Guild Wars 2

I’m going to break this down into three very short sections for simplicity’s sake.

Start of Season

• Placements: 6 wins, 4 Losses, placed in Gold T2
• Next few weeks was a relatively, not-so-frustrating climb, eventually reaching a peak Rank of #16.

Mid Way, Leading to End of Season

• I kept queueing and ended up deranking myself from #16 to off the leaderboards, over night! (drunk rip, but that’s no excuse).
• Darn near took the rest of the reason just to break out of Gold.
• Started to play with the same players over and over… to the point where I had to queue dodge a person for having him on my team 3 games in a row… first season I ever had to queue dodge some one.

“Top” Players Still Cheating the System

• Players having multiple alts, friends of friends queueing solo together just to throw if they fight against each other.
• Player’s on alts throwing the game if they get teamed up with another player in top 25.
• I guess queue dodging is a bigger thing than I thought because nearly everyone on my Friends list does it.
• I realized I should have used my gf’s alt account just to prevent myself from playing more games once reaching my desired rank. I realize that now.
• More players are succeptible to “throw games” on alt accounts. (an obvious, but much needed statement)

Season Conclusion & what we Need

• Matchmaking has been the worse… it’s 99% the cause of this season’s frustrations.

• We need a vote / report system in-game just like LoL and Overwatch. You report for player Idle/Match manipulation/etc and enough votes automatically judge a player for a temp Dishonor or temp Ban on their IP address.

• You’re literally at the mercy of Matchmaking so continuing to queue and play games is just a dice roll. For most people it’s not worth the continued effort unless you’re near the desired rank you want to be in.

• You’ll play the same top 25 players and/or their alts over and over. I preferred
  S2/S3’s matchmaking imo.

I still like the PvP league system and it’s what was needed in S1, but everything mentioned above darn near made this season the most frustrating one yet, next to season 1.

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

“All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.” – I’d love that. :]

+1 what reikou said.

It makes no sense that 2 pro duos can fight each other and lets say they win 50/50 because they’re both really good duos they end up eventually BOTH losing rating because the other 6 players are low rated.

No wonder people bloody q dodge man.

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

Not sure about the hard ranges, but number 2 definately should be impemented. It’s almost industry standard in most competative team games to get rewarded and penalised based on team rating.

meanwhile in LOL and DOTA2… soloQ only, need phone number attached to account to be able to play pvp. ( prevent alt /smurfing)

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

That’s not a bad idea, but it won’t solve the true problem on the long term. What may happen is :
- never ending queues at some hours ( HoH/GvG/HB in GW1 had that, the same happened in tournaments in GW2 4 years ago)
- players never really going down in leaderboard, since in average, players won’t lose many points. #250 players will be hard to catch if they
- players might use multiple accounts on dead hours even more since they will be getting 10 points instead of 1/2

To be honest, i think the system itself right now is good, they did put a lot of effort, and you have to consider it was implemented recently. However, huge matchmaking problems and fiasco is generally happening in silver-platinum where you team people that know how to play on capture points with people that perform a good build, but have literraly no clue how to play . The only thing that should be done is :
- adding a capture points tutorial
- removing OP builds and getting back to the old meta

Population is too low to come to any accurate conclusions about Anet’s MM system.

Problem is they gutted their PvP pop by castrating the rewards & killing off Pro League.

The balance issues and all the other problems with GW2 PvP can be played around while still maintaining an overall pleasant experience for the user

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

Well, the hard limit on players only in your division simply can’t work at least not with the current set up. Otherwise on NA naru would not be alowed to play with anyone else since he is the only one in legend and I doubt there are 10 people in legend on EU, let alone getting them all to q at the same time. Not to mention low tier platinum, is closer to top tier gold then top tier gold is to low tier gold so seems silly to mark a hard boundary in that regard.

100% agree with everything Saiyan said, and Reikou’s #2 point. The match manipulation is a real factor this season, and imo the prevalence is a result of discouraging players who have reached a desired rank from continuing to play at that rank any more than necessary. The logic is like this:

Step 1: Player reaches a desired top 250 spot, or hits plat, or high gold, or whatever they are aiming for.

Step 2: Player is punished more for losses than they gain for wins, thanks to points being calculated off PERSONAL MMR instead of TEAM MMR.

Step3: It is therefor undesirable for player to play any more games than are strictly necessary to avoid decay and remain on the boards (assuming they are actually after a board position. Player gets on an alt account!

Step 4: Player is already on an alt account, and:
A: Duo queues up with a buddy, but hey, player messes around more on this alt and therefor has lower MMR here, and how about that, it helps out their buddy to climb, because player is drawing a lower MMR opponent for them!

B: Player sees one of those top ranking competitors on his team, and man, they really don’t want to give that competitor a win. I’m not saying player is throwing the match per se, but their heart isn’t in it, they just don’t play that hard. Or for some people, they do just straight out throw.

