Jump to content

Psr Update And Changes - Jun 2020


490 replies to this topic

#81 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 04 June 2020 - 01:05 PM

View PostAndrzej Lechrenski, on 04 June 2020 - 12:59 PM, said:


This is a valid point about the group drops, but I think the unfair part of it has more to do with the handicap it places on the team with the potato-carrying group than it does with somebody getting something they don't deserve (although the psychology behind why somebody would want to game a system just to make life harder on themselves is a bit baffling to me altogether).

On the other hand, a group bringing such a weak player will also suffer for it, over a large enough number of matches, and drop in tier as well. If neither of those things happen... well, a win is still a win.


A fair point. My biggest concern is that groups are a non-random element in a system that depends upon randomness to work. I haven't seen an inkling of a suggestion that PGI is prepared to address this.

#82 MechWarrior414712

    Member

  • PipPipPipPipPipPip
  • 446 posts

Posted 04 June 2020 - 01:06 PM

View PostGerdau, on 04 June 2020 - 01:02 PM, said:

Ditto!

Better rewards for INDIVIDUAL PERFORMANCE, please!

[color=#959595]A player winning with a low match score should drop. They were a drag on their team and only won because of their teammates. Same thing with a losing player scoring 400+. They did more than anyone else to win the match but were penalized by their team.[/color]


Personally I think win should never reduce your rating BUT instead if you worry someone doesn't do anything instead raise the required score needed to gain. People can easily do 0-100 damage (very low match score) if they play slow assault like Anni & game ends in less than 2 minutes cuz of a stomp. Shouldn't penalize if they don't even HAVE to do anything to win.

#83 Termin8rSmurf

    Member

  • PipPip
  • Wrath
  • Wrath
  • 45 posts
  • LocationNew Zealand

Posted 04 June 2020 - 01:07 PM

A good scout/spotter/NARC/Tag/UAV placement etc should be rewarded well too. Things like these are often critical to the team's ability to secure the win. I've seen some awesome gameplay by lights or mediums who were spotting, tagging, NARCing, Popping UAVs, etc, which were absolutely critical to the team winning, but they were rewarded with a ridiculously tiny match score, while those who secured all the kills got high scores.

Perhaps some kind of adjustment to how match scores are calculated.

Someone who sits at the back in full LRM-Tard mode, flinging LRMs across the map, cleaning up a 500+ match score, while the guy who puts hit tiny metal butt in the sling to get the locations of the enemy, to risk a sudden fiery death from discovery behind the lines, Narc/tag/visual spotting and placing UAVs on the enemy position, should be rewarded much higher. Make the brave little fellow's rewards suit his contribution to the team, and make the LRM-Tards actually have to work much harder to achieve a decent match score.

I hate LRM-Tards as much as the next man, but can often be found in a cheeky narcer or tagger, bringing in the vital info for them. Reward the brave little mechs more than the LRM boats.

#84 MechWarrior414712

    Member

  • PipPipPipPipPipPip
  • 446 posts

Posted 04 June 2020 - 01:10 PM

View PostTermin8rSmurf, on 04 June 2020 - 01:07 PM, said:

*snip*


Yikes. Having narc doesn't mean you can't do damage. It's time to stop playing a comp narc build, it's easy to have 4mpl and narc on most useful light/pseudolight.

Besides shouldn't reward something that takes literally no skill; narcing is literal no skill in quick play.

#85 Bradleyusm

    Member

  • PipPipPip
  • The Patron
  • The Patron
  • 50 posts

Posted 04 June 2020 - 01:10 PM

If matchscore is going to be used as the metric, it should be top 12 players in a match(regardless of win/loss) should go up and bottom 12 players in a match should go down. This is true zero sum and accounts for win/loss in the fact that winning gives you an automatic match score bonus. Surely sometimes you will die early doing something that wins the match for your team but overall you will go up if you are playing well consistently. I know all of the players that I know and respect as being good will be tier 1 with the above system and most if not all bad 'tier 1' players will never make it back to tier 1 with the above system.

I'm sure there are tons of things wrong with my idea and matchscore is probably not the best metric but just an idea.

#86 Ensaine

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • Bad Company
  • 831 posts
  • LocationOn a frozen rock .....

Posted 04 June 2020 - 01:11 PM

The system should depend on individual performance, not the team.

#87 Andrzej Lechrenski

    Member

  • PipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 96 posts

Posted 04 June 2020 - 01:14 PM

View PostPaul Inouye, on 03 June 2020 - 12:09 PM, said:


[color=orange]Regarding Stomp Numbers: [/color]

But first, let's look at some numbers. We mentioned stomp stats that are matches that end with a score of 12-4 or worse for the losing side. While this metric isn't perfect, it does give us a good baseline for comparison purposes.

