PVP Scoring works how?

[Trevor Boyer.6524]

@Lord of the Fire.6870 said:Okay this how for I know it works

First you have your 10 matchups which put you somewhere in the rang.Then you have about 20 games in which the amount of points added and reduced are contently deceasing from +-25 to +-12/13

The amount of points you gain or lose doesn't have much to do how you perform and the rank difference little impact the scoring.

1.) The least amount of reduced point I ever had was -9 and this was matchup against a group 2 whole ranks above me and my team. This is pretty broken if anything say like silver 3 plays against gold 3 in case of S3 loses they should lose only 1 point and G3 should only gain 1 point. On the other side winning from S1 should give them 26 points giving the difference in rang naturally G3 loses then the same amount of points.

Before someone cries this is unfair for the G3 guys this matchup is already horrible and in other games it might works differently there they take the difference between winner and looser and transfer a certain % of point from the looser to the winner. This also means when you reach the top it becomes really hard to gain point only tournament giving the top players then chance to gain more points.

2.)From my personal experience I know how the matchmaking calculate a groups strength is also broken . What it does is taking the best player in one group and then tries to build a group that match it. This also works in unranked if the rank difference is great enough this can result in an awful experience.

3.) Even with everything out of the way there is no personal scoring what you could do it is for each top stats you get one point less reduction.

EDIT:

Basically how the system works now is your ranking is determinant after 30 rounds from there on moving upwards is super hard and need a lot of grind also your performance are considered near to none at this point.

This is also one of the explanation why weekends PvP sucks so much(on EU at least) there people who plays mainly at the weekend but their scoring is actually in your rank range because the fight a lot only each other during the weekend. The current system doesn't punish them enough to put them into to their actual rank.

Same can be said about bots also the current system scores your own performance so less and preference grind this support bots a lot.

For support players even if they nearly deleted the possibility to play them I will say this is also bad if you fall under gold 3 in EU your are screwed because people often don't group up there and your personal performance don't count your end up getting sucked downwards .

Also I think the reason GW2 got kicked out of the ESL has also something to do with such a system.

1. GW2 optionally left the ESL when the financial investment was no longer worth the profit.
2. Ranked rating had absolutely nothing to do with ESL, as everything ESL related was ran with organized 5 man teams in tournament format only.

But you're right on one thing. Everything about Glicko is terrible for a 5v5 ranked mode that pairs you randomly with other players. Literally everything that could have gone wrong with the theory behind Glicko in GW2, did go wrong.

[Bazsi.2734]

@John.8507 said:Finally won a match after 5 losses, and got a measly 10 points? Whats the point:O

Rating change is only based on who you've beat/lost to. The game paired you up with a team that was much weaker than you(at least based on their rating). Also if you really hit the wall around 1100, you do not want to climb just yet. If you were to be matched against 1600-1700-ish rated plat players, that would just be a curbstomp.Play against players on your own level, improve and you'll climb in the process.If you aren't climbing, thats because you're not improving.

[Lord of the Fire.6870]

@Bazsi.2734 said:

@John.8507 said:Finally won a match after 5 losses, and got a measly 10 points? Whats the point:O

Rating change is only based on who you've beat/lost to. The game paired you up with a team that was much weaker than you(at least based on their rating). Also if you really hit the wall around 1100, you do not want to climb just yet. If you were to be matched against 1600-1700-ish rated plat players, that would just be a curbstomp.Play against players on your own level, improve and you'll climb in the process.If you aren't climbing, thats because you're not improving.

I wouldn't call this problem and solution . I can show you when I play weekend 10-15 loses in a row and this happens only weekends. Well we could argue if we/I too low in rank or them I can only call this kind of situation nightmarish . The current system is stupid, idiotic and unfair because it doesn't put the players in their actually rank fast enough . It is also unfair because it doesn't include the actually rank of the enemy team vs you in the point system. On the other hand if you have enough lucky encounters with out you contributing much you can go up to the highest rank.