Conclusion: as a result of the way personal MMR effects points gained/lost at the end of a match, I think you can see a clear incremental decision making process that would lead to behaviors that taken in whole looks an awful lot like deliberate match manipulation. And as we all know, there’s no real recourse for players to report and see anything get done about this with any consistency.

Solution: Swap to TEAM MMR for determining points gained/lost so that higher ranked players only have to maintain that 50% win rate against equal opponents to maintain rank… you know, the 50% win rate that the MMR system is attempting to generate to begin with.

Much logic.

100% agree with what Saiyan has already said.
I’ll leave my feedback here in support of this thread.

My experience in S6:

• 4 wins and 6 losses on placement, placed gold 2. I had two DC losses and two double thief comp losses. The last 2 losses were legit losses. ~ It is my opinion that it is time to implement a no class stacking function within the algorithm and get some community feedback on how to better handle DCs during ranked seasons.
• My first 200 games played were all played between 1500 and 1600 rating. I was duo que’ing with a couple of buddies that were lower rated than I was. I had roughly a 75% win rate for these first 200 games yet I could not progress past gold 3 because I was gaining +4s and +6s on wins and anywhere from -14s to -20s on losses. You can view proof of this in this thread: https://forum-en.gw2archive.eu/forum/game/pvp/5-on-wins-20-on-loss-what/first#post6547324 and no, we weren’t smurfing. They were new players that I had made acquaintance with and decided to give them a run in ranked. Though with this current rated gain/loss system, I won’t be able to play with them next season if I want to progress. ~ I understand that basing rating gains/losses on personal MMR is to discourage smurfing but it’s happening anyway as Saiyan has already explained and this system punishes legit players who are trying to run solo que, that aren’t match manipulating. This is actually encouraging match manipulation as an almost mandatory aspect of playing the leaderboards. The rating gains/losses need to be based on average party MMR/Rating for obvious reasons. What else can be done to clean up match manipulation? I don’t know, it’s a difficult task indeed but arenanet should probably start fishing for community feedback on how to diminish the strong effects that it has on the leaderboards.
• So I had a couple other higher rated buddies who wanted me to que with them. They explained to me the smurfing tactics and how to play the leaderboards as such. Honestly before this explanation, I thought only a few people were doing this kind of stuff but come to find out, it’s pretty common place nowadays. But seeing as how I have no HoT alt to help them do what they wanted to do and then in turn do for me what I was doing for them, they were uninterested in que’ing with me. They would run a match or two a night and then quickly leave to duo up with someone who was able to run the smurf alt play with them. This includes: placing your alt low, que with your buddy to help him get easy matches. Then your buddy places his alt low and ques with your main to help you get easy matches. When both alt’s ratings get too high, duo que together on the alts and throw matches to lower the rating back down again, all the while killing other player’s ratings while you do it. It’s real, it’s happening all over the leaderboards and it’s a problem. Then I had to make a decision, upgrade my gf’s vanilla account to HoT so I could smurf with everyone? or represent a legit environment and refuse to do it? I decided to represent a legit environment.
• Started looking for new friends who wanted to play legit, that were around my same rating. We had the same outcome though, win 3 and lose 1 = stay same position. God forbid we have a 50% win rate one night, which usually results in negative score with how this system works. God forbid you have a truly bad night and lose more than you win = drop an entire division. Why was this such a problem in S6 though? Why was match making and rating gain/loss so bad? The problem is too much match manipulation. Platinum players smurfing in on bronze accounts, creating lopsided as hell matches and then those same accounts get too high of rating to smurf so they start throwing matches. The biggest problem occurs when your match maker selects some guy who is supposed to be gold tier weight to support your team but when he gets into the match, he just throws instead of playing at that gold tier weight. Essentially resulting a 4v5 that is virtually impossible to carry.

The system used in S6 is the system GW2 spvp deserved during year one, back when there were enough players to make it work. I think the system for S6 is great to be honest. The only thing that funked up S6 was all of this match manipulation. I understand that we are working with a smaller player base and that puts limitations on what any match making algorithm could do. Everyone needs to understand this if they want to give feedback! I can’t think of too many suggestions for the algorithm at this point but three things are certain, in my opinion:

Rating gain/loss needs to be based on average party vs. average party, not individual
“No class stacking” needs to be implemented into ranked
We have got to find a way to discourage smurfing without punishing legit players

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

Been preaching it for 2 seasons now. The game punishes you not only with teammates that are divisions below you but also huge rank losses because it sees only your personal MMR vs average enemy MMR and doesn’t take into account that you had way lower ranked team overall. You literary just play in negative (unless you queue with top players at 5 am and farm low ranks) – if you win, you gain little because enemy team was lower ranked, if you lose you lose a lot because system doesn’t care that your team sometimes hundreds of ranks below enemy team.

