Jump to content

Psr Update And Hold On Patch.


713 replies to this topic

#221 cougurt

    Member

  • PipPipPipPipPipPipPip
  • CS 2023 Silver Champ
  • CS 2023 Silver Champ
  • 691 posts

Posted 09 June 2020 - 04:17 PM

View PostHorseman, on 09 June 2020 - 02:20 PM, said:

Which lead to matchscore anyway. Accuracy translates to damage and kills. Ability to spread damage translates to more scoring opportunities.

yeah, no need to overcomplicate things. the only issue i see with the way match score is currently calculated is that it's a bit indiscriminate in how it rewards damage. i don't think it's a big deal since generally speaking if you're doing a lot of damage it's likely to be a fairly hefty contribution regardless, but obviously there's a little more nuance to a player's performance than sheer damage output. the problem with trying to reward non-combat actions or things like damage taken is that their value is very dependent on context and isn't easily measurable, so the only thing you can really look at is whether the match was won or lost.

#222 That Token Canadian Guy

    Member

  • PipPipPip
  • Ace Of Spades
  • Ace Of Spades
  • 83 posts
  • LocationSomewhere Ville

Posted 09 June 2020 - 04:28 PM

I have to say , Russ is wrong about it working fine at current , if there are that many people playing at one time , I would not see the same people over and over again . The tiers have been slammed together for a long time now . Now the issue is tonnage balance , granted team work makes it easier to kill stuff...BUT when one team gets 6 assaults because the other team gets a team of 2, the team of 2 are lights .... only 1 assault on that team? Sorry to say , doesn't make sense to me . From what I am reading however , seems like they really didnt have a plan of attack , they must of thought it would be simple . So in order to shut us up , he wants us to do the work.

Maybe Russ should start playing his own game for once . We all know he gave up on it after a series bad decisions and pandering to people who didnt even buy anything and chased off the ones who did. The reason we have a low pop is because the 10% who complained on the forums got what they want. The other 90% we too busy playing this game and enjoying it . NOW , the 90% pretty much got chased off and the 10% still didnt buy a thing because they were just plain bad at this game .

I seen Kras was willing to take up the mantel of the PSR . Let him do it , he plays and has a base understanding what ACTUALLY has to be done . Start pandering to the actual players that play . They will tell you only a few things need to be changed to create balance. Nerfing weapons because bad players can not use them correctly is just sad .

I personally would wait for a good match of people my skill level . I had no problem in the past waiting 20min. Just knowing the guys are your team at least have a clue is a big step . Its no fun getting slammed every game because rookie players do instant nascar . Assaults get ditched , and its 50/50 who gets the edge up . Rinse , repeat , extremely boring . Tactics YES PLEASE!.
Thats my rant , hate me if you want , I am not the only one who feels this way.

#223 Andrzej Lechrenski

    Member

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

Posted 09 June 2020 - 04:32 PM

Paul,

Is there any way to get more information on how match maker actually works? One particular question I have is if the Tiers need to be directly used by it. For instance, in the past you have talked about "valves" and seemed to have fine-grained control over the "tier separation" of teams, using numbers that appear to be absolute values of PSR.

c.f. https://mwomercs.com...overies-week-1/

Quote

The current values are giving us the following:

Average wait time in QP Queue: ~1 min 30 seconds (up from ~48 seconds)
Average Tier separation between teams: ~300 (tightened from ~1800)
Average Weight Class difference: +/- 1 or 2 in a given weight class (5 heavies, 3 assaults for example) (tightened from 12)


Furthermore, are there upper and lower limits to the values of PSR that the servers record for each player (i.e. does a player maxed out in Tier 1 stop gaining PSR)? What are the PSR cut offs between Tiers? Are they all equally spaced? Must they be? Or, if you really do have fine-tuning control, what is the granularity of that? Can it be a continuum?

I'm sure I can think of many more questions, so a basic rundown of what sort of control and record-keeping you guys have on MM on your end would be greatly appreciated.

