actually we don´t not need class distribution balance alone but also traits/stats distribution balance
the game needs to know how many are tank how many are roamers how many are glasscanon how many are condi and so on to calculate only on mmr is not the best if you dont look at how the classes in a team are build
This 3% probably happens towards the top of the leaderboard, when a legendary/high platinum player can’t find a team to play with, they get some substitute. Where there’s a high pool of players queueing simultaneously, a substitute doesn’t make a great difference, but at the top it could be difference of 200-300 rating because there are 250 players within the 2100-1760 range in NA. Thus we get matches that are completely one-sided.
i expect that 3% means something over 100 MMR average diffrence by assuming lower average isn´t considdered a problem ? The problem is more divers it´s teams on one side, class distribution and MMR diffrences. Even a single substitute with way lower MMR on one side propably leading to less then 30 team average diffrence will create a lopsided and preset match (one player way lower then the rest is usualy causing a loss independet of team average). It´s not a slight tweak on substitute mechanic.
I wrote what shall be done above.
I just got rolled over by an esl player in unranked. I’m in bronze 2. I know it wasn’t ranked. This kind of match doesn’t inspire confidence in the system.
Yes, I tend to play on off hours. Still, this kind of match is not unusual for me.
I just want to say thank you to Evan for his wonderful transparency. Here we have a poster who documented a match-making issue that hindered his enjoyment of this game, and Evan acknowledged the problem and is actively trying to fix it.
Anyone who says Zerk is the average Joe build is an average Joe.
Yes its great. I jsut hope it´s not only doen on the surface for those 3%. It´s an oportunity to optimize a bit more in MM that doesn´t look to complicated. Of course its an additional module added.
MOAR DATA: ~3% of matches have substitutes which could* cause the original problem in this thread. We are testing fixes now.
What is % of such matches in the ratings 1700+?
I have a feeling that due to nature of dodging, sniping, afking during the season etc. that is happening there, the % is way higher than 3%. I might be wrong though.
I agree. Human nature tells me propable each match has substitutes. Of couse it might be that they are much closer in rating. If there are a lot of players available substitutes with close MMR will be much more available. If the 3% means similar rating gaps yes i belive it but this would indicate a very bad state and explane a lot bad matches.
To be honest the matches this season are as terrible as the seasons before. I play one account on NA and one on EU.
The majority of all the matches are not even close. Which means that one team must be packed with “good” and one with “bad” players.
I don’t see another reason why I win one match by 300+ points and lose the next one by 400+. It’s happening on both of my active accounts.
Can this be a random effect? I think that’s most unlikely.
[orz] below mediocre – we sponsor Arenanet
Piken Square EU, maybe soon on your server.
To be honest the matches this season are as terrible as the seasons before. I play one account on NA and one on EU.
The majority of all the matches are not even close. Which means that one team must be packed with “good” and one with “bad” players.
I don’t see another reason why I win one match by 300+ points and lose the next one by 400+. It’s happening on both of my active accounts.
Can this be a random effect? I think that’s most unlikely.
The problem actually is that it is random. Team formations should in fact not be random.
MOAR DATA: ~3% of matches have substitutes which could* cause the original problem in this thread. We are testing fixes now.
The asterisk after “could” is suspicious.
The astrix just means that there are substitutes in 3% of matches, but just the possibility of a substitute won’t mean you will get an imbalance, it could find a good substitute equivalent to the original intended player. So there’s a 3% chance a substitution happens on average, and only if one does is it possible to get an imbalance, but not guaranteed. Entirely appropriate * I think
MOAR DATA: ~3% of matches have substitutes which could* cause the original problem in this thread. We are testing fixes now.
and double staked duos?
this is not only substitutes fault, the “good” vs “bad” is a constant patron all wide matches, reviewing a random player history great part of games will be stomps, stomp victory or stomp defeat and something have to be made to adress this
Its good that evan looks at it but that 3% is only the tip of the iceberg. I would go that far that all players in a matc hshoudl be within 100 MMR. So the best and worst is maximum 100. With people not gettign ready this will lead to long Q times offpeak and Q breakoffs and thats what seems to be seen the lesser problem then blowouts. Well i prefer no match over a preset blowout.
Have one to add. I’m low gold so was my team. We played against a top 100 duo and no one was under platinum on the other team. on the plus side I only lost 4 MMR points.
I am sitting in a match facing Sind/Rom duo (top 5 player), they are like 500 ahead of me, i have teammates in gold (according to my teammates info) and 1 afk. I lose wooping 11 ranks (1 win worth of points vs normal team) for this. Evan where does it fit, please tell me? In what world 2200 = 1700? 11 points loss vs team that is 500~ ranks above me. Is this what you call fair pvp system?