[Bazsi.2734]

@Lord of the Fire.6870 said:

@Bazsi.2734 said:

@John.8507 said:Finally won a match after 5 losses, and got a measly 10 points? Whats the point:O

Rating change is only based on who you've beat/lost to. The game paired you up with a team that was much weaker than you(at least based on their rating). Also if you really hit the wall around 1100, you do not want to climb just yet. If you were to be matched against 1600-1700-ish rated plat players, that would just be a curbstomp.Play against players on your own level, improve and you'll climb in the process.If you aren't climbing, thats because you're not improving.

I wouldn't call this problem and solution . I can show you when I play weekend 10-15 loses in a row and this happens only weekends. Well we could argue if we/I too low in rank or them I can only call this kind of situation nightmarish .

Takes 30 games for me to reach my rating on average. That is fast enough for me. If you take much much more than that, maybe you're overestimating your own skill in relation to the playerbase.

It is also unfair because it doesn't include the actually rank of the enemy team vs you in the point system.

I don't know what you mean by this, it literally does.

On the other hand if you have enough lucky encounters with out you contributing much you can go up to the highest rank.

And some people win the lottery. You're hearing about it all the time, yet in actual reality it never happens.

[sitarskee.5738]

It's hard for me to tell. There are seasons where I get placed gold3/plat1 after 10 first games and keep it for the whole season and there are seasons where I get placed silver2/3 and have a freaking hard time winning games there and stay there most of the time until I quit of frustration.

So yeah, PvP is pretty random unless you're very skilled.

[Smoosh.2718]

https://clips.twitch.tv/VivaciousMoldyLyrebirdSpicyBoy

Tbh This is how Ive been looking at it for quite a while with PvP, the rank system means nothing, the balance is out of wack and you can abuse builds to carry your wins or make the game not fun for others.

I'll do PvP ranked or unranked but I'll never see a leader board as an epeen stroker, since there is no 1v1 ranking or no 5v5 team ranking, which would actually mean something. Ranking based on how good your fellow randoms were leaves a very bad taste in the mouth. So best thing to do, play the game, have fun and don't get so hell bent on believing the ranking means something in this game.

[Lord of the Fire.6870]

@Bazsi.2734 said:

@Lord of the Fire.6870 said:

@Bazsi.2734 said:

@John.8507 said:Finally won a match after 5 losses, and got a measly 10 points? Whats the point:O

Rating change is only based on who you've beat/lost to. The game paired you up with a team that was much weaker than you(at least based on their rating). Also if you really hit the wall around 1100, you do not want to climb just yet. If you were to be matched against 1600-1700-ish rated plat players, that would just be a curbstomp.Play against players on your own level, improve and you'll climb in the process.If you aren't climbing, thats because you're not improving.

I wouldn't call this problem and solution . I can show you when I play weekend 10-15 loses in a row and this happens only weekends. Well we could argue if we/I too low in rank or them I can only call this kind of situation nightmarish .

Takes 30 games for me to reach my rating on average. That is fast enough for me. If you take much much more than that, maybe you're overestimating your own skill in relation to the playerbase.

It is also unfair because it doesn't include the actually rank of the enemy team vs you in the point system.

I don't know what you mean by this, it literally does.No it doesn't it takes your point ratio in the matchup who you fighting is irrelevant as yourself said before. What other PvP games do is take the ranking of who you won/lost to and gives / takes from you according to it points + how you performed. This varies from game to game a bit but in general the good games do this this way.

On the other hand if you have enough lucky encounters with out you contributing much you can go up to the highest rank.

And some people win the lottery. You're hearing about it all the time, yet in actual reality it never happens.If you looking in this form you can find core eles in platin rank which tells you everything.

True you can still have this luck with this other system but the probability is much lower because you can't feed on lower ranked this much and when you lose once against such a team you fall harder then in the current system. Naturally this works also in reverse. This also works against match up manipulation where people tries to snipe noobs to farm points/rank at certain times.

@"Smoosh.2718" said:https://clips.twitch.tv/VivaciousMoldyLyrebirdSpicyBoy

Tbh This is how Ive been looking at it for quite a while with PvP, the rank system means nothing, the balance is out of wack and you can abuse builds to carry your wins or make the game not fun for others.