Queuing solo is absolute nightmare if you care even tiny bit about game. And if you don’t care you end up quiting… like i did.

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

I will take longer queue’s any day of the week vs the system we have now. +1

we can queue in pve, who cares about q time honestly or you could even play a game with better pvp tabbed out while you wait for q lol xD (ANets defense will be queue times which they don’t want to be high)

It’s a joke I am still getting matched with people in placement and I was gold 3 I think at one point now stuck in gold 1-2 hell my ratio is like literally 1:1 win for loss every time not even trying anymore just play for lulz it is seemingly impossible to climb. I don’t want to duo. Either I want to full solo, or play full team.

ArthurDent.9538:

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

Well, the hard limit on players only in your division simply can’t work at least not with the current set up. Otherwise on NA naru would not be alowed to play with anyone else since he is the only one in legend and I doubt there are 10 people in legend on EU, let alone getting them all to q at the same time. Not to mention low tier platinum, is closer to top tier gold then top tier gold is to low tier gold so seems silly to mark a hard boundary in that regard.

You have to look at it in the longer term.

Yes in the short term some players may face forever queues (because quite frankly the game is dead.) But if the actual gameplay is fun due to matches between players at a similar skill level, there will be more people back to play PvP.

Additionally, when the higher-ranked players are matched exclusively against each other, it will give other players a chance to actually rise up to those higher tiers in rating.

Right now I’d argue that the biggest issue right now, is not balance or rewards or whatever, but the fact that the rote gameplay and matches that are produced just are not fun to play simply because of the HUGE skill difference between given players in a match.

Top players get matched with/against bads, and get frustrated with their team, and frustrated with the game.

Lower ranked players get matched with/against higher tiered players and get farmed. It just isn’t fun and players get frustrated with the game.

When enough people get turned off from the game because of above frustrations, it leads to a dead game.

Yes, you can turn this around and say people at the high end of MMR will face frustrations in actually finding a match. And I would agree, in the short term. Perhaps the restrictions on legend can be relaxed a little bit to include T3 Plat. But a Legend vs Silvers, or even golds should never happen.

I feel ya, 100%, only I’m An EU player. Went 9/10wins placement, placed gold mid tier2, climbed to 1894 peak with a #12 spot, and eversince easter I have lost more then 60matches with barely Any wins. Season ends tomorrow, I’m gold and not even top250 anymore with thé most kitten matchmaking you could think of.

Idk What they did around easter, but they kittened up good <3

I noticed same thing actually. That little patch around Easter. I started getting very frequent lopsided matches after that while solo que. Went on a cool 20-30 game lose streak and didn’t really give a kitten to climb again after that. Funny thing is, my first 200 games were solidly played between 1500 and 1600, seemed like match making was actually working for once.

Whatever happened at easter patch was too bad.

one big problem with team average rating would be duo players how would you calculate 1846rating duoquen with a 426rating? team rating way down to give higher some bonus pts?
other way just make both ratings for calculations equal of the higher rating ie 1846 and 1846 which in return would discourage players to question with anyone below them…. so nothings different from now.. ok you got my vote for team average mmr then…

They should get rid of seasons. Just put the achievements into unranked so you can grind your Legendary backpack and give back the option to farm Ascended gear too – without having to pve.

The glicko kitten never really worked. Most of the unranked games are far more balanced than 90% in season.

End this “competitive” experiment. It failed. If you want to follow this path you need an actual balancing team that balances on a daily or weekly basis not patches things every three month based on the raid situation in pve.

everything is wrong about pvp.. Matchmaking, class balance, point system, rewards everything..

It doesn’t make sense you lose 50-500 or 499-500.

GW2 PvP code looks like created by a student or intern. Just throwing dice.. GW2 never will be a good pvp game Anet hasn’t talented developers and they ruined too many things they can’t fix anymore.

Nevermind lets remove mender amulet..

Saiyan.1704:

• I realized I should have used my gf’s alt account just to prevent myself from playing more games once reaching my desired rank. I realize that now.

:) Hopefully everyone would realize this.

Reikou.7068:

1) Matchmaking had hard limits on its search ranges.

its not mmr that’s at fault here. its the fact that you can q with someone who is rank 20.

imo, for each tier, you need a minimum rank req.
r0-19 – bronze
r20-39 – silver
r40-59 – gold
r60-79 – platinum
r80^ – legendary

in addition to better quality matches this would help eliminate smurfing/q baiting.

Going off of what Reikou mentioned, there should be a more elegant solution.

Personally, and I’m aware that it’s a recurring complainy from both seasons with this system now, but I just finished my push at the minimum game count and hit the “+5 for a win, -16 for a loss” scenario for every match I played.

Yes, the population is dwindling and yes, a hard cap (which I thought was + or – 150 rating but I don’t have tje data to support that and it isn’t the point) would make outlying skill ranges miserable to queue in.

