Psr Update And Changes - Jun 2020
#121
Posted 04 June 2020 - 04:09 PM
I still debate whether the game result of the win/loss for the team should actually be a factor.
If my team wins, great, load me up with c-bills and salvage.
If we lose, well, the spoils go to the victor.
But as an individual player in that team if I am regularly performing under the bar, which is about that 150 - 250 match score, then I should be dropping in PSR until my match score begins to level.
We have various teamwork and activity bonuses that impact match score, so as a player if I am getting those then I am at least trying to function as part of a team and in doing so, will be improving my match score.
Add to that doing some serious work in fighting the opposition and I should be pushing into the +/- 0 PSR change if not a + increase of PSR.
If we want to encourage other roles on the battlefield then adjusting some of those bonuses should be looked at to help bring up the match score.
#122
Posted 04 June 2020 - 04:15 PM
you are lucky enough to be on winning team so you can go up in points?
#123
Posted 04 June 2020 - 04:23 PM
denAirwalkerrr, on 04 June 2020 - 10:06 AM, said:
In theory, ~15% stomps where you are on the losing team. Of course, us humans are naturally biased to notice the losing stomps more that the winning matches overall. XD
#124
Posted 04 June 2020 - 04:26 PM
Now there is one thing I'd like clarified. Will the match maker look only at tier when matchmaking or will it also match according to different PSR within the tier? Reason I'm asking is that even if people now stay in their correct tier there is still a huge difference between someone like me (95th percentile) and someone like Vorteex (99th percentile).
I don't want the matchmaker thinking that we are equals, if there is two 99th percentile players in the pool the MM should understand that they should be on opposite sides of the team. Basically there needs to be granularity and PSR needs to be open ended so the top 5-10% don't stack up against a maximum value where they can no longer be told apart.
And guys, stop obsessing over the assymetry between losing and winning. PSR has absolutely nothing to do wirth being "correctly" "rewarded" or "punished" in individual matches. It is about trending over many matches towards your correct PSR so that the MM can predict your impact. A player who is too low in PSR will win more than he loses, and go up. A player who is too high will lose more than he wins and go down. Your W/L is actually about your own impact over time, not matter what you happen to believe or how obsessed you are about losing individual matche despite doing well etc. That is why all the good players on the Jarls list have positive WLR ratios, it's because they have an average positve impact on winning.
Now there are multiple ways you could make the formula, but as long as it's zero sum and works the same for everyone you will approach the correct value over time. Good players will trend upwards and bad players will trend downwards until they are where they should be.
Increasing in PSR is not a "reward", and decreasing is not a "punishment". Get that notion out of your head because it's a completely idiotic way to think about it. There is nothing good about having a high PSR and nothing bad about having a low one, you just want it to correspond as closely as possible to your average impact on winning, that's what it's for and that's what will give you the best matches. It's not a rank, it doesn't measure "true skill", it's simply a matchmaking tool.
As far as tier separation, I honestly don't care about it. Mixing tiers is fine as long as the mix is the same on both teams. In other words as long as someone on the other team corresponds roughly to my skill level and so on for each teammate, I'm perfectly fine with playing with all tiers.
There is also some reason to believe that a mix of skill levels on both teams reduce the snowball effect, the reason being that the weakest players on both sides tend to die first and then there is a decent chance for the remaining better players to turn the match around. If the skill level is even the first kills matter much more and the snowball effect is stronger.
Anyways let's hope it works out. I'd also be interested to hear more about how groups are matched in this system.
#125
Posted 04 June 2020 - 04:33 PM
Calculating tiers on a rolling average score of the last 20 or 50 matches removes the deep, dark trough that a bad run of luck or unco-operative teams could otherwise cause.
#126
Posted 04 June 2020 - 04:37 PM
#127
Posted 04 June 2020 - 04:39 PM
#128
Posted 04 June 2020 - 04:54 PM
Magic Pain Glove, on 04 June 2020 - 11:17 AM, said:
So I feel like this is where we're all going wrong: PSR isn't about "punishment" or "reward", and everyone keeps treating it like it is. PSR is ONLY about "are you better than everyone else you played with and need a bigger challenge, or are you worse than everyone you played with and need less of a challenge?". To that end, I feel as though winning brings its own natural bias in match score: if you're winning, you're not getting killed as effectively as your enemies, and you have more time to do damage, spot enemies, etc, resulting in a higher match score at the end of the match. If you're losing, then your team is dying faster (or you died first) and you have less time for everything that generates match score.
Thus, the PSR advancement should be independent of win/loss, because the variation in match score between those will take care of that on its own. PGI should decide on a "target" match score they feel is representative of a fair game. If you get above that, you go up, and if you get below that, you go down. It's gonna be hell on their 500 match score challenges, though, because "fair" seems to be around 300 match score...