When we first implemented the queue merge, we mentioned that the current stomp rate increased by 5%. But as some of you mentioned, that 5% was not clear as to the number of actual stomps, just an increase in an unknown value.

How did we get the 5% increase? We took an average across a number of days prior to the merge and a number of days after the merge.

Prior to queue merge, our sample set gave us this:

1123 Stomps out of 3755 matches in an average 24hr period resulting in a 29.91% stomp rate.

Post queue merge, our sample set gave us this:
935 Stomps out of 2751 matches in an averages 24hr period resulting in a 33.99% stomp rate.

This was the 5% increase we saw.

When we first made some tuning changes to the match maker for Tier/PSR balance and tonnage balance, the sample set showed this:

1801 Stomps out of 5506 matches in an average 24hr period resulting in a 32.71% stomp rate. A 1.28% decrease in stomps.

When I did the stealth tweak to Tier/PSR and tonnage balance in the match maker, it resulted in this:

1057 Stomps out of 3453 matches in an average 24hr period resulting in a 30.61% stomp rate. A bigger reduction from peak of 33.99% down by 3.38%.

Essentially, overall stomp numbers increased by only 0.7% compared to pre-queue merge.



I very strongly disagree with how the stats were handled here. Percentages can't just be subtracted like that (although, even with subtraction, I'm not sure where the quoted 5% came from). 33.99/29.91 = ~ 1.136 or a ~13.6% increase.

Please just report the straight numbers in the future. I'm sure you are aware that a lot of people are looking for ways to call you guys dishonest, and handling things this way doesn't help that.


Quote


To change this, we will be implementing the following:

Current PSR values:

Player LOSES:
Match Score: 0-100 goes down in PSR by -2
Match Score: 101-250 goes down in PSR by -1
Match Score: 251-400 does not move.
Match Score: 401+ goes up in PSR by +1

Player WINS:
Match Score: 0-100 does not move.
Match Score: 101-250 goes up in PSR by +1
Match Score: 251-400 goes up in PSR by +3
Match Score: 401+ goes up in PSR by +5

New PSR values:
Player LOSES:
Match Score: 0-100 goes down in PSR by -5
Match Score: 101-250 goes down in PSR by -3
Match Score: 251-400 goes down in PSR by -1
Match Score: 401+ does not move.

Player WINS:
Match Score: 0-100 does not move.
Match Score: 101-250 goes up in PSR by +1
Match Score: 251-400 goes up in PSR by +3
Match Score: 401+ goes up in PSR by +5



To me, "zero-sum" would mean that for as much and as many as one team goes up, the other team goes down. Tying the magnitude of the increase or decrease to match score only accomplishes the goal of "zero-sum" if those match score bins are statistically equal on both teams, which is something I expect is unlikely.

Quote


We are making these changes to build the best matches with the tools and numbers we have.

We will continue to monitor the stomp rate, game win/loss scenarios and tier matching to make sure the new PSR system is working as intended.



I know I sound critical here, but I really do appreciate what you are doing. Perhaps I missed it, but when is the ETA for the PSR reset?

#88 OldSchoolCav

    Member

  • PipPipPip
  • Ace Of Spades
  • 87 posts
  • LocationAustin

Posted 04 June 2020 - 01:17 PM

Would drastically punishing discos who don't reconnect be too much? I get one or more in at least 50% of my games and it's always an assault...

#89 RRAMIREZ

    Member

  • PipPipPipPipPip
  • FP Veteran - Beta 1
  • FP Veteran - Beta 1
  • 183 posts
  • LocationIn the Blob

Posted 04 June 2020 - 01:19 PM

Andrzej said:

Perhaps I missed it, but when is the ETA for the PSR reset?

June 9th

#90 Captain Caveman DE

    Member

  • PipPipPipPipPipPipPip
  • The Carnivore
  • The Carnivore
  • 519 posts

Posted 04 June 2020 - 01:24 PM

first off: the proposed change isn't zero-sum. pls don't call it that. it is an improvement over what was there, but it can (and needs) to be better;

just to spitball 2cents here: do 'places' in the team, winning and losing;

the best in-team gets X% plus when he wins, the 2nd slighly less and so on; the worst 1-4 of the winning team should go down, at least a little.
do the same, vice versa, on the losing side; the best 1-4 performers should go up a little, the rest down in various degrees.



imo it's fair & easy; all that "carrying the worst performers" via 'win means they can't go down' is to distort where a person should stand - and reopens the whole can of worms you had before, again.

better to get an (more or less) 'accurate' assessment where somebody stands, and then a matchmaker can place them where they need to be, to make it a more balanced match.