So, since rating gained or lost is a function of deviation, which is also tied to matchmaking, then the matchmaker shouldn’t be able to make matches in which the highest rated player on each team would lose more than they gain from a match, which would of course “trickle down” and everyone in the match would receive the same treatment for their match gains/losses.

Implemented correctly, and this would reduce many players frustrations with queuing at the top end of the ladder, on top of it being a direct improvement to the match quality (we’ve all had those games where we think “how could I have possibly been expected to carry THAT GUY,” and those are the players that we shouldn’t be queued with, and which makes the algorithm take more away for a loss than a win even though it’s what chose to place us together to begin with, which is a result of the algorithm making too loose or too inaccurate skill matchups).

I have to say I still do not understand rankings. I played daily, had roughly a 60/40% win/loss ratio, played since the week of season start. I made it to Silver T2

A guild mate played 1 week, had a 40/60% win/loss ratio. Less matches than me and he was Gold T3 in that time.

How is it some players can rank through with less matches than others? Its like if my team wins +10-15 points. If we lose, -17-25 points.

As for matchmaking, Bronze players getting paired into a Gold team is crazy. Or even better, a player without a Bronze rank yet paired onto a Gold team. The MMR seriously needs work.

Another point, if a person ditches during the match they should receive an automatic 1hr Dishonor for the first offense…2nd offense same week, 24hr Dishonor…3rd offense, 72hr Dishonor. If they manage to do it a 4th time, Dishonor for 1 week.

Reikou.7068:

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

I’m going to presume this is true since it’s consistent with anecdotal evidence, and if it’s true it is 100% a mistake. Like, your model is inconsistent and you fail your statistics exam kind of mistake.

It means your matchmaking is using a fallacious basis and your leaderboards don’t sort properly.

It would be good to get a confirmation to the contrary, because if the above is true GW2 isn’t using Glicko or Glicko-2 or anything of the sort, it’s using its own custom MMR that’s basically just wrong.

Ensign.2189:

Reikou.7068:

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

I’m going to presume this is true since it’s consistent with anecdotal evidence, and if it’s true it is 100% a mistake. Like, your model is inconsistent and you fail your statistics exam kind of mistake.

It means your matchmaking is using a fallacious basis and your leaderboards don’t sort properly.

It would be good to get a confirmation to the contrary, because if the above is true GW2 isn’t using Glicko or Glicko-2 or anything of the sort, it’s using its own custom MMR that’s basically just wrong.

I didn’t really understand what you were assuming to be true here. The OP said that the wins/losses should be based on your team average MMR, which is what matchmaking supposedly uses to pair groups of people, i.e.:

Team A Team B
1500 1400
1200 1500
1300 1500
1700 1600
1800 1500
———- —-——
1500 1500 AVG

However, if Team A loses to Team B, the player with 1800 rating loses more points than that player with 1200 rating on his team.

In essence, the player with the highest rating is expected to carry the lower rated player with whom his MMR averages out with based on the teams.

This is sometimes impossible, which is why the current system is overly harsh as you climb ratings. I think the OP’s suggestion that the ratings gained/lost being based on average team MMR would encourage people to actually play more.

Right now if you climb up that high, you get -30 for a loss and +5 for a win. If based on average team MMR, if you are 50/50 in games, you will be staying at the same personal MMR rather than bleeding MMR over time.

Then again, one can make the argument that if you cannot maintain an 80% win ratio, you’re not good enough to carry your teams and therefore you don’t deserve your high rating.

EDIT: as far as my experience this season, I won 7 and lost 3 placement matches and got placed into Gold T2. I remained basically between Gold T1 and Gold T3 all season, occasionally getting into the leaderboards. I got paired with and against the top players a few times which is both a blessing and a curse, since it’s either an easy win or an easy loss. I think with a more tier restrictive matchmaking algorithm, a new win/loss calculation, and a greater player base, the major problems people are complaining about this season can be solved.

Animism.9803:

Ensign.2189:

Reikou.7068:

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

I’m going to presume this is true since it’s consistent with anecdotal evidence, and if it’s true it is 100% a mistake. Like, your model is inconsistent and you fail your statistics exam kind of mistake.

It means your matchmaking is using a fallacious basis and your leaderboards don’t sort properly.

It would be good to get a confirmation to the contrary, because if the above is true GW2 isn’t using Glicko or Glicko-2 or anything of the sort, it’s using its own custom MMR that’s basically just wrong.

I didn’t really understand what you were assuming to be true here. The OP said that the wins/losses should be based on your team average MMR, which is what matchmaking supposedly uses to pair groups of people, i.e.:

Team A Team B
1500 1400
1200 1500
1300 1500
1700 1600
1800 1500
———- —-——
1500 1500 AVG

However, if Team A loses to Team B, the player with 1800 rating loses more points than that player with 1200 rating on his team.