I am Gold 2 and yesturday there was a 250 ap player in my team, twice, announcing to the team that he is a healer/buffer guardian and that he never lost a game so far. We lost that game.
How can a 250 ap account playing a guardian (not dh) healing staff buffer player enter gold 2 ranked games…multiple times?
Its good that evan looks at it but that 3% is only the tip of the iceberg. I would go that far that all players in a matc hshoudl be within 100 MMR. So the best and worst is maximum 100. With people not gettign ready this will lead to long Q times offpeak and Q breakoffs and thats what seems to be seen the lesser problem then blowouts. Well i prefer no match over a preset blowout.
Agreed.
Character: Henry rank million/Duke Henry
Necromancer/Casual Warrior
[Team] Best WvW guild of all time. EASILY.
I am sitting in a match facing Sind/Rom duo (top 5 player), they are like 500 ahead of me, i have teammates in gold (according to my teammates info) and 1 afk. I lose wooping 11 ranks (1 win worth of points vs normal team) for this. Evan where does it fit, please tell me? In what world 2200 = 1700? 11 points loss vs team that is 500~ ranks above me. Is this what you call fair pvp system?
The point loss does not matter. You should never be placed in a game with those players. The system should be much slower about expanding its matchmaking criteria.
Official winner of solo queue MMR leaderboards – EU
I am sitting in a match facing Sind/Rom duo (top 5 player), they are like 500 ahead of me, i have teammates in gold (according to my teammates info) and 1 afk. I lose wooping 11 ranks (1 win worth of points vs normal team) for this. Evan where does it fit, please tell me? In what world 2200 = 1700? 11 points loss vs team that is 500~ ranks above me. Is this what you call fair pvp system?
The point loss does not matter. You should never be placed in a game with those players. The system should be much slower about expanding its matchmaking criteria.
I agree however it also really bothers me how crappy the rating gains are. I have won matches vs top rank 1 before, but even if you win vs duoqs that are so much higher rated than you are, you still gain almost nothing in comparison. The rating gap is so huge yet it is absolutely unrewarding to win vs such teams and extremely punishing to lose vs them. For some reason the system assumes 1700 equals 2200 when it calculates rating gains/losses – wth?
I am sitting in a match facing Sind/Rom duo (top 5 player), they are like 500 ahead of me, i have teammates in gold (according to my teammates info) and 1 afk. I lose wooping 11 ranks (1 win worth of points vs normal team) for this. Evan where does it fit, please tell me? In what world 2200 = 1700? 11 points loss vs team that is 500~ ranks above me. Is this what you call fair pvp system?
The point loss does not matter. You should never be placed in a game with those players. The system should be much slower about expanding its matchmaking criteria.
I disagree, im 100% ok with playing players 100-200 points ahead of me and it happens daily if i play more then 2-3 games.
What i find unfair is losing 10-15 points when none of them were placed on my team.
Like if i had 2 ESL players on my team and there was 2 on the other, ok.
Whats happening is 4 of the top 250 players on the other team and gold players on the other.
maybe there aren’t that many players in the pool to begin with, so the matchmaking system grabs players usually outside of its normal parameters.
just maybe.
So you’re saying that just because the player pool is small, that no attempt to balance out the two teams should be made, and even if the lower rated teams beat the odds and won, they should barely gain any rating for it?
Doc Von Doom – Asuran Necromancer
Gate of Madness
Contribute to the Wiki MetaBattle Builds
First game today, game instantly popped after clicking to queue. Lost 500-20.
Now mostly my games are very one-sided, Normal is to have 500-100 loss or win, I’ve been trying to track the results a bit and around 2/12 the loser gets over 300 points. The games are almost always one-sided from the start. I haven’t seen something like comeback in two weeks or so.
First game today, game instantly popped after clicking to queue. Lost 500-20.
Now mostly my games are very one-sided, Normal is to have 500-100 loss or win, I’ve been trying to track the results a bit and around 2/12 the loser gets over 300 points. The games are almost always one-sided from the start. I haven’t seen something like comeback in two weeks or so.
Just to provide some more stats to the last 100 games- based on end score:
58 games > 150 points difference.. so ‘42’ games could be considered close.
10 games > 400 points difference
21 games > 300 points difference
47 games > 200 points difference
Even 10% of unwinnable games > 400 points difference is bad.
21% of unwinnable games > 300 points difference : also bad.
It’s from figures pulled from gw2pvp.de – I’m finding my win/loss ratio on a profession is forcing. This is from a climb from ~1200-1700 down to 1600 as majority of last games were plain unwinnable.
Not even maybe lol… This is a small kitten pool. Let’s call it a tub
this idea of tub intrigues me. I daresay we take this metaphor even further. suds shall henceforth be direct damage builds (since the tub is the total of players, the water must be the players, and suds keep it all clean). what say you good sir?