Edited by Andrzej Lechrenski, 09 June 2020 - 04:34 PM.


#224 StrikerX22

    Member

  • PipPip
  • The Warden
  • 26 posts

Posted 09 June 2020 - 04:39 PM

I'd like to kick off with just spitting out the system I have in mind.

My_PSR_Gain = My_MS/My_PSR + BonusOrPenaltyForWinningLosing - Average(P1_MS/P1_PSR, P2_MS/P2_PSR, ... P24_MS/P24_PSR).

Example:
(Numbers are inflated from likely actual numbers for clarity. You can imagine multiplying by the PSR average across teams here, to see the MS/PSR part as essentially just a form of MS.)

Quote

Winners [Total from this team's MS/PSR is 2890. Grand total is 4930. Grand Average is approx. 205.4.]

P1_MS/P1_PSR = 500; add 50 for winning, subtract 205.4 (grand ave) = 344.6 is PSR gain (multiply by some constant_A as wished).
P2_MS/P2_PSR = 450; add 50 for winning, subtract 205.4 = 294.6
P3_MS/P3_PSR = 350; + 50 - 205.4 = 194.6
P4_MS/P4_PSR = 350; + 50 - 205.4 = 194.6
P5_MS/P5_PSR = 300; + 50 - 205.4 = 144.6
P6_MS/P6_PSR = 250; + 50 - 205.4 = 94.6
P7_MS/P7_PSR = 200; + 50 - 205.4 = 44.6
P8_MS/P8_PSR = 190; + 50 - 205.4 = 34.6
P9_MS/P9_PSR = 150; + 50 - 205.4 = -5.4
P10_MS/P10_PSR = 100; + 50 - 205.4 = -54.4
P11_MS/P11_PSR = 50; + 50 - 205.4 = -105.4
P12_MS/P12_PSR = 0; + 50 - 205.4 = -155.4

Losers [Total from this team's MS/PSR is 2040. Grand total is 4930. Grand Average is approx. 205.4.]

P1_MS/P1_PSR = 1000; Subtract 50 for losing, subtract 205.4 = 744.6 is PSR gain (multiply by some constant_A as wished).
P2_MS/P2_PSR = 300; Subtract 50 for losing, subtract 205.4 = 44.6
P3_MS/P3_PSR = 200; - 50 - 205.4 = -55.4
P4_MS/P4_PSR = 150; - 50 - 205.4 = -105.4
P5_MS/P5_PSR = 120; - 50 - 205.4 = -135.4
P6_MS/P6_PSR = 90; - 50 - 205.4 = -165.4
P7_MS/P7_PSR = 70; - 50 - 205.4 = -185.4
P8_MS/P8_PSR = 50; - 50 - 205.4 = -205.4
P9_MS/P9_PSR = 30; - 50 - 205.4 = -225.4
P10_MS/P10_PSR = 20; - 50 - 205.4 = -235.4
P11_MS/P11_PSR = 10; - 50 - 205.4 = -245.4
P12_MS/P12_PSR = 0; - 50 - 205.4 = -255.4


For this to work well, PSR should try to reflect how many multiples of effectiveness over another player one is. For a rough estimate, I might guess from match score in faction play that the best player there is 10-20x better than the worst player. This ratio can be reflected in their PSR differences, so that it represents a more "real" quantity, and thus works well in the above formula. To facilitate this, one example could be keeping 1 as lowest possible PSR, and aiming very high skill being around 20, using whatever multiplier to both those for clarity as necessary. I would avoid 0 for both logical and div0 error reasons.

One thing to note is the average would be the cutoff for a positive PSR change, but winners get an edge in the positive direction, losers the opposite.