In essence, the player with the highest rating is expected to carry the lower rated player with whom his MMR averages out with based on the teams.

This is sometimes impossible, which is why the current system is overly harsh as you climb ratings. I think the OP’s suggestion that the ratings gained/lost being based on average team MMR would encourage people to actually play more.

Right now if you climb up that high, you get -30 for a loss and +5 for a win. If based on average team MMR, if you are 50/50 in games, you will be staying at the same personal MMR rather than bleeding MMR over time.

Then again, one can make the argument that if you cannot maintain an 80% win ratio, you’re not good enough to carry your teams and therefore you don’t deserve your high rating.

EDIT: as far as my experience this season, I won 7 and lost 3 placement matches and got placed into Gold T2. I remained basically between Gold T1 and Gold T3 all season, occasionally getting into the leaderboards. I got paired with and against the top players a few times which is both a blessing and a curse, since it’s either an easy win or an easy loss. I think with a more tier restrictive matchmaking algorithm, a new win/loss calculation, and a greater player base, the major problems people are complaining about this season can be solved.

Animism got it correct here.

After thinking about it, I realize there is another alternative solution that would not require any changes to the MMR Calculation formula, AND still allow for players in the outlier MMR ranges to have somewhat fast queues.

That solution would be to basically remove any and all matchmaking what so ever, and make it a completely soloqueue random arena.

No more “even teams,” no more class-stacking, no more duo queues, nothing. Just take the first 10 players who queue up, shuffle them around, and throw them in a match. Then MMR gains/losses can be calculated off personal MMR (like our current system,) and still make any sort of sense.

Even teams break glicko if there is a big skill gap. They need to put similar rated players on the same team. They either need longer queues, more people, or reduce the team size.

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

You lose 50% of MMR based on your own and 50% based on your team’s MMR. Would that work?

Tomiyou.3790:

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

You lose 50% of MMR based on your own and 50% based on your team’s MMR. Would that work?

It would “work” in so much as the current MMR calculation system “works” I guess.

But I’m not sure why you’re looking for a 50% change, when it would be both needlessly complicated, and still provide worse results than calculating mmr gains/losses 100% based off of average team MMR.

Alt accounts and win sharing has a big impact. The ease of alt accounts and like you said, how people will just instant quit/give up against a counter team or top player(s) and it doesn’t affect them.

Despond.2174:

Alt accounts and win sharing has a big impact. The ease of alt accounts and like you said, how people will just instant quit/give up against a counter team or top player(s) and it doesn’t affect them.

Yes but people play alt accounts because the system is so punishing to players who have high MMR. There is no incentive to play once you reach the rating you want because it’s more advantageous to just camp that rating.

Reikou.7068:

Tomiyou.3790:

Reikou.7068:

All of which wouldn’t matter and wouldn’t happen if

1) Matchmaking had hard limits on its search ranges. Specifically search for players to play with/against ONLY in your division. Nothing higher, nothing lower.

OR

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

You lose 50% of MMR based on your own and 50% based on your team’s MMR. Would that work?

It would “work” in so much as the current MMR calculation system “works” I guess.

But I’m not sure why you’re looking for a 50% change, when it would be both needlessly complicated, and still provide worse results than calculating mmr gains/losses 100% based off of average team MMR.

I was thinking of these 2 problems:
“- players never really going down in leaderboard, since in average, players won’t lose many points. #250 players will be hard to catch if they”
“Then again, one can make the argument that if you cannot maintain an 80% win ratio, you’re not good enough to carry your teams and therefore you don’t deserve your high rating.”
I’m pretty sure Anet considers both of these as a potential problem and splitting loss of MMR based on team’s and a player’s own MMR 50/50 would certainly give more control over both. Numbers could be tweaked ofc (60/40….)

ANET tried putting similar ranked players on the same team. That was s2-s3 and it was miserable for many players.

I’m sure the win streakers enjoyed it.

The reality is that most people will quit a game they consistently lose. All or virtually all serious sports go to great lengths to provide a level playing field.

That hasn’t happened in PvP for the most part. Teams vs. Solos, deliberately skewed matches, exploits such as throwing and smurfing… So, we come to the situation we have now. The rewards aren’t high enough to be worth taking beatings for the pleasure of the “top” players. Hence, population goes down.

As to being able to vote a teammate dishonor. That is so blatantly open to abuse that I am reasonably confident ANET will see the problem and not implement it.

This whole discussion about bad MM is irrelevant. The MMR system is simply not a good way to do a matchmaking in a team based game. The more people you have in your group, the more prone to errors MMR is. This system is generally great in 1v1. I never had any bigger issues with matchmaking in games like Starcraft II, Street Fighter IV etc. But in games like GW2, WoW, Overwatch, LoL etc. it’s always a series of dissapointments.