So you’re saying that just because the player pool is small, that no attempt to balance out the two teams should be made, and even if the lower rated teams beat the odds and won, they should barely gain any rating for it?
I merely stated a possible why, my good chap! such declaration unto my most humble self is most untoward!
So you’re saying that just because the player pool is small, that no attempt to balance out the two teams should be made, and even if the lower rated teams beat the odds and won, they should barely gain any rating for it?
I merely stated a possible why, my good chap! such declaration unto my most humble self is most untoward
You posted about small populations being the issue as if everyone in the game doesn’t already realize this in response to a thread discussing the issue of teams being stacked and gaining unreasonably low rating for beating better teams while losing too much rating for losing to those better teams. I was merely inquiring as to how your low population comment addressed the issues brought up in the thread.
Doc Von Doom – Asuran Necromancer
Gate of Madness
Contribute to the Wiki MetaBattle Builds
You posted about small populations being the issue as if everyone in the game doesn’t already realize this in response to a thread discussing the issue of teams being stacked and gaining unreasonably low rating for beating better teams while losing too much rating for losing to those better teams. I was merely inquiring as to how your low population comment addressed the issues brought up in the thread.
well I suppose my response doesn’t address said issues! hopefully it helped to inform some unwitting denizens! I would have thought my intent clear on the matter.
Team Deathmatch for PvP – Raise the AoE cap for WvW – More unique events for PvE
maybe there aren’t that many players in the pool to begin with, so the matchmaking system grabs players usually outside of its normal parameters.
just maybe.
still doesn’t justify abysmal rating gains if you manage to win vs those teams and absurd rating loses
Heh I had a -13 loss last night when I had no one in the top 250 (doubt even plat) and I vs’d the #3 duo q. Lost 592-10 but I lose 13 rating sigh lol. I was under 1.7k MMR solo q.
Evan will never agree that there are multiple problems with the system unless a top popular player says it first. Getting him to “test” something…look at the effort it took after how many weeks of Season 5? Season’s pretty much over.
System abuse will remain. Top players will remain where they are, rest will have to grind for it.
If you aren’t playing Mesmer or Thief this season, good luck. Have fun.
Evan feel free to poke around the hilarious fluctuation on my account. I’m garbage!
More of matchmaking’s greatest matches, ANET I really wanted to enjoy this season but this is so tiring… I decide to q a few games and 2nd game of the evening I get into this?
2xDuo top10 and top100 obviously vs a full solo of 1550-1600 rating??? WTF
Evan will never agree that there are multiple problems with the system unless a top popular player says it first. Getting him to “test” something…look at the effort it took after how many weeks of Season 5? Season’s pretty much over.
System abuse will remain. Top players will remain where they are, rest will have to grind for it.
If you aren’t playing Mesmer or Thief this season, good luck. Have fun.
Evan feel free to poke around the hilarious fluctuation on my account. I’m garbage!
Actually, I’m not a top player, but I created a thread and provided data from the game and presented logical questions concerning the matchmaking system and the way it stacks teams, and Evan dropped by to let me know he’ll be looking into the issue soon as he addresses the issue that was brought up in this thread.
Doc Von Doom – Asuran Necromancer
Gate of Madness
Contribute to the Wiki MetaBattle Builds
I agree however it also really bothers me how crappy the rating gains are. I have won matches vs top rank 1 before, but even if you win vs duoqs that are so much higher rated than you are, you still gain almost nothing in comparison. The rating gap is so huge yet it is absolutely unrewarding to win vs such teams and extremely punishing to lose vs them. For some reason the system assumes 1700 equals 2200 when it calculates rating gains/losses – wth?
Your mistake is that you believe – incorrectly – that rating loss or gain is related to the MMR of your team vs the other time.
MMR changes are based on two things:
The first is the “win or lose” prediction from the respective MMRs vs the actual win or lose, which determines if you gain or lose MMR from this fight.
The second is how confident the MMR system is of your ranking, which determines how big the change is. So, a large MMR change means the system thinks you are far from your correct MMR, and a small change means it thinks you are close to your correct MMR.
Which is to say: your result was close to the expected result, like most of them have been recently, leading to high confidence your MMR is correct, leading to small changes as a result of the outcome of the fight.
I agree however it also really bothers me how crappy the rating gains are. I have won matches vs top rank 1 before, but even if you win vs duoqs that are so much higher rated than you are, you still gain almost nothing in comparison. The rating gap is so huge yet it is absolutely unrewarding to win vs such teams and extremely punishing to lose vs them. For some reason the system assumes 1700 equals 2200 when it calculates rating gains/losses – wth?
Your mistake is that you believe – incorrectly – that rating loss or gain is related to the MMR of your team vs the other time.
MMR changes are based on two things:
The first is the “win or lose” prediction from the respective MMRs vs the actual win or lose, which determines if you gain or lose MMR from this fight.