#129
Posted 04 June 2020 - 05:04 PM
If this is really the case.
#130
Posted 04 June 2020 - 05:11 PM
MWO Playerbase on June 9th when PSR changes release exactly as written in the first post despite days of "debate" which essentially boil down to keeping the individual based matchscore system we already have.
#131
Posted 04 June 2020 - 05:25 PM
If you haven't played in a while, your score moves back towards zero, maybe as much as 5% per day you haven't played. This gives a decreasing movement towards the middle.
Good and bad players will quickly move to the correct level. The rest will be somewhere in the middle, which is the goal. Adjustments can be made by changing up match score valuation.
Remove tier display to get over the ego issue. Better yet, show everyone as tier 1 to reinforce how it doesn't matter.
Match maker can use simple number ranges for grouping. Starting at zero and moving up and down reduces the chance that a top level player that hasn't played in a year will be clubbing seals because the lowest that player will fall is the middle of the pack. Same for poor players, they won't be fresh meat for long.
There will be lots of volatility in the middle, so if you get on a hot streak, matches will get tougher, have a bad day and the matches get easier.
Edited by denofsteves, 04 June 2020 - 05:28 PM.
#132
Posted 04 June 2020 - 05:56 PM
I think you need to take W/L out of it and base it on MS only. Those who do well go up, those who don't go down, that makes it fair for everyone.
Those who are consistently good will rise and those who are consistently bad will not.
#133
Posted 04 June 2020 - 06:08 PM
C337Skymaster, on 04 June 2020 - 04:54 PM, said:
So I feel like this is where we're all going wrong: PSR isn't about "punishment" or "reward"
Don't get caught up in the semantics. People are just using reward and punishment to describe the net effect on PSR. Scoring multiple kills and lots of damage while the rest of your team does **** all in a stomp loss isn't an indicator of poor skill. Thus, your PSR should probably still go up if you're top of the losing team by a big margin. That's a "reward" if you like (although some people might feel like rising in rank is a punishment when they face more top players.....) or call it something else. It's not like you're getting extra payout, xp or special items for coming in top of the losing team. But your skill rating shouldn't be impacted negatively when you showed more skill than your entire team and half the enemy, right?
Ballistic Panicmode, on 04 June 2020 - 04:23 PM, said:
In theory, ~15% stomps where you are on the losing team. Of course, us humans are naturally biased to notice the losing stomps more that the winning matches overall. XD
Stomps are bad no matter which side you're on. Winning 12-1 against drunkards and whirly gigs flitting about the map isn't fun and doesn't generate high matchscores (if I can instagig someone from behind with 60 pts of damage because they're oblivious to everything but the target they're tracking, I'm getting a much lower matchscore than if I take out a twisting, moving, jumping player who splashes damage everywhere - and the latter is much more satisfying).
#134
Posted 04 June 2020 - 06:16 PM
Almost no fighting. 5 players on the winning team under 100 match score.
Also, this system is hardly zero-sum. A very good, 12-11 match (not that they happen that often) can easily have most players either staying put or going up, even on the losing team. To be zero-sum, you'd need to have a scale from -11 to +11. Best player gets a +11. Worst gets a -11. the two most mediocre players both get a 0. THAT would be zero-sum.
I'm a fan of the rolling averages as well, and upping the effects things that aren't necessarily damage/kills have on match score. I absolutely love scouting. And I don't just mean taking a fast mech and playing Squirrel!, or trying to backstab people. I mean scouting, finding the enemy, relaying where they are, getting/holding locks etc. But in terms of match score, that's almost never worth anything, even it's a key in winning.
I also often laugh at people who want to just do a lot of damage -- I've seen plenty of times where half of their damage was essentially wasted. Counter to that, I've been leveling mechs recently, and playing m,y AC20 Hunchie, I've also noticed that my match scores have been lower, even when getting multiple kills, because I'm not sandblasting my opponent's armor. I'm hitting a few time in one location, and getting kills on relatively undamaged mechs. Damage is SUCH a big factor in match score. And there is useful damage and useless damage. As someone else pointed out earlier in this thread, LRM boats (I know from experience, having played one or two) can easily rack up match score without getting any kills. Just sandblast armor all over the place.
#135
Posted 04 June 2020 - 06:24 PM
If you get 100 or less (maybe make it 150) and win you aren't rewarded.
Honestly i'm fine with that. If you win half your games and play average you will be average.
You win half your games and play well, you will go up
You win half your games and are a miserable player you go down
It's good enough.