I'll do PvP ranked or unranked but I'll never see a leader board as an kitten stroker, since there is no 1v1 ranking or no 5v5 team ranking, which would actually mean something. Ranking based on how good your fellow randoms were leaves a very bad taste in the mouth. So best thing to do, play the game, have fun and don't get so hell bent on believing the ranking means something in this game.

My point.

What I could add here is the realization why we have no possibility on players joining a running game until a certain point or surrender vote is it would manipulate the points at the end of the matchup which is everything the current system is running on because of this it is not allowed. :s

[Bazsi.2734]

@Lord of the Fire.6870 said:

@Bazsi.2734 said:

@Lord of the Fire.6870 said:

@Bazsi.2734 said:

@"John.8507" said:Finally won a match after 5 losses, and got a measly 10 points? Whats the point:O

Rating change is only based on who you've beat/lost to. The game paired you up with a team that was much weaker than you(at least based on their rating). Also if you really hit the wall around 1100, you do not want to climb just yet. If you were to be matched against 1600-1700-ish rated plat players, that would just be a curbstomp.Play against players on your own level, improve and you'll climb in the process.If you aren't climbing, thats because you're not improving.

I wouldn't call this problem and solution . I can show you when I play weekend 10-15 loses in a row and this happens only weekends. Well we could argue if we/I too low in rank or them I can only call this kind of situation nightmarish .

Takes 30 games for me to reach my rating on average. That is fast enough for me. If you take much much more than that, maybe you're overestimating your own skill in relation to the playerbase.

It is also unfair because it doesn't include the actually rank of the enemy team vs you in the point system.

I don't know what you mean by this, it literally does.No it doesn't it takes your point ratio in the matchup who you fighting is irrelevant as yourself said before. What other PvP games do is take the ranking of who you won/lost to and gives / takes from you according to it points + how you performed. This varies from game to game a bit but in general the good games do this this way.

No I didn't say anything like that, actually READ what I've typed out. Here is a thread where the OP is full of links and explanation on how glicko-2 works:https://en-forum.guildwars2.com/discussion/58477/glicko-2-algorithm-put-into-code-updated-conclusion-about-win-streaks/p1(If the OP of this thread is still reading, in the linked material is the exact math that answers your original question)

Just think about it, if your own rating wouldn't be a factor in who you're playing against, what would be the point to having matchmaking at all?

On the other hand if you have enough lucky encounters with out you contributing much you can go up to the highest rank.

And some people win the lottery. You're hearing about it all the time, yet in actual reality it never happens.If you looking in this form you can find core eles in platin rank which tells you everything.

Which tells me what exactly? That good players can carry less viable specs into "high" ratings? We have balance problems but its not THAT bad. I myself wouldn't play this game if choosing a spec would determine my rating.

True you can still have this luck with this other system but the probability is much lower because you can't feed on lower ranked this much and when you lose once against such a team you fall harder then in the current system. Naturally this works also in reverse. This also works against match up manipulation where people tries to snipe noobs to farm points/rank at certain times.

Everything seems to have a different definition with you... match manipulation is where you have at least one mole in the enemy team, and that player throws the match so you can win. Playing at certain times and being matched up against noobs certainly does not qualify as manipulation by itself.

[Rose.2593]

@Bazsi.2734 said:

@John.8507 said:Finally won a match after 5 losses, and got a measly 10 points? Whats the point:O

Rating change is only based on who you've beat/lost to. The game paired you up with a team that was much weaker than you(at least based on their rating). Also if you really hit the wall around 1100, you do not want to climb just yet. If you were to be matched against 1600-1700-ish rated plat players, that would just be a curbstomp.Play against players on your own level, improve and you'll climb in the process.If you aren't climbing, thats because you're not improving.

I'm pretty sure I'm improving personally, my placement matches went really well. But then ended up losing 7 in a row so went right down again. You always loose more points when you loose than you gain for a win, Which yes I get your supposed to win more than you lose. But just a bit annoying that its so hard to get back to where you were after a few losses.

[Luthan.5236]

The most annoying thing is when you lose hard fight against strong opponents ... and losing the match. When then it gives you a big rating loss. I know it is supposed to give less rating loss against stronger enemies. And we don't see the ratings.