Edited by Captain Caveman DE, 04 June 2020 - 01:29 PM.


#91 Phits

    Rookie

  • Moderate Giver
  • Moderate Giver
  • 6 posts

Posted 04 June 2020 - 01:33 PM

This kind of win/loss system only works in smaller games where teams are around 5 players. But now everyone has to hope and pray that they drop with either a group or 11 others that know what they are doing. This system will just make it harder for good players to advance if they keep getting dropped in with either 6 or 7 new players that just don't know what they are doing yet or trolls that go around and just bring down the team cause they don't care. One person can not go and carry an entire team vs 12 opponents so if they do really well but still lose cause they were on a bad team they get punished for it. This system would work really well if there was a ranked queue and a regular one but the player base isn't big enough for that so it needs to be based on individual performance not the performance of 11 other random people. Just saying that this will only encourage people to play in groups and it will make solo players suffer for it.

#92 MechWarrior414712

    Member

  • PipPipPipPipPipPip
  • 446 posts

Posted 04 June 2020 - 01:39 PM

View PostOldSchoolCav, on 04 June 2020 - 01:17 PM, said:

Would drastically punishing discos who don't reconnect be too much? I get one or more in at least 50% of my games and it's always an assault...

It's time to not play the game if you feel like you DC so much that you can't climb in rating

#93 L1f3H4ck

    Member

  • PipPipPipPipPipPipPip
  • 738 posts

Posted 04 June 2020 - 01:44 PM

This is good news. I really liked the whole process.

This is what I've been missing a lot over the years. You know it's not really a niche franchise, right? Wouldn't it be sad if this turned out to be too little, too late after all?

Good trend, more of this please.

Oh, and just to throw that out here, stomps are totally in line with lore.

https://www.sarna.ne..._Loss_Groupings

And in my personal experience, the merge has made even stomps more engaging.

#94 Rkshz

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,866 posts
  • Twitch: Link
  • LocationOdesa, Ukraine

Posted 04 June 2020 - 01:48 PM

only one question - why is the calculation of the RSR independent of the Tiers?
it is logical to assume that the higher the tiers, the more difficult it is to get a plus and the easier it is to go into a minus - this is called competition

even with the new PSR calculation system, a weak player with good luck will be able to get Tier 1 and he will be a problem for the team because it is not his tier skill... lrm masters are a good example Posted Image

Edited by Rkshz, 04 June 2020 - 01:55 PM.


#95 OldSchoolCav

    Member

  • PipPipPip
  • Ace Of Spades
  • 87 posts
  • LocationAustin

Posted 04 June 2020 - 01:49 PM

View PostI O O percent KongLord, on 04 June 2020 - 01:39 PM, said:

It's time to not play the game if you feel like you DC so much that you can't climb in rating


I'm complaining about DC's not worried about my PSR due to my inability to stay connected.. .

Edited by OldSchoolCav, 04 June 2020 - 01:49 PM.


#96 MechWarrior414712

    Member

  • PipPipPipPipPipPip
  • 446 posts

Posted 04 June 2020 - 01:55 PM

View PostOldSchoolCav, on 04 June 2020 - 01:49 PM, said:


I'm complaining about DC's not worried about my PSR due to my inability to stay connected.. .


Well i'm sorry dude but in that case if the system drops you down - it's working as intended.

Think about it like this; if you really dc so much that even when you win every game where you don't DC, it pretty much means you're not contributing well over half the time.

#97 pvt Hudsoff

    Rookie

  • Bad Company
  • 6 posts

Posted 04 June 2020 - 02:03 PM

View PostFirewired, on 04 June 2020 - 10:54 AM, said:

Thanks for the follow-up and background information, much appreciated.

Couple of things...

1. If I were to make a suggestion, base your statistical analysis on a larger sample size than just 24-hours.

Human beings, like it or not, live on a week-to-week schedule. We ran into challenges at work with statistical analysis pre-vs-post implementation because the Architect (who was NOT an Engineer) based his results analysis on a Monday result vs. a Tuesday result. He had not collected sufficient pre-change data to compare versus post-change. We insisted the changes be rolled back (they were because we ARE Engineers and we have authority over Architects).

The data was then gathered for a full 7-day period, then changes re-implemented and the data gathered to be compared day-vs-day (as in Monday pre- vs Monday post-implementation, same for Tue-Tue, Wed-Wed, and so on).