50 (arbitrary) is the bonus for winning and -50 for losing, when Team's total PSR is even with the other team's. Otherwise, the number tilts more positive to the lower PSR side. For example, say Team A has double the PSR Team B has, so 2/1 (A/B) ratio. Square that, for instance, 4/1 (A/B). That would make Team A's win bonus 50*(1/4)=12.5, loss -50*(4/1)=-200, with Team B's win bonus 50*(4/1)=200, loss -50*(1/4)=-12.5. This rewards you for team upsets. (This is necessary for instance in the scenario that you are say, tier 1 but the rest of your team is mostly tier 2 somehow, while the enemy is mostly tier 1. If not for the bonus from your Team PSR being lower, you would be graded the same as the enemy team, even though you're carrying.) With this, if winning/losing affect MS already, that should probably be removed in favor of this.

Other thoughts not directly related to this will be in my next post.

#225 StrikerX22

    Member

  • PipPip
  • The Warden
  • 26 posts

Posted 09 June 2020 - 05:06 PM

Other thoughts: (Please note the system I proposed above, if you don't mind, press like if you agree, etc. for consensus.)

Should MS changes tweak the meta?:
The goal is to win the match, so if killing the other team allows you to theoretically meet the goals of the game mode "off screen", then so be it. We should not be encouraged to do various things that /seem/ maybe fitting for the game mode unless it actually will win us the game. If it changes the meta, that's only because people are playing for MS rather than winning, which is improper. This is another argument for a strong weighting of W/L for PSR growth. That said, if it's going to be used anyway, you want it to be properly in line with the actual value of the action.

Mineomatic said:

I like running Light Mechs sometimes but with the current PSR system I feel I put my team at a disadvantage each time I do and having different tier brackets for each class of mech would eliminate that some.

Having different tier brackets for each weight class sounds like a great, fun idea. It would mean having 3 more variables to track for each player in the database, though, if that's a problem. It would be a simple and nice solution. I'm not experienced enough to say whether each weight class is actually "balanced" when played properly.

Rolling "Average" (not sure of the exact implementation in mind):
This kind of system is probably a pretty important aspect, though I'm not sure how much data the database records currently. It's possible this would take a lot more space than they'd like to support at this stage (you would track every match's change in PSR so you can drop off the last value when you get a new one, and do something with those, right?).

Team Sizes:
I'd like to see different sizes of matches randomly thrown in, not just to help MM, but to be interesting. Maybe just play with the number of lances each team has. This could potentially even be used for making "imbalanced" matches of fewer higher tier players on one team vs more lower tier ones. I wouldn't mind seeing lance sizes fluctuating between 3 and 4, although if my very old lore memory serves, 4 was a typical group number. I honestly think this extra varying factor would be lots of fun (though 4v4 might be inconsistent in non-skirmish modes, necessitating good weight class matching, but 6v6 might be okay as is).

"I don't get justice every match. I played well and lost but didn't get rewarded.":
Wins and Losses are the ultimate identifier of success, the (proper) reason we play the match, and covers all blind spots MS misses or miscalculates. They may not give justice immediately, but they do *eventually*. They are the true measure of skill in the end. Yes, having a smaller player base probably hinders it, but it should not stop it. Better players will win more eventually, period. It just doesn't feel great on the way. Of course, the community wants it to be competitive AND feel good, so we can find a happy medium.

--Groups: That said, Having personal skill rated does help take care of being carried in group drops. I also think it would be fair to consider groups to be a bit higher skill level than their average. One good way might be to add a flat value to the group based on the highest tier (I do not agree with considering the whole group that highest tier, as that's very binary).

"PSR means personal skill rating!":
Let's not get caught up in the naming conventions. What we want is "ability to help team win", which is why wins are a great way of (eventually) determining that.

"Do PSR changes first, see how it works, then do MS.":
Except that using a PSR system based on MS means that it will work less well without fixing the system as a whole. I believe that some small tweaks could be made without confusing things too much. It is worth noting as an issue, of course, but it works both ways.