And there might be players that play below their "true" rating (still having to go up in rating after some loss streak maybe - therefore you might get pwned hard by a team where the system thinks they are weaker than they should be ... giving you a bigger rating loss when you lose.

Is is really discouraging at the end of a match. When you think "wow they pwnd us hard ... okay 12-13 rating loss" (when you usually get these numbers) ... then suddelny -15 appears.

On the other hand pretty fun to have it vice-versa: Pwning the enemy hard and expecting some low rating gain against super weak enemies ... and system gives you a +14 or +15. :D

They could as well announce the expected gain/loss at the beginning - so you know how hard you have to try. If you get "bad start" and yo know you might lose big rating (system expecting you to win) more people might motivated to try harder ... instaed of giving up early. (Some cry after a bad startd and get demotivated - which means they perform worse than they could.)

[Megametzler.5729]

GLICKO is a very sound and verified rating system. It can be criticized for a couple of things (not measuring the "clearness" of a victory/loss for example), but it remains pretty accurate in general (in the example, both sides are not valued so you dont lose more for a clear loss).

However, the matchmaker struggles with several things indeed, mainly the low population, but other factors as well (duoQ, class switching/stacking etc.). Here comes a significant RNG factor into the equation - people suddenly duoQing, playing different builds and all that.

Many people confuse these two systems and do not know what is doing what. Read these:http://www.glicko.net/glicko/glicko.pdfhttps://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm

[BadHealer.3608]

The problem is that the points depend on the rank of you, the people in your team and the opposite team. This makes it very hard to understand. So if you are platinum and play with legendary player against bronze players you will get not many points if you win, but will lose a lot if you are losing. On the other hand, if you are playing with bronze players against legendary players and win, you will get a lot of points, but losing wont lose that many points.

An additional information about how much rank different where in your match would be a good way to understand the points, but this isn't in the game.

[TenzKu.1509]

So, the number of points you loose and gain is based on how 'good' the system thinks the other team is.  This often results in really close games getting scored really poorly.

 

I get that.   It's broken...

 

But, why can't it be better?  I mean come on.  All those micro-transactions can't pay for somebody to look at the code and go... 

 

Gee.  We need to find a way to factor in point spread and individual performance.  I mean, really now -- figure it out.

 

[Eddbopkins.2630]

21 minutes ago, TenzKu.1509 said:

So, the number of points you loose and gain is based on how 'good' the system thinks the other team is.  This often results in really close games getting scored really poorly.

 

I get that.   It's broken...

 

But, why can't it be better?  I mean come on.  All those micro-transactions can't pay for somebody to look at the code and go... 

 

Gee.  We need to find a way to factor in point spread and individual performance.  I mean, really now -- figure it out.

 

Development time and hours allocated to the pvp game mode are not allowed by anet. Pvp is not something arenanet wants to development and maintain and improve. The proff is in the very bare minimum of work and development pvp gets.

[Tanbin.9402]

its a decent matchmaker, its just there's virtually no consistent rank population in this game, its scattered, so often times you'll get 2 gold 1-2 players as a plat2 while your other team may have almost all plat players

 

it's a coin flip in this game, having a duo helps a lot though because it helps stabilize that rng

[razaelll.8324]

9 minutes ago, Tanbin.9402 said:

its a decent matchmaker, its just there's virtually no consistent rank population in this game, its scattered, so often times you'll get 2 gold 1-2 players as a plat2 while your other team may have almost all plat players

Acording to the algorithm published this is not the case

[Tanbin.9402]

7 hours ago, razaelll.8324 said:

Acording to the algorithm published this is not the case

It uses Glicko-2, which is used in a myriad of platforms. Would you be concerned about the RD (deviation)?

<Ratings>
  <Rating default="1200" min="100" max="5000" max-change="300" profession-ratio="0.0"/>
  <Deviation default="350" min="30" max="350" period="3d" max-periods="20"/>
  <Volatility default="0.06" min="0.04" max="0.08" system-constant="0.5"/>
</Ratings>

RD is defaulted at 350 from GW2's configuration, which is standard for Glicko-2.