#136
Posted 04 June 2020 - 06:34 PM
1. I don't believe this is a zero sum system. In order to have that, I believe you would have to have the total number of PSR change be zero sum in each match. I'm not an expert, but this stuff can very easily go in unexpected directions. I believe the proposed new system might actually result in people falling in rank on average once they reach to around their correct tier, as you tend to have lower match score on games you lose, so you will tend to inflate that negative number more than the positive one when winning, assuming you have half wins, half losses, at that correct tier. Maybe that would all work itself out still, though? I don't know that we need a strict zero sum system. And yeah, as denofsteves notes, that probably means people not playing move down. I don't think this is necessary, but correct me if I'm wrong. If they were to go down, I'd only start a week after not playing, and limit it to a certain % less from rating when last played, like 5 or 10% eventually total. I don't think skills leave that fast or much, generally, and this isn't a rank system for rewards or something.
2. I'm inexperienced, but I feel group play is more complicated than most are saying. Issues seem to be as follows:
-A- They tend to work better together. [ This is not inherently a bad thing, as the game is team-based, but might alienate solo players. ]
-B- They may tend to have their own discord and not interact with the team as a whole. [ This means they have worse teamwork, not better, so...? A potential negative, though. ]
-C- They can somewhat artificially inflate someone's rating. [ However, they are also affected by having a less skilled player, so they hurt their own rating as well. Furthermore, matchmaking should still take the skill difference into account, although with a smaller player base it does become more difficult and allows for wide gaps in skill in a single game. It can also be argued that a high skilled player has a disproportionately larger effect on the team than the negative effect of the low skilled player. This can still technically be accounted for, if enough players exist (which is a problem regardless but potentially a bit worse here). ]
3. The PSR system is clunky, in several ways:
-A- It uses arbitrary whole numbers for tiers of match score, rather than smoothly distributing.
-B- It uses 2 ratings systems at once: Win/Lose, and Match Score. Win/Lose are not arbitrary, while Match Score is absolutely arbitrary, and has many dials to adjust for balance, but is probably not actually balanced at the moment (some of this can come from certain strategies/weapons/etc. being overpowered / underpowered; otherwise, value needs tweaking).
-C- I don't remember anymore, I might've melded it with -B-.
...[ In a perfect world, match score would accurately reflect our impact, and even potential impact, on the game. With that perfect knowledge, we would actually set winning/losing aside, as they do not reflect your impact on your team, technically. You could technically have the worst teams every game and never win because of it. However, it is probably not a very realistic goal. It is complex and difficult to tune correctly. Wins and Losses are easy to read and will likely level things out over time, even though it feels bad in certain scenarios. This is not overemphasizing team play per se, but rather assuming you'll have normal luck if you play enough that it will even out properly, as it is a pretty reliable metric long term. I do feel like it would be nice to have match score weigh in a little bit, even if it is arbitrary, though... most everyone wants justice, so this would help quell people's anger a bit. However, it will never satisfy most of them when they do well but get an insufficiently good PSR change. Subconsciously, people expect perfect judgement on your part, or worse, on their part. ]
TL;DR with pseudocode:
In the end, I believe the new PSR system may be close to the correct direction. Perhaps match score should be taken out entirely from it. It should probably instead track the average PSR on your team, compared to the average PSR on the enemy team, over a long period of W/L (rather than individual games). Something like the following:
Quote
If Win, Then MyPSRGain = BasePSRFromWinning * (A + B *AveragePSROtherTeam/AveragePSRMyTeam) * 20/Minimum(MyGamesSinceNewSys, 20)
Else MyPSRGain = -BasePSRFromWinning * (A + B * AveragePSRMyTeam/AveragePSROtherTeam) * 20/Minimum(MyGamesSinceNewSys, 20)
Where A and B are constants for setting a base flat value (A) and a multiplier (B) to help when say, 2x the PSR doesn't mean 2x effectiveness (one way or the other). Note that for losing, Base PSR is negative winning value, and AveragePSRMyTeam switches position in fraction. If you are using whole numbers for some reason and don't want to switch to floating point, then please inflate the PSR ratings by a couple digits to have a smoother gradient of progression. Accuracy doesn't have to be perfect, either, if you're worried about floating point errors; we just want close.
Open valves for a while at start, to help things settle. Set initial PSR based on current tier to help settle things quickly (experience does tend to mean better) and to not offend players as easily (exact amounts probably don't have to be large or specific, just significantly different). With MyGamesSinceNewSys multiplier for first ~20 games, this should work out well I think. I believe this PSR system is around the best. If necessary, group play can be limited to within 2 tiers of each other, like a group of tier 3 to tier 1. Not everyone will like that but it will address the situation somewhat for solo players. Another option is to give a slight, positive flat boost to solo players each game; not because they deserve it, but because you want them to stay, so let them know.
This doesn't take the player's PSR directly into account as it doesn't make sense to here, like it would if using match score (though it is involved in the team's average). Poster Dogmeat1 was using median instead of average. Is there a particular reason for that?