I never played a single team based game with competetive mode where I would not get infuriated by the matchmaking. When you 1v1, the MMR is stricly and entirely tied to your own performance. You NEED to carry yourself or you loose. In team based games, even if you are better than everyone else on the map, you can still loose and in fact you have a pretty decent chance to loose. Your personal skill increases your chances to win somewhat but nothing more. If you play A LOT of matches, by principles of statistics you will be placed more or less where you belong. But you need to play a lot or you need some luck. In 1v1 games I was placed where I felt I belong pretty fast, after 20 games maybe. After that, as I gained or lost MMR, I felt my oponents getting better or worse. In team based games I often gain some MMR and then I get kitten on by the matchmaker and I tank. I had games in GW2 where I bunkered a point, oftentimes held it against 2 people and was generally busy but we still lost because rest of my team was running around without idea. I also had games where I was trying some new class and was playing really bad but I still won because I was carried by my decent team. This does not generally happen in 1v1 games. But will keep happening in 5v5 and similar formats and there is nothing that A.net or anyone else can do about it.

TLDR: I dont think we can blame A.net for such state of matchamaking as it is in atm. Because thats the state of every single matchamaking system in every single game based on team combat.

I think that’s a misunderstanding. PvP is a “team” game in the sense one might pick up a couple of random partners to play cards or some ball game.

It is not a “team” game, in the sense of uniforms and scrimmages.

I believe ANET’s mistake has been in paying too much attention to organized team play. GW2 is a game that wasn’t packaged or sold as a hard core PvP arena.

In GW1, I have a hero group. Yes, that’s team play. I play Monk or Mesmer and have a definite role. My heroes have specific team roles. But that’s seven AI characters and me. A full human team is far rarer. It’s often less effective, too.

Animism.9803:

However, if Team A loses to Team B, the player with 1800 rating loses more points than that player with 1200 rating on his team.

Right, that is how it seems to work. That means that A.Net’s implementation of Glicko is mis-specified. It’s simply wrong – everyone on team B’s rating should move up or downward the same amount (with allowances for differences from rating volatility).

What they are doing, with the higher rated player losing more on losses than the lower rated player on the same team, is not Glicko or Glicko-2. It’s mis-specified. It makes everything about matchmaking and ratings inconsistent (matchmaking and match prediction is based upon rating numbers representing something specific, but they don’t actually represent what they are supposed to – because they are computed wrong).

It means players are rated incorrectly, which means matchmaking isn’t building teams appropriately, creating blowouts that it doesn’t think are blowouts – garbage in, garbage out.

Updating their MMR algorithm to actually be a correct implementation of Glicko-2, or some consistent derivative, should be really high priority. This should have been fixed years ago – there’s really no excuse for a high budget studio like A.Net not to run their MMR and matchmaking code past a statistician at some point.

Ensign.2189:

Reikou.7068:

2) All MMR gained/lost from a match is calculated based off of average TEAM MMR, and not personal MMR.

I’m going to presume this is true since it’s consistent with anecdotal evidence, and if it’s true it is 100% a mistake. Like, your model is inconsistent and you fail your statistics exam kind of mistake.

It means your matchmaking is using a fallacious basis and your leaderboards don’t sort properly.

It would be good to get a confirmation to the contrary, because if the above is true GW2 isn’t using Glicko or Glicko-2 or anything of the sort, it’s using its own custom MMR that’s basically just wrong.

This is not anecdotical evidence. You should try to duoq with someone who is 100 or so ranks below you. Ask them how many ranks they gain/loss for every match with you. I promise you, it will be something like: you gain 10, they gain 20, you lose 20, they lose 15. The system currently does base rank gains/loss on personal MMR vs avg MMR of the enemy team. It does not take into account your teammates at all and given wonky MM that creates matches like plat/gold/gold/gold/gold vs legend/legend/plat/plat/plat on constant basis it becomes really punishing for no good reason.

Season 2 and 3 did have the best matchmaking or at least the matchmaking that made the most sense. Nothing enrages people more then being forced to play with people that are more then a tiny bit worse then they are. They’ll never go back to it though because of the bads that were losing like 20 or 30 games in a row under that system because there teammates were other equally bad people. Also, without pips not sure how you could do it. Seasons 2 and 3 I believe was where you were grouped with people right at your MMR level and faced whoever was + or – within a certain pip range in the queue. I guess instead of pips they could look at MMR range. I’m guessing queues would increase but I could live with that. And yes they do need a report system so players can police themselves like in LoL and Overwatch. They don’t have the guts to do that either. They would tell you it would get abused even though there is no evidence its abused in those other games. Those self policing tools are powerful incentives for people to try 100% at all times and more importantly to want to learn how to play the game correctly. You want to keep going to far point five times straight in a game and lose every time and kill your team I hope you enjoy the mini ban that is coming down the pike for you if we had those tools.

