipPipPipPipPipPipPip
  • The God of Death
  • The God of Death
  • 7,518 posts

Posted 24 June 2020 - 05:41 AM

View PostBlechreiz, on 24 June 2020 - 05:35 AM, said:

Well, that's where we have to agree to disagree Posted Image

Problem is, and the reason why I doubt that I would be able to change your opinion, that, in order to come up with a comprehensive simulation of the future distribution of players across the different tiers, I think we simply do not have the necessary data Posted Image

Hence I don't think that the posted analysis can be used to reliably predict the player distribution in, let's say, 3 to 6 months after the tier reset and the implementation of the new PSR calculation.


If you're talking about the graph I linked for (1) I agree. Xiphas already said in his post his model is very simplified. It does show that Jay Z's solution diverges, this part is absolutely correct and it's the point I noted in (1). There's no simulation of the expected WLR for players across the skill spectrum nor a prediction of stomps but I did that in my simulation linked. You can note a decided absence of such support for the first post. I think Jay Z tried to do a sim but it confirmed what we know, he can correct me if I am wrong.

Edited by Nightbird, 24 June 2020 - 05:51 AM.


#40 Xiphias

    Member

  • PipPipPipPipPipPipPip
  • Littlest Helper
  • Littlest Helper
  • 862 posts

Posted 24 June 2020 - 05:46 AM

View PostTheUltimateGhost, on 24 June 2020 - 01:39 AM, said:

well, and what if the current avg matchscore is 222, right in that zone that doesn't go to extremes?
(pulled from https://leaderboard.isengrim.org/stats)

It's not the average that matters, it's the individual players averages that actually generate that average. That figure was generated using a distribution of simulated players pulled from Jarl's list data. Only players who are actually right at the average end up staying in the middle over time.

View PostBlechreiz, on 24 June 2020 - 01:57 AM, said:

Sorry, but I think your models aren't entirely correct since they seem to be based on the assumption that the average players performance will be static across all tiers. This is not the case.

Apart from the extremes at the top and bottom of the MS/performance scale, there's a strong indication that people will get higher AMS in the lower tiers 4 & 5 and will get lower scores in tier 1 & 2. This, imo, will prevent the envisioned scenario where you will end up with 60% of active players in T5 and will lead ultimately to a normal distribution.

I think, even though I voted for a different approach, the proposed changes would be a good change. Maybe the implementation and reset could be combined with some longer lootbag-type event to encourage people to play more?

One think that's not really clear to me is the situation of people starting the game; would they start in T3 or T5? Remembering the rather "brutal" start and steep learning curve for me, I think it wouldn't be a good idea to let them start in T3 where they could end up facing T1-players.

If you're referring the the first linked model, that's actually mine and admittedly has quite a few limitations (I plan to build a more robust one given the time).

That said, the absolute matchscore values don't matter because players are not being compared to a absolute matchscore threshold. If we picked values based on historic data to move players up or down that could work, and you would see exactly what you're saying (players would move into a higher tier, get matched against harder players, and get lower matchscores, stopping them from moving up further).

However, in this case relative matchscores are being used, how well a player scores is going to be dependent on how good their teammates and the enemy team are. To move up you need to basically be on average better than half the rest of the playerbase, otherwise you are going to scoring in the lower half more often than not (relative to the rest of your team/enemy team) and that will push you down over time.





2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users