"Damage contributes too much to Match Score.":
What should we actually be measuring: effectiveness on taking out a mech. Try this: Give a much heavier MS to contributing damage to components that are destroyed, more so if it contributed directly to killing the mech (both side torsos would count if clan XL mech dies to losing both side torsos. Caveat for back vs front armor: only count for this bonus if that side of armor was completely pierced (hence, it actually contributed). Reward general damage only a little.

--Tanking: This can also address the call for tanking damage giving MS: If tanking the damage does not result in that component being destroyed, then you "successfully" tanked the damage (exception being you get bonus still by tanking front/back damage when if armor doesn't get pierced, even if component is destroyed from other end). Losing a component not contributing directly to your death could give a smaller bonus. You can give a smaller reward for tanking any damage, as it can still be valuable, but not as much.

"This idea doesn't need to affect MS because it already gives you a chance to get MS other ways":
The same could be said about a lot of MS rewards. You still have to capitalize on the chances. Also, it's hard to keep it "simple" at this point since MS is anything but. Rather, it's trying to be "robust".

"AMS contributes too much to MS.":
This really depends on how much AMS contributes to a win. It has nothing to do with skill. The skill issue comes into play in how skillful do you have to be to contribute x MS worth of help for the win? That part is a balance issue. If it is too effective for skill involved, it needs to be nerfed (just remember you give something up by having it in your mech, though maybe not enough). If it doesn't help as much as MS says it does, then nerf the MS, yes. (I won't assume as I'm new, but it's hard to be sure of what might have been.)

Perhaps later I will have a close look at all MS factors and try to come up with numbers that seems fairly rigorously logical, but I lack experience so we'll see.

#226 BrrtWarthog

    Member

  • Pip
  • Bad Company
  • Bad Company
  • 12 posts

Posted 09 June 2020 - 05:28 PM

View PostPaul Inouye, on 08 June 2020 - 02:52 PM, said:

win - Base value for winning.
loss - Base value for losing.
teamkills - Loss for team killing.
spottingassist - Gain for spotting an enemy and that enemy takes damage. (Press R)
componentdestroyed - Gain for destroying a component on an enemy.
death - Loss for dying.
capturewin - Bonus gain for winning by capture.
captureassist - Gain for helping others capture a capture point by being in the capture radius.
suicide - Loss for suiciding.
saviorkill - Gain for helping a teammate under fire and you get the kill shot on his highest damaging opponent.
defensivekill - Gain for helping a teammate under fire and you get the kill shot on an opponent damaging them.
uavkill - Gain for any kills happening under a UAV you deployed.
uavlockeddmg - Gain for any damage done to a target that is under your UAV.
uavdetection - Gain for any new enemy detected by your UAV.
counterECM - Gain for countering enemy ECM.
counterECMLockedDmg - Gain for damage done to enemies under your counter ECM.
turretkill - Gain for killing a turret.
killblow - Gain for getting the killing blow on your enemy.
killassist - Gain for damage done to enemies upon kill but you didn't explicitly get the kill shot.
teamdmg - Loss for team damage done.
damagedone - Gain for damage done to enemies.
killmostdmg - Gain for getting the kill shot and you did the most damage to your enemy.
solokill - Gain for killing an enemy without the assistance of your teammates.
scouting - Gain for targeting enemies without damage being done.
flanking - Gain for being out of LoS to your enemy and behind enemy line.
capture - Gain for capturing a capture zone.
capturepulse - Gain for time you are capturing in a capture zone.
firstcapture - Gain for capturing the first capture zone in a match.
brawling - Gain for being in combat agaist multiple opponents.
tagdmg - Gain for any damage done to an enemy you have tagged.
tagkill - Gain for any kill done to an enemy you have tagged.
narckill - Gain for any kill done to an enemy you have narced.
hitandrun - Gain for attacking an enemy and escaping their LoS for an amount of time.
tagstealth - Gain for tagging an enemy behind enemies and not being targeted by them.
lanceformation - Gain for time spent near lancemates.
protectmedium - Gain for killing an enemy who is attacking a medium class teammate.
protectlight - Gain for killing an enemy who is attacking a light class teammate.
protectproximity - Gain for killing an enemy who is near any teammate.
powercell_pickup - Gain for picking up a power cell.
powercell_dropoff - Gain for dropping off a power cell.
incrusion_destruction - Gain for destroying objective objects in Incursion.
kill_powercell_carrier - Gain for killing a power cell carrier.
ams_missile_destroyed - Gain for missiles destroyed by your AMS system.
 
matchscore_scale - A multiplier used to reduce the sum of all match score activity to keep numbers reasonable.



