Psr Community Formula V1.0
#1
Posted 30 June 2020 - 09:54 AM
As mentioned, today is the day the new PSR formula kicks in.
Everyone has been reset to a PSR value of 2500 with a minimum of 0 and maximum of 5000. As everyone plays matches, they will start to be distributed based on their performance in the match compared to the other players in the match.
This new formula keeps the amount of PSR movement from a given match zero-sum. I will post updates on the curve that is created for everyone to follow as we examine how the system is working for the next couple of months.
If there are issues, or the formula isn't doing what's expected of it, we can look into the other suggestions that were put forth in the previous discussion threads about PSR.
We are looking forward to seeing the outcome of the current formula and hope to see it work out.
-Paul
#2
Posted 30 June 2020 - 09:56 AM
I think for the foreseeable future most people will be in Tier 3 because it takes >100 games for most pilots to move out of it. If you're able to present a plot with PSR by games_played, it would better show how people are queuing up. This does mean a pilot who played 1000 games will weigh more than a pilot who played 10 games but that's OK if this accurately presents their presence in the queue.
Lastly, a PSR by avgMS graph like the one below would be helpful to show how everyone is moving.
Note: image below is from a simulation, not from actual matches.
Edited by Nightbird, 02 July 2020 - 07:41 AM.
#3
Posted 30 June 2020 - 10:00 AM
Inquiring light pilots would love to know.
#4
Posted 30 June 2020 - 10:11 AM
Paul, two known problems must be addressed as soon as possible. The most important is the stabilization.
If the MM knows a player's PSR before a match, it can stabilize. See the last portion of my thread about how to stabilize the PSR. It uses only two variable inputs: Jay Z's calculated shift and player PSR.
The two big steps forward are:
- Can we make the XP bar stabilize now?
- Can we tune up Jay Z's X,Y,C parameters to values with at least an analysis backing it?
https://mwomercs.com...stabilizing-it/
Priority 1 Making PSR stable
Addresses the PSR function(# matches) divergence and how to stop it.
Priority 2 X,Y,C Tune up
Addresses tuning the X,Y,C to represent Skill fairly across the MS and WLR components
UPDATE:
Both are now available as a nice Google Document:
https://docs.google.com/document/d/1tNau95Nj0KKjHEHfl90c8MqrVpji8GiDAUdrA0T7bLU/edit#heading=h.5a9hn5stz0e1
Links to the full analysis of both stability and tuning included.
Edited by Cluster Fox, 27 July 2022 - 07:37 PM.
#5
Posted 30 June 2020 - 10:32 AM
Edited by Kat Baran, 30 June 2020 - 10:32 AM.
#6
Posted 30 June 2020 - 11:16 AM
#7
Posted 30 June 2020 - 03:37 PM
even though the matches were loses, my performance was still good enough to move up, 'cept that ;last match, too many my team blocking and using me as a meat shield
#8
Posted 30 June 2020 - 08:22 PM
Edited by Tarl Cabot, 30 June 2020 - 08:23 PM.
#9
Posted 30 June 2020 - 08:36 PM
#10
Posted 30 June 2020 - 09:48 PM
@Paul, please leave the old graph in place and just add a new one with a time stamp, so we can see the progress.
#11
Posted 01 July 2020 - 07:25 AM
based on my own experience, with the way the new system is working, I like it.
win or lose, if I preform poorly, I go down, and when I do well I go up.
haven't been in that score/performance range to been carried up on a win yet, hard for me to tell, without looking and doing the math of everyone's scores to see who was carried, I just know if I did bad and see that green up arrow , , I was carried, haven't seen that yet
I get a lot of the no movement on the losing side, the yellow equal, had it a couple times on the winning (probably did something I shouldn't had and died quick, or couldn't hit anything)
this is going to take a little time to get players in their perspective tier levels
as for the new players, I'm not going to disagree they still need to start in Tier 5, but right now, who are they going to match up with, my guess, tier 3's. yes, we're all currently tier 3, till that tier spread is further spread from each tier level, that's where they'll be for right now, At the same time, I don't think they should start with a 0 either, don't think any one started from the the bottom, ( I didn't start playing till 7 months after live started, so if at the git-go, if it did, I didn't know, even though I did not ever pay any attention to tier levels anyways), so starting recruits at 1500 may be the better choice
ETA:
forgot to mention an idea, not sure if it will work here, or suggested already, but is some places when you arrow over a bar, you get the actual numbers, like if I was to arrow over the tier bar, I would be able see my current PSR,
wouldn't mind being able to do that.
Edited by Xaat Xuun, 01 July 2020 - 07:31 AM.
#12
Posted 01 July 2020 - 09:25 AM
The matches in which comp groups are present are entirely unfun, however as if they are on your team they are boring easy wins and if they are on the other team they are boring predictable losses in both instances stomps all the way.
I've had some real nail-biter photo finish matches today so at the very least it's something of a change.
I've also seen full teams with low matchscores i didn't think possible in assault/heavy mechs (sub 50? How?!?) so it's been hit or miss.