The second is how confident the MMR system is of your ranking, which determines how big the change is. So, a large MMR change means the system thinks you are far from your correct MMR, and a small change means it thinks you are close to your correct MMR.
Which is to say: your result was close to the expected result, like most of them have been recently, leading to high confidence your MMR is correct, leading to small changes as a result of the outcome of the fight.
and this is bad thinking. a ELO, glico or what is it is a ELO evolution, main source of improvement/decreasement is the diference of you vs your oponent, losing whith a far better oponent conducts to 0 or irrelevant change a victori vs a better one is max profit, a lose vs one worse oponent is devastating and that thing that only thing not implemented properly is that makes top ones be hapy whith that matchmaking, they play again worse ones whith some ¿legal? manipulations duo, ofpeak hours, qskiping other tops… and in the rare case they loose the damage is minimum, they are stablished in top by system.
a proper thing is if they lose those matches they have to be devastated for mmr lost throwing several positions and not willing to have another match with this devastating efec, prefering the 50-60% win ratio that is the correct thing playing whith her truly equals and having low mmr gain loses
mmr system confidence have to be a factor to matchmaking, not to mmr calculation
Your mistake is that you believe – incorrectly – that rating loss or gain is related to the MMR of your team vs the other time.
MMR changes are based on two things:
The first is the “win or lose” prediction from the respective MMRs vs the actual win or lose, which determines if you gain or lose MMR from this fight.
Technically not a mistake, as the win or lose prediction takes into account both team’s MMR. Indirectly related if anything.
The mistake is assuming that rating change is SOLELY tied to team MMRs. And it’s also probably more about the team’s average MMR as opposed to an individual player. While you may face off against a rank 1 player, their team isn’t all made up of rank 1 players. If you legitimately win a match with a team average of 1700 against a team average of 2200, your rating gain will definitely not be minimal.
I agree however it also really bothers me how crappy the rating gains are. I have won matches vs top rank 1 before, but even if you win vs duoqs that are so much higher rated than you are, you still gain almost nothing in comparison. The rating gap is so huge yet it is absolutely unrewarding to win vs such teams and extremely punishing to lose vs them. For some reason the system assumes 1700 equals 2200 when it calculates rating gains/losses – wth?
Your mistake is that you believe – incorrectly – that rating loss or gain is related to the MMR of your team vs the other time.
MMR changes are based on two things:
The first is the “win or lose” prediction from the respective MMRs vs the actual win or lose, which determines if you gain or lose MMR from this fight.
The second is how confident the MMR system is of your ranking, which determines how big the change is. So, a large MMR change means the system thinks you are far from your correct MMR, and a small change means it thinks you are close to your correct MMR.
Which is to say: your result was close to the expected result, like most of them have been recently, leading to high confidence your MMR is correct, leading to small changes as a result of the outcome of the fight.
and this is bad thinking. a ELO, glico or what is it is a ELO evolution, main source of improvement/decreasement is the diference of you vs your oponent, losing whith a far better oponent conducts to 0 or irrelevant change a victori vs a better one is max profit, a lose vs one worse oponent is devastating and that thing that only thing not implemented properly is that makes top ones be hapy whith that matchmaking, they play again worse ones whith some ¿legal? manipulations duo, ofpeak hours, qskiping other tops… and in the rare case they loose the damage is minimum, they are stablished in top by system.
a proper thing is if they lose those matches they have to be devastated for mmr lost throwing several positions and not willing to have another match with this devastating efec, prefering the 50-60% win ratio that is the correct thing playing whith her truly equals and having low mmr gain loses
mmr system confidence have to be a factor to matchmaking, not to mmr calculation
Judging by the lack of capitalization (among other things) I’m guessing this was typed from a mobile device. Never the less, that’s one hell of a run on sentence XD
The issue is that instead of Evan trying to fix and balance Pvp, he’s trying to fix a stupid matchmaking algorithm that won’t ever work because you can’t possibly balance randoms vs teams.
Try as he might, he will fail everytime. The sooner he realizes that and begins spliting true random queue from team-queue, the better.
It’s a nice dream to have such an algorithm but it ain’t gonna happen. Just admit it.
The issue is that instead of Evan trying to fix and balance Pvp, he’s trying to fix a stupid matchmaking algorithm that won’t ever work because you can’t possibly balance randoms vs teams.
Try as he might, he will fail everytime. The sooner he realizes that and begins spliting true random queue from team-queue, the better.
It’s a nice dream to have such an algorithm but it ain’t gonna happen. Just admit it.
There’s a different team that handles balance. Evan isn’t a one man show for everything PvP.
Found a couple minor bugs in addition to the lack of reshuffling. I’ve also worked on a new version of the matchmaker that will improve roster size and profession matching. We will likely trial it during the off-season.