View PostPaul Inouye, on 08 June 2020 - 02:52 PM, said:

Each of these triggers has a value assigned to it. The values are essentially the weighting of the variable in the overall sum of all actions listed. At the end, the number is multiplied by a decimal multiplier that reduces the overall sum to a number that is manageable.


Quote

If the community can agree on weighting of these events, we will review them on our side to make sure they're fair to as many play styles as possible. Once that is done, we'll implement the numbers and try them out.


That list doesn't mean much unless we also know how each data point is weighted in the current calculation. Can you include that data as well? Unless we know the current weight of any given data point, any suggestions made so far in this thread are purely guesswork (however well educated/researched those guesses may be). If you're asking the community to come up with a new calculation, we need to know exactly how the calculation works currently, so we know which suggestion(s) being proposed would have the most positive or negative impact(s).

Without knowing those weighting values, here's my 2 cents: Currently, winning a match has a greater positive impact on match score than losing has a negative impact (hence the upward movement bias). All other things being equal (scores attained for all the data points other than winning/losing), winning and losing should have an equal and opposite impact on match score, in some fashion. Remove the upward bias that currently exists through simply being on the winning team.

#227 cougurt

    Member

  • PipPipPipPipPipPipPip
  • CS 2023 Silver Champ
  • CS 2023 Silver Champ
  • 691 posts

Posted 09 June 2020 - 05:58 PM

View PostStrikerX22, on 09 June 2020 - 05:06 PM, said:

"Damage contributes too much to Match Score.":
What should we actually be measuring: effectiveness on taking out a mech. Try this: Give a much heavier MS to contributing damage to components that are destroyed, more so if it contributed directly to killing the mech (both side torsos would count if clan XL mech dies to losing both side torsos. Caveat for back vs front armor: only count for this bonus if that side of armor was completely pierced (hence, it actually contributed). Reward general damage only a little.

--Tanking: This can also address the call for tanking damage giving MS: If tanking the damage does not result in that component being destroyed, then you "successfully" tanked the damage (exception being you get bonus still by tanking front/back damage when if armor doesn't get pierced, even if component is destroyed from other end). Losing a component not contributing directly to your death could give a smaller bonus. You can give a smaller reward for tanking any damage, as it can still be valuable, but not as much.

i like these ideas, the question is whether PGI is capable or willing to implement them. i would also probably include some additional requirements to the tanking criteria, since it's not very useful if your team isn't able to capitalize on it. something like being in close proximity to teammates when you take damage, or the attacker taking damage within a certain period of time.

#228 T e c h 4 9

    Member

  • PipPipPip
  • CS 2022 Participant
  • CS 2022 Participant
  • 77 posts
  • LocationBehind you

Posted 09 June 2020 - 06:45 PM

View PostnRom, on 08 June 2020 - 08:43 PM, said:

People arguing for a zero-sum system are saying matchmaking is poor due to poor sorting of player skill by the tier system (ie the reason games are bad are because the good players are stuck in T3). I think odds are it's because the player base is too small and the skill range too diverse to consistently make matches in a decent amount of time. I say this as someone who keeps MWO on the shortlist of games to play over the past 7 years when I have some time not mowing grass.


This may be off topic, but this an argument for going to 8v8 I would think.

The PSR needs to be reworked and reset. I'm not sure how to best do that, but there are others in the community that I'm sure could help PGI figure that out. Hopefully, PGI will listen. So far, this communication from them is encouraging.