 

µ = (r − 1200)/173.7178

φ = RD/173.7178

 

This assumes a player would play 1 ranked game and induce a deviation of 0.06 volatility over time, of course these calculations aren't updated after every match (that would be dumb). 

 

µ = rating

 

Playing against n opponents with ratings µ1, . . . , µn

 

v =   Xn j=1 g(φj ) 2E(µ, µj , φj ){1 − E(µ, µj , φj )}   −1

 

Reverse engineering RD from Glicko-2 is basically: rating change = function (game result, RD, rating difference)

 

To visually see this you can do this in a sandbox, just npm glicko2's javascript package based on GW2's configurations. Make dummy matches based on the deviations -> New Player -> https://codepen.io/superkaratemonkey/pen/JjNdgME

 

On paper and in the sandbox you can see the ratings work fine, and are good, if the population is good. With an inconsistent population, the matchmaker is forced to find and search players from different rank divisions to accommodate for the lack of rank you currently are at. This search is done on both ends, if you are in gold 1-2 and manage to find yourself in a match against top 5's. Or you are with top 5's and a few gold 1-2's and find yourself against all plat players like I mentioned. 

 

The elo system is mathematically inferior to glicko-2 on paper, and they both obviously reach the same goal but Glicko-2 is designed to reach that goal faster, which makes sense given the amount of time a season is compared to LoL's Elo system vs GW2's Glicko-2 system. Faster rating computation is technically better for quicker seasons but comes at a cost with lower population.

 

With all of this, I would say the matchmaking itself is indeed fine, and works as it should. With the given population, within the pseudo-code is where the issues arise due to lack of population.

def tryMakeMatch(target, queue, config)

It will attempt to match against qualified opponents, and then will use padding and a filtering process to find whatever they can after the fact. I could be wrong on this though. 

 

Quote

Over time, padding is added to your player rating. While this may decrease match quality, it helps ensure that outliers still receive matches.

 

I generally feel the padding is indeed more prevalent due to population concerns more so than adjusting to find you a better match in the matchmaking protocol due to being an "outlier". 

 

With that I would say the system itself that's in place is fine, and works fine on other game platforms as well. But those games have populations that span far greater than ours with exponentially more members within their rank divisions to accommodate for the matchmaking algorithm.

 

You'll constantly find in this game being matched with players that shouldn't be anywhere near you in the matchmaking algorithm especially during specific hours throughout the day/night. A lot of players will abuse this and que at night, because they know padding and filtering from the example code above will work in their favor to help farm up rank.

 

Edited July 2, 2021 by Tanbin.9402

[razaelll.8324]

56 minutes ago, Tanbin.9402 said:

It uses Glicko-2, which is used in a myriad of platforms. Would you be concerned about the RD (deviation)?

<Ratings>
  <Rating default="1200" min="100" max="5000" max-change="300" profession-ratio="0.0"/>
  <Deviation default="350" min="30" max="350" period="3d" max-periods="20"/>
  <Volatility default="0.06" min="0.04" max="0.08" system-constant="0.5"/>
</Ratings>

RD is defaulted at 350 from GW2's configuration, which is standard for Glicko-2.

 

µ = (r − 1500)/173.7178

φ = RD/173.7178

 

This assumes a player would play 1 ranked game and induce a deviation of 0.06 volatility over time, of course these calculations aren't updated after every match (that would be dumb). 

 

µ = rating

 

Playing against n opponents with ratings µ1, . . . , µn

 

v =   Xn j=1 g(φj ) 2E(µ, µj , φj ){1 − E(µ, µj , φj )}   −1

 

Reverse engineering RD from Glicko-2 is basically: rating change = function (game result, RD, rating difference)

 

To visually see this you can do this in a sandbox, just npm glicko2's javascript package based on GW2's configurations. Make dummy matches based on the deviations -> New Player -> https://codepen.io/superkaratemonkey/pen/JjNdgME

 

On paper and in the sandbox you can see the ratings work fine, and are good, if the population is good. With an inconsistent population, the matchmaker is forced to find and search players from different rank divisions to accommodate for the lack of rank you currently are at. This search is done on both ends, if you are in gold 1-2 and manage to find yourself in a match against top 5's. Or you are with top 5's and a few gold 1-2's and find yourself against all plat players like I mentioned. 