Ensign.2189:

Right, that is how it seems to work. That means that A.Net’s implementation of Glicko is mis-specified. It’s simply wrong – everyone on team B’s rating should move up or downward the same amount (with allowances for differences from rating volatility).

What they are doing, with the higher rated player losing more on losses than the lower rated player on the same team, is not Glicko or Glicko-2. It’s mis-specified. It makes everything about matchmaking and ratings inconsistent (matchmaking and match prediction is based upon rating numbers representing something specific, but they don’t actually represent what they are supposed to – because they are computed wrong).

It means players are rated incorrectly, which means matchmaking isn’t building teams appropriately, creating blowouts that it doesn’t think are blowouts – garbage in, garbage out.

Updating their MMR algorithm to actually be a correct implementation of Glicko-2, or some consistent derivative, should be really high priority. This should have been fixed years ago – there’s really no excuse for a high budget studio like A.Net not to run their MMR and matchmaking code past a statistician at some point.

It does not matter what sort of rating system you use. In 5v5 you will always have many matches which will be out-sided stomps not fair to some of the people in the loosing team. Why when I play 1v1 games after being properly rated, I don’t get such one sided matches unless I get successfuly cheesed (which is my fault because knowing how to deal with cheese is part of being skilled)? Why I happen to see one-sided stomps every day when I play GW2, LoL, Overwatch, HotS, WoW etc. etc.?

“top” players cheating the system went way deeper than that. Pretty sure since s5 people have been boosting accounts for gold or even cash. Especially seeing this season leaderboard, I’m 100% positive at least half of the people in top 25 paid to be on there. Either through gold or cash. And top 3 100% cheated/paid to get their spot. Having the “God of PvP” title just basically meant “hey I cheated”. These top players even bold enough to stream themselves boosting multiple accts openly and ANet still didn’t do a thing (if you watch twitch you know the ones I am talking about). And even if they do I’m pretty sure these players don’t care anyway. Most of them even openly said they are only playing the game cause someone paying them. Just face it, PvP is dead. With next expansion things may perk up where people come back for a while but as long as they don’t permanent IP ban booster and require verified unique phone number per account to PvP the game mode will forever be a meme.

Not saying people don’t cheat/smurf but HALF of the top 25 paid for their spots? That’s pretty funny and false lol.

I see at most maybe 5 people that could’ve cheated out of 25 and that’s a maybe.

Cynz.9437:

This is not anecdotical evidence. You should try to duoq with someone who is 100 or so ranks below you. Ask them how many ranks they gain/loss for every match with you. I promise you, it will be something like: you gain 10, they gain 20, you lose 20, they lose 15. The system currently does base rank gains/loss on personal MMR vs avg MMR of the enemy team. It does not take into account your teammates at all and given wonky MM that creates matches like plat/gold/gold/gold/gold vs legend/legend/plat/plat/plat on constant basis it becomes really punishing for no good reason.

Yep, then the rating and matchmaking system is simply wrong. Broken systems produce garbage, which it uses as future inputs in a long cycle of garbage.

A.Net should probably fix that before looking at anything more detailed. Hard to talk about improvements to Glicko when you don’t have Glicko working yet.

fishball.7204:

Not saying people don’t cheat/smurf but HALF of the top 25 paid for their spots? That’s pretty funny and false lol.

I see at most maybe 5 people that could’ve cheated out of 25 and that’s a maybe.

It really doesn’t matter how many do that. The box is open and if Anet doesn’t ban that behaviour now you will have a lot more people doing this even on lower rankings which destroys pvp even more.
I really don’t understand why competitive games always draw scum like these people out of their holes and the game designers tolerate it.

This kind of behaviour destroyed any game I played in the last 30 years.

Ensign.2189:

Cynz.9437:

This is not anecdotical evidence. You should try to duoq with someone who is 100 or so ranks below you. Ask them how many ranks they gain/loss for every match with you. I promise you, it will be something like: you gain 10, they gain 20, you lose 20, they lose 15. The system currently does base rank gains/loss on personal MMR vs avg MMR of the enemy team. It does not take into account your teammates at all and given wonky MM that creates matches like plat/gold/gold/gold/gold vs legend/legend/plat/plat/plat on constant basis it becomes really punishing for no good reason.

Yep, then the rating and matchmaking system is simply wrong. Broken systems produce garbage, which it uses as future inputs in a long cycle of garbage.

A.Net should probably fix that before looking at anything more detailed. Hard to talk about improvements to Glicko when you don’t have Glicko working yet.

The calculation should be against your rating not the teams. However, there shouldn’t be huge variations in the rating of team members. The issue is with the matchmaking not glicko. Right now we are sacrificing short queue time for poor quality matches.

Also, we really aren’t using glicko it is more ELO. This season they disabled the key difference between the two systems which is increasing rating deviation due to inactivity.

Faux Play.6104:

The calculation should be against your rating not the teams. However, there shouldn’t be huge variations in the rating of team members. The issue is with the matchmaking not glicko. Right now we are sacrificing short queue time for poor quality matches.