#229 Nightbird

    Member

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

Posted 09 June 2020 - 07:38 PM

My PSR proposal:

https://mwomercs.com...mmunity-psr-mm/

#230 Brauer

    Member

  • PipPipPipPipPipPipPipPip
  • 1,066 posts

Posted 09 June 2020 - 08:16 PM

View PostJPhoenix12, on 09 June 2020 - 04:15 PM, said:

MM seems fine, just find some way to break the never ending nascaring and MWO will be much improved.


Bring a group.
Take a power position.
Gun everything down that exposes.

#231 denofsteves

    Member

  • Pip
  • Mercenary Rank 6
  • Mercenary Rank 6
  • 19 posts

Posted 09 June 2020 - 09:15 PM

I believe that PSR is the more important item to address at this time. Match score influences PSR, but MS is a different system from PSR, and is not directly used by the MM.

Regarding previously mentioned PSR solutions, I prefer Andrzej's Solution 2.

View PostAndrzej Lechrenski, on 08 June 2020 - 07:38 PM, said:

Solution 2

Use match score only in a relative (within the same match) sense and give priority to W/L ratio. Winners gain PSR, losers lose PSR. Here is an idea of what that might look like using Russ' PSR numbers:

Winning team:
player 1 +24
player 2 +22
player 3 +20
player 4 +18
player 5 +16
player 6 +14
player 7 +12
player 8 +10
player 9 +8
player 10 +6
player 11 +4
player 12 +2

Losing Team
player 13 -2
player 14 -4
player 15 -6
player 16 -8
player 17 -10
player 18 -12
player 19 -14
player 20 -16
player 21 -18
player 22 -20
player 23 -22
player 24 -24


Winning is the objective in this game. In team sports, the fun is in the playing with like-minded enthusiasts, where the camaraderie and shared experiences are why players keep coming back. Winning is the icing on top. It feels good to win. It feels bad to lose. It feels less bad to lose if it was a hard-fought game and you know you did the best you could; that your teammates know you did the best you could. This is a team game, so rewarding the team that wins is the better way.

View PostAndrzej Lechrenski, on 08 June 2020 - 07:38 PM, said:

With the player pool as small as it is, there is clearly never going to be a 'perfect' solution,...


The player counts we have been shown are not high enough for a MM that is both timely and exclusive. At peak times, we approach a minimum-viable number of players for the MM to work best. The rest of the time, well, we keep playing "for the love of the game".

#232 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,209 posts

Posted 09 June 2020 - 09:22 PM

View PostGeneral Vulcan, on 09 June 2020 - 04:28 PM, said:

I have to say , Russ is wrong about it working fine at current , if there are that many people playing at one time , I would not see the same people over and over again . The tiers have been slammed together for a long time now . Now the issue is tonnage balance , granted team work makes it easier to kill stuff...BUT when one team gets 6 assaults because the other team gets a team of 2, the team of 2 are lights .... only 1 assault on that team? Sorry to say , doesn't make sense to me . From what I am reading however , seems like they really didnt have a plan of attack , they must of thought it would be simple . So in order to shut us up , he wants us to do the work.

Maybe Russ should start playing his own game for once . We all know he gave up on it after a series bad decisions and pandering to people who didnt even buy anything and chased off the ones who did. The reason we have a low pop is because the 10% who complained on the forums got what they want. The other 90% we too busy playing this game and enjoying it . NOW , the 90% pretty much got chased off and the 10% still didnt buy a thing because they were just plain bad at this game .

I seen Kras was willing to take up the mantel of the PSR . Let him do it , he plays and has a base understanding what ACTUALLY has to be done . Start pandering to the actual players that play . They will tell you only a few things need to be changed to create balance. Nerfing weapons because bad players can not use them correctly is just sad .