If you do anything at the end or at the start of a new month, then only compare that data to those same dates next month, as human beings do chaotic things near the end and start of a month that can throw data analysis off compared to other dates of the month.

This method generally collects and compares reliable data consistently.

Yes, it takes longer to gather and compare the data, but the more reliable data results are worth the time, and it does not require any more work effort than shorter duration data comparisons if automated.

2. I think (I may be wrong) people were hoping zero-sum would mean Win vs. Loss would be completely ignored in PSR movement per-match.

While your proposed changes technically fit into the definition of zero-sum, they still reward a player for performing poorly and winning, and penalize a player for performing well and losing.

MWO already has a system for rewarding poor performance and winning: Cbills. They get more Cbills if they win regardless of their performance. That should really be enough.

PSR is a Pilot Skill Rating of an individual Pilot's performance. It is not a Pilot Win Rating.

If you feel that you HAVE to reward players for winning then add the Pilot Win Rating to their profile page and remove the Tier Rating, that should make them happy.


THIS.

It is all about performing not winning. Match making should be based ONLY on players performance. W/L is irrelevant.
Match Score can be a more or less ('cause it can be exploited) of an indication. W/L should not. PSR should not be driven by W/L ratio at all IMO.

#98 Constalation

    Member

  • Pip
  • Bridesmaid
  • Bridesmaid
  • 17 posts

Posted 04 June 2020 - 02:04 PM

View PostPaul Inouye, on 03 June 2020 - 12:09 PM, said:


1123 Stomps out of 3755 matches in an average 24hr period resulting in a 29.91% stomp rate.

Post queue merge, our sample set gave us this:
935 Stomps out of 2751 matches in an averages 24hr period resulting in a 33.99% stomp rate.

This was the 5% increase we saw.

When we first made some tuning changes to the match maker for Tier/PSR balance and tonnage balance, the sample set showed this:

1801 Stomps out of 5506 matches in an average 24hr period resulting in a 32.71% stomp rate. A 1.28% decrease in stomps.

When I did the stealth tweak to Tier/PSR and tonnage balance in the match maker, it resulted in this:

1057 Stomps out of 3453 matches in an average 24hr period resulting in a 30.61% stomp rate. A bigger reduction from peak of 33.99% down by 3.38%.

Essentially, overall stomp numbers increased by only 0.7% compared to pre-queue merge.



Uhh, you guys do realize that isn't a 5% increase? A difference of % is read as a percentage point. So that is a 5 percentage point increase, not an increase of 5%.

The math is: (endvalue-startvalue)/(absolutevalue of startvalue)*100%

(33.99 - 29.91) / (abs29.91) * 100% = 13.64

So the actual percentage increase is 13.64%, not 5%. The actual percentage point change is also 4.08%, not 5 %.

I honestly don't think a 30% stomp% is a good number to stay at. Preferably we would see less than 10% stomps. I suppose we will see if the PSR change lowers that or not.

Also, is there a specific reason why 5-12 is not considered a stomp? I can't see any discernable difference between a 5-12 or 4-12 match unlike a 6-12 match.

Edited by Constalation, 04 June 2020 - 02:10 PM.


#99 OldSchoolCav

    Member

  • PipPipPip
  • Ace Of Spades
  • 87 posts
  • LocationAustin

Posted 04 June 2020 - 02:08 PM

View PostI O O percent KongLord, on 04 June 2020 - 01:55 PM, said:

Well i'm sorry dude but in that case if the system drops you down - it's working as intended.

Think about it like this; if you really dc so much that even when you win every game where you don't DC, it pretty much means you're not contributing well over half the time.


Not sure whose English is to blame here - it might be mine. I don't have a DC problem. I DC maybe once every 6 months or so. I hate being in matches with players who DC. Would it be possible to penalize them? Maybe automatic -10 PSR for DC? Is that too harsh?

#100 MechWarrior414712

    Member

  • PipPipPipPipPipPip
  • 446 posts

Posted 04 June 2020 - 02:24 PM

View PostOldSchoolCav, on 04 June 2020 - 02:08 PM, said:


Not sure whose English is to blame here - it might be mine. I don't have a DC problem. I DC maybe once every 6 months or so. I hate being in matches with players who DC. Would it be possible to penalize them? Maybe automatic -10 PSR for DC? Is that too harsh?


Well again - it works as intended, you bite the bullet but according to what Paul says they'll drop a lot with what is proposed, discos go down as intended.





11 user(s) are reading this topic

0 members, 11 guests, 0 anonymous users