I have a system in mind if match score is to be used, but I don't think that is the correct direction to take, as discussed above. Nevertheless, if it is definitely going to be used, then I will summarize:
- It gives winning team and losing team pools of PSR points, which is determined by the outcome of the match (bonus for win, then mostly by kills, and depending on game mode, some from details like Conquest point totals) and PSR ratio between teams. Aside from showing team skill, pyrrhic victories aren't great in-universe probably.
- Then those pools are divvied out based on match score and PSR within the team.
- Finally, you subtract (Total_PSR_given_to_all_Players / 24) from each player to make negative scores where they theoretically belong.
- (Note: Make sure to balance ways to get match score. [hard, many problems])
#137
Posted 04 June 2020 - 06:39 PM
VixNix, on 04 June 2020 - 02:56 PM, said:
turn the wrong corner and you are dead - low match score
get left behind by nascar and you are dead - low match score
no control of the team you get put on.
no one has any motivation to do anything except go to the same place on the map and then rotate...
no balance on the teams
no balancing of equipment on the teams
constantly get stuck on long range maps in short range mechs
etc etc
You never had control of teams, but despite that everyone goes up before, now not as much
Left behind by nascar, well you weren't going up anyway
No balance on teams and equipment, oh well its usually 50/50
Which brings me to the point. Most people sort of have that w/l of 1 going on.
If you get 250+ you net 2 points per pair of games
Less than 250 you lose 2 points per pair of games.
If you get lower than 250 you will eventually get 250 as you go down
If you get more than 250 you might eventually go down to 250 as players get tougher.
The really good players, sorry boys, nothing we can do about that, you're just going to go up because you can legitimately positively bias all outcomes.
[Redacted].
Hey its not perfect, for true zero sum the above commentators are probably better ideas, but meh.
Edited by GM Patience, 05 June 2020 - 02:05 PM.
#138
Posted 04 June 2020 - 06:40 PM
Review the allocation of match score for achieving mission and team outcomes, such as capping points, scouting, escorting the fatties, etc. Increase rewards for component destruction; if possible, rewarding weapon destruction would be an ideal addition. Reward team and mission outcomes more highly than sheer damage.
Assign player tiers based on ranking all players using the rolling average match scores of their most recent 20 matches. Update the 'tier' field in the database every day. It doesn't need to be exact.
There is no 'negative' scoring. There is no perceived punishment - no offensive little down-arrows on the match outcome screen. It's simply a matter of comparative ranking.
Players will quickly settle to a level which is about right for their skill level and will be easily able to climb out of a hole if they've been goofing around, levelling a mech or learning new game roles like learning to pilot a light.
Match make games to get an approximate equal mix of players of tiers in mechs of weight classes - equal numbers of same level players in same type of mechs would be ideal, but it obviously needs to be a bit fuzzy.
If you want to play high level competitive matches with your buddies, there's private queues for that. If you want to keep the existing player base playing, it's not going to be feasible to have one half playing boring meta and the other half playing dum-dum league. The top players will get bored and the bottom players will get frustrated.
Edited by NumberFive, 04 June 2020 - 06:44 PM.
#139
Posted 04 June 2020 - 06:54 PM
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: 401-600 does not move.
Match Score: 601+ goes up in PSR by +2
Player WINS:
Match Score: 0-50 goes down in PSR by -2
Match Score: 51-100 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
The bolded font indicates the changes from what Paul proposed in OP.
1. To reward only the best players on a loosing team.
2. To not reward non-team players, nor the AFK on winning teams.
#140
Posted 04 June 2020 - 06:55 PM
Thebackson, on 04 June 2020 - 02:30 PM, said:
1-3 get +5
4-6 get +3
7-10 get +1
11-14 get 0
15-18 get -1
19-21 get -3
22-24 get -5
This is also a net 0 PSR on the scores but rewards performance independent of whether you win or lose but also naturally favors the winning team since the winners are typically going to have a higher match score. At the same time it rewards a great player on a losing teaming and lowers the really bad players on a winning team.
The number of nights where I have had 8-12 loses in a row while at the same time pulling top 1-3 damage and match scores over 300 while still losing is staggering, under the proposed system I would do nothing but lose psr in those 8-12 matches.
Based on my previous post with a PSR rating system based off of pilot skill instead of who wins or losses I wanted to post an example scoring from a match.
https://imgur.com/a/A1WYHDS
You can see the winner has a more + than - yet the AFK NTG-H on the winning team ends up getting a -5 while 2 people on the losing side got a +5 due to being very effective even though their team loss. It is a net 0 system and pushes good player up to tier 1 and not just players who happen to be on the winning team.
From there you can adjust what gives a good match score separately if you want to reward certain types of play. IE reduce what AMS adds to match score and maybe give higher bonus to hit and run for an example.
Edited by Thebackson, 04 June 2020 - 07:15 PM.
4 user(s) are reading this topic
0 members, 4 guests, 0 anonymous users