I personally would wait for a good match of people my skill level . I had no problem in the past waiting 20min. Just knowing the guys are your team at least have a clue is a big step . Its no fun getting slammed every game because rookie players do instant nascar . Assaults get ditched , and its 50/50 who gets the edge up . Rinse , repeat , extremely boring . Tactics YES PLEASE!.
Thats my rant , hate me if you want , I am not the only one who feels this way.

As I've already said, skill >> tonnage. For example I'm bad Assault pilot. When I play Assault, I can perform worse, than when I play Heavy. And some Light pilots, I met in my matches, could tear enemy team apart. Size doesn't matter, you know. Assaults are just big slow target dummies. It's harder to kill Light, that spreads damage across several 20 armor parts and dodges 90% of damage, than 2-shot Assault's ST, that requires 70DMG to destroy. Because even some Mediums (Clan ones mostly) have such amount of firepower today. And it's even easier to kill IS one due to ST-death penalty. Once your armor is stripped, even small scratch on ST one-shots you. Personally, I can't play Assault as tank. Only as support. So there is nothing bad in having 6 noob Assaults vs 1 pro in one match.

Edited by MrMadguy, 09 June 2020 - 09:27 PM.


#233 K4I 4LL4RD LI4O

    Member

  • PipPipPip
  • The Warden
  • The Warden
  • 78 posts

Posted 09 June 2020 - 09:57 PM

Ok so no changes to PSR so far I get that. If you really want to help with stomps and problems with balancing at the moment, then there is something pretty simple you can do.
On Mining Collective, HPG and Canyon Network change the Spawn points. Right now if you are in a slow mech and get dropped in Alpha lance you are dead, no matter how good you are, lights will swarm you and eat you alive. Change that and you will have better results.
Thx

#234 SkubaSteve

    Member

  • Pip
  • Shredder
  • Shredder
  • 15 posts

Posted 09 June 2020 - 09:58 PM

Thank you this is much closer to the fair system weve wanted for years. Each player should be judged individually, not based on a win or loss due to the rate of pugs bird watching and this looks like it might do that. Wonderful

#235 Saint OZ

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,701 posts

Posted 09 June 2020 - 10:00 PM

View PostHorseman, on 09 June 2020 - 02:20 PM, said:

Which lead to matchscore anyway. Accuracy translates to damage and kills. Ability to spread damage translates to more scoring opportunities.

Agree in accuracy, but amount of taken damage should be particular counted.

#236 Nightbird

    Member

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

Posted 09 June 2020 - 10:12 PM

was asked to move my post here:

Suppose all the Match Score PSR suggestions to date have major flaws, and also suppose that there is a WLR-based MM solution that will work (at least no one is managing to present flaws), which would you prefer?

A Match Score system that is no better than what we have today, or a WLR system that will work but will knock you down if your team is unluckily bad. (Suppose that a MS solution will improve things from 15% good matches today to 20%, and a WLR solution will improve it to 50%) Which would you pick?



Details (may cause headaches!)

Match score solutions so far have the problem that
1) the PSR will go off to +- infinity for good bad players
2) the rate people go off is determined by number of games played, not skill
3) if we cap the PSR movements, people will stop at the cap like Tier 1 today
4) is as time progresses, moving between tiers is harder and harder

WLR solution:
https://mwomercs.com...thread-we-need/
Based on this thread, just set PSR=WLR (vary how many past matches WLR is based on 20-100) and you'll get much better matches. This change will be transparent to the users, keep the PSR EXP bar on the frontend. Only the MM in the backend will need to know it to make better matches. This is to prevent an unlucky loss from bothering the player.

A related thread to explain why MS is not a good predictor when used the same way:
https://mwomercs.com...and-suggestion/
Finally using WLR is not the best I can come up with, adding other factors such as mech tonnage, weapons range, map, and game mode is definitely possible, but PSR=WLR is the easiest

Edited by Nightbird, 09 June 2020 - 10:17 PM.