There's an actual reason to compete hard again, though so thanks for that!
#13
Posted 01 July 2020 - 10:03 AM
Too many games where a four man group went thru and killed our team leaving literally 13 minutes on the clock. 0-12 makes it sound closer than it was.
Some games where I was wondering why I didn't have anything to shoot and launched a UAV to find a target and realized, oh, games over. 12-1 or 12-2 at best. Reminds me of picking my wife up at work, taking her to lunch at the buffet and finding the Denver Broncos were eating there, and we were behind them.
#14
Posted 01 July 2020 - 10:41 AM
With X = 5, Y=20 and C=1, Match score is about 3X too important. Because those were initially suggested without much analysis, to cater for the "feel good" factor.
Aka two example players:
Player A is 65th percentile MS and 50th percentile WLR
player B is 50th percentile MS and 95th percentile WLR
Player A is treated as the same skill as Player B.
Doesn't seem right to me...
Edited by Cluster Fox, 01 July 2020 - 10:45 AM.
#15
Posted 01 July 2020 - 11:10 AM
Paul Inouye, on 30 June 2020 - 09:54 AM, said:
As mentioned, today is the day the new PSR formula kicks in.
Everyone has been reset to a PSR value of 2500 with a minimum of 0 and maximum of 5000. As everyone plays matches, they will start to be distributed based on their performance in the match compared to the other players in the match.
This new formula keeps the amount of PSR movement from a given match zero-sum. I will post updates on the curve that is created for everyone to follow as we examine how the system is working for the next couple of months.
<<<Image from https://static.mwomercs.com/downloads/PSR_Chart_001.png>>>
If there are issues, or the formula isn't doing what's expected of it, we can look into the other suggestions that were put forth in the previous discussion threads about PSR.
We are looking forward to seeing the outcome of the current formula and hope to see it work out.
-Paul
Hi there... I got trapped out of the loop during some of the discussion. Would you please add what the Formula is to your opening post, or at least reply back to this thread with what the Formula is? Right now, things are not making much sense to me at all, and other's posts are also failing to make any sense to me.
~D. V. "really puzzled about the new PSR Formula and its' effects, as well as puzzled about what others say" Devnull
#17
Posted 02 July 2020 - 06:25 AM
OneTeamPlayer, on 01 July 2020 - 09:25 AM, said:
The matches in which comp groups are present are entirely unfun, however as if they are on your team they are boring easy wins and if they are on the other team they are boring predictable losses in both instances stomps all the way.
I've had some real nail-biter photo finish matches today so at the very least it's something of a change.
I've also seen full teams with low matchscores i didn't think possible in assault/heavy mechs (sub 50? How?!?) so it's been hit or miss.
There's an actual reason to compete hard again, though so thanks for that!
And depending on how many games players put in, you will see lots of those for a few months. Why?
- No Seeding utilized but stuck EVERYONE in Mid-Tier 3/PSR
- Above SOO wrong, primarily for new players who had just started (Paul's infamous graph showing tons of Tier 5 over last 2 months), and for previously new players who may come back to an existing account. Now All of them are in Tier 3... Think about that.. being a newbie and literally being thrown to the wolves.
- Tier 5 players should be reverted back to Tier 5, or at least Tier 4, especially those who may not have recently played.
- Seeding should have been utilized, and if not the wide range then between Tier 2-4
- Groups - Change to Max 3. Prefer max 2. I used to call for this, to allow Duos option to opt into Solo queue w/o opt out of group, but if group size is reduced, I dont see PGI going to max 2 now since there is no separate QP Group queue now.
- Either remove Tonnage, change to 1 per weight class or tighten up the tonnage. Would help change the amount of firepower and armor a "comp" group can bring to the table while ensuring that others bring something more. When faced off, it may not change the end result but would change the negative perception the other side won cause they had 5 assaults vs 1 assault.
#18
Posted 02 July 2020 - 07:30 AM
(No) cbills, no XP, no SP… whatever your skill, it's sounds like a challenge to compete with all the others 2500 PSR.
But what is the % of alt and what is the % of "real new comers" and the % of "real T5" in Paul diag on T5 population… that's the question.
(for instance I have an alt with few games in last months, to remember me how it starts in order to "coach" a friend that migth have joined)
#19
Posted 02 July 2020 - 08:28 AM
When PGI changed from Elo to PSR/Tier, players were seeded based on their Elo rating. Here, using PSR rating it would not have worked due to the how their previous PSR was setup but PGI did have other stats to use, primarily avg MS over X amount of game at least.
The bad part will be the large number of players who are low PSR who do not play many games per month, come back and get dropped into Tier 3 during this time period... Not good for Player Retention.
And just to repeat it.. Player Retention.. The PSR change may be good for player retention down the road but the initial setup to dump everyone into Tier 3 instead of spreading them out, even if 2-4, will likely not be good for player retention.
#20
Posted 02 July 2020 - 11:26 AM
Edited by Dionnsai, 02 July 2020 - 11:26 AM.
4 user(s) are reading this topic
0 members, 4 guests, 0 anonymous users