 

The elo system is mathematically inferior to glicko-2 on paper, and they both obviously reach the same goal but Glicko-2 is designed to reach that goal faster, which makes sense given the amount of time a season is compared to LoL's Elo system vs GW2's Glicko-2 system. Faster rating computation is technically better for quicker seasons but comes at a cost with lower population.

 

With all of this, I would say the matchmaking itself is indeed fine, and works as it should. With the given population, within the pseudo-code is where the issues arise due to lack of population.

def tryMakeMatch(target, queue, config)

It will attempt to match against qualified opponents, and then will use padding and a filtering process to find whatever they can after the fact. I could be wrong on this though. 

 

 

I generally feel the padding is indeed more prevalent due to population concerns more so than adjusting to find you a better match in the matchmaking protocol due to being an "outlier". 

 

With that I would say the system itself that's in place is fine, and works fine on other game platforms as well. But those games have populations that span far greater than ours with exponentially more members within their rank divisions to accommodate for the matchmaking algorithm.

 

You'll constantly find in this game being matched with players that shouldn't be anywhere near you in the matchmaking algorithm especially during specific hours throughout the day/night. A lot of players will abuse this and que at night, because they know padding and filtering from the example code above will work in their favor to help farm up rank.

 

Its highly unlikely to have 2 gold players in 1 team and all plat+ players in the other.. it is much more likely to place 1 gold player in each team, thats what i ment. 

Edited July 2, 2021 by razaelll.8324

[Tanbin.9402]

4 hours ago, razaelll.8324 said:

Its highly unlikely to have 2 gold players in 1 team and all plat+ players in the other.. it is much more likely to place 1 gold player in each team, thats what i ment. 

Ah ok sorry I understand now

[JusticeRetroHunter.7684]

15 minutes ago, Tanbin.9402 said:

Ah ok sorry I understand now

 

Hey Tanbin. The Filtering process that occurs in the 2nd part of the matchmaker is the scoring setup which determines which team you are going to be on. So i think the way it works is that the matchmaker selects 10 players to start with first as if all 10 players are on a single team (and it uses rating and padding and all that to find those 10 people)...it will then use that scoring process to tell the matchmaker which person will belong to which team. The Glicko System might work, but the filtering process effects the result that the Glicko system produces.

 

I don't know the impact of each component of the scoring system has on the matchmaker...but I'm pretty sure that all it does is weight the rating with the highest relative score. In other words, rating is just one of a number of components in how the matchmaker creates matches.

 

nice link btw. Really useful 👍

 

[razaelll.8324]

4 hours ago, Tanbin.9402 said:

Ah ok sorry I understand now

No worries at all, buddy. My post was not very clear either.

[Faux Play.6104]

On 5/16/2020 at 2:06 PM, John.8507 said:

How does the scoring work.

Played 3 matches today, 1 match lost 500 - 475 so that was close, but lost -17 points. 2nd match Won match 500 - 227 and only gained 7 points, then won again 502 - 352 and gained 13 points.

So even though I won 2, lost 1, I only overall gained 3 points?

Doesn't seem like will ever move up, and most the time you lose more points for a loss than you gain for a win. It should be the other way round.

The link below has a high level summary of the rating system.  As you play more matches your rating deviation will go down enough that when you are playing teams at your rating level you will gain/lose around 11-13 pts per match.  Since ratings can change fairly quickly, and there is the reset at the beginning of the season there will be people that are outside their rating level in some matches.  When I ran the numbers from the wiki, the rating deviation settles at around 60.  It is equivalent to a standard deviation so that means that the system is 95% sure your rating is +/- 2 deviations or +/- 120 rating pts.

PSA how placements work - Page 2 - Player vs. Player - Guild Wars 2 Forums

[Lightfoot.5438]

 

 

..

Edited July 6, 2021 by Lightfoot.5438
clarity