#237 Surn

    Member

  • PipPipPipPipPipPipPipPip
  • Hero of Kurita
  • Hero of Kurita
  • 1,074 posts
  • Twitter: Link
  • LocationSan Diego

Posted 09 June 2020 - 10:53 PM

Match Score proposed values : https://docs.google....dit?usp=sharing
Posted Image
Includes 3 data points not on the list, but easily available.. one already reported.

Examples are a decent match and a terrible mistake match involving a teammate running in front of me just as I launched an Artillery Strike on the center ramp of mining collective!
First Match Score was the actual match score from that game:


Using the current game match scores and player overall PSR.

PSR Calculation
If lose:
PSR=( (player match score * (MedianPSR of players in game/PlayerPSR))/ team median match score ) - ( 1 + (winning team median Match score/ losing team median match score))

Example, good player gets 300 match score in loss
(300 * (.75)/ 200 ) - (1 + 275/200) = 1.125 - 2.375 = -1.125

If win:
PSR=( (player match score * (MedianPSR of players in game/PlayerPSR))/ team median match score ) + ( 1 + (winning team median Match score/ losing team median match score))

Example, good player gets 300 match score in win
(300 * (.75)/ 200 ) + (1 + 275/200) = 1.125 + 2.375 = + 3.5

The formula could be simplified by making win =1 and loss =-1

Simplified Formula:
If not win then win = -1
PSR=( (player match score * (MedianPSR of players in game/PlayerPSR))/ team median match score ) + (win)* ( 1 + (winning team median Match score/ losing team median match score))

Edited by Surn, 09 June 2020 - 10:59 PM.


#238 Sniper09121986

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Sickle
  • The Sickle
  • 2,161 posts

Posted 09 June 2020 - 11:08 PM

View PostAnomalocaris, on 08 June 2020 - 06:46 PM, said:

current population

View PostKnight Captain Morgan, on 08 June 2020 - 07:01 PM, said:

TLDR: the population has dropped enough

View PostVellron2005, on 09 June 2020 - 01:05 AM, said:

there simply isn't enough players still playing this game

View PostBud Crue, on 09 June 2020 - 02:43 AM, said:

What more do we need given the population?

View PostEatit, on 09 June 2020 - 11:29 AM, said:

All matches will still be unbalanced due to Low population MM pulling from all tiers.


I am sure there have been other people, but this is all I bothered to quote. As I said in a different thread:

Quote

People leave because a game is bad in some areas, and if the game is good then people have no reason to leave. The low population is the effect, not the cause. Not that I object against a marketing campaign, it just needs to come after the game-breaking issues have been addressed. There might be skill gaps with low population, but they will get filled once people see a game they can actually enjoy.


The matchmaker works with what it has, but its core principles of operation should not depend on the game population. What is good for the goose is good for the gander. If the matchmaker works as advertised people would return and the gaps in PSR will get filled.

#239 MrMadguy

    Member

  • PipPipPipPipPipPipPipPipPip
  • 2,209 posts

Posted 09 June 2020 - 11:11 PM

Yeah, it would be really great, if MS would be real measure of performance in match. But I don't care, how it will be implemented. Only thing, that I need now: if I have bad match - my rating should drop. And it should happen as soon, as possible. That's it.

Yeah, true zero-sum would be great. As current system has just one flaw. It has fixed MS target. And it's way to easy to achieve, even while playing badly. This means, PSR will be biased towards increasing anyway.

#240 Horseman

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Mercenary
  • The Mercenary
  • 4,736 posts
  • LocationPoland

Posted 10 June 2020 - 12:00 AM

View Postcougurt, on 09 June 2020 - 04:17 PM, said:

yeah, no need to overcomplicate things. the only issue i see with the way match score is currently calculated is that it's a bit indiscriminate in how it rewards damage. i don't think it's a big deal since generally speaking if you're doing a lot of damage it's likely to be a fairly hefty contribution regardless, but obviously there's a little more nuance to a player's performance than sheer damage output.
It probably should emphasize components, KMDDs and especially solo kills a bit more and raw damage somewhat less.





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users