The matchmaker cannot possibly provide high quality matches with garbage ratings being fed into it. Ratings, and rating differences are supposed to represent likelihoods of outcomes, and when the matchmaker is paired with a consistent rating system the ratings will converge to make the predictor true.

But that doesn’t happen because the matchmaker and the rating systems are inconsistent.

For instance, the matchmaker has baked into it some logic that says that if you put high platinum player into a mid gold game (that is, a player with a +200 rating difference), that it would expect the platinum player’s team to win 57% of the time. That’s a bit lopsided, but not so unacceptable that it can’t fire the match off. Unfortunately, that assumption isn’t true – because the player rating update is inconsistent, the platinum player’s team actually has a 76% chance of winning the match.

What is the matchmaker supposed to do when it can’t accurately evaluate the quality of a match?

The matchmaker can only create fair matchups if it can accurately predict the outcomes. If the ratings being given to it don’t mean what it thinks it does, and update in a way inconsistent with those assumptions, it’s going to keep spitting out crappy matches with no correction mechanism.

Faux Play.6104:

Also, we really aren’t using glicko it is more ELO. This season they disabled the key difference between the two systems which is increasing rating deviation due to inactivity.

They aren’t using either system – again, Elo is a consistent system, while A.Net’s special snowflake rating system is not.

A.Net’s special system does not find, or try to find, fair match-ups for its top players – it instead finds weak opponents to feed to those players in blowout after blowout. That is probably not the design intent of their rating system, but it’s what theirs does.

Ensign.2189:

Faux Play.6104:

The calculation should be against your rating not the teams. However, there shouldn’t be huge variations in the rating of team members. The issue is with the matchmaking not glicko. Right now we are sacrificing short queue time for poor quality matches.

The matchmaker cannot possibly provide high quality matches with garbage ratings being fed into it. Ratings, and rating differences are supposed to represent likelihoods of outcomes, and when the matchmaker is paired with a consistent rating system the ratings will converge to make the predictor true.

But that doesn’t happen because the matchmaker and the rating systems are inconsistent.

For instance, the matchmaker has baked into it some logic that says that if you put high platinum player into a mid gold game (that is, a player with a +200 rating difference), that it would expect the platinum player’s team to win 57% of the time. That’s a bit lopsided, but not so unacceptable that it can’t fire the match off. Unfortunately, that assumption isn’t true – because the player rating update is inconsistent, the platinum player’s team actually has a 76% chance of winning the match.

What is the matchmaker supposed to do when it can’t accurately evaluate the quality of a match?

I agree, except reverse that 57% and 76% example for soloQ players. But you’re absolutely right in regards to that last sentence.

Ensign.2189:

The matchmaker can only create fair matchups if it can accurately predict the outcomes. If the ratings being given to it don’t mean what it thinks it does, and update in a way inconsistent with those assumptions, it’s going to keep spitting out crappy matches with no correction mechanism.

Emphasis for truth!

Ensign.2189:

Faux Play.6104:

Also, we really aren’t using glicko it is more ELO. This season they disabled the key difference between the two systems which is increasing rating deviation due to inactivity.

They aren’t using either system – again, Elo is a consistent system, while A.Net’s special snowflake rating system is not.

They’ve tried both systems. Backways, front ways, all kind of ways. But just placing a number on a person then “tweaking” that number depending on W/L ratios against teams Anet deem are “easy” or “hard” to win means there can be LOTS of outlying errors that’ll simply yield the opposite result.

Ensign.2189:

A.Net’s special system does not find, or try to find, fair match-ups for its top players – it instead finds weak opponents to feed to those players in blowout after blowout. That is probably not the design intent of their rating system, but it’s what theirs does.

Currently it looks for players of the same division.. but top 10 are grouped together due to “matchmaking looking around”, when it finds no more top 10, it picks the lower divisions because going higher isn’t an option. So top 10 are up against the next mmr and that’s top 25 and beyond.

Normally this would be “farming” the lower ranks but Anet added the -30 for a loss, +3 for a win. So that a single loss can really hurt the top 10. Naru + his carrying teef Amon must have beat me 8 times this season but I managed to add 2 losses under their belt, resulted in marginal gains if any.

To bad the top 10 tend to queue at 11:00pm to 3:00am, or before dawn, or any other time a duo is on for them to queue dodge, etc. Oh, except for when they’re on their alt accounts.

Thank you FalseLights for your post.

I agree with most of your points.

Saying FACTS help, but proposing solutions matters.

1- The rating system need to get better. The pip system of season 1 was a MASTER PIECE of what the rating system should be. Losing game at 100 to 500 versus low MMR players should drop your rating so much (if you have a good mmr), you would need 10 game to take that over. This would prevent intentional throwing from the get go, because it’s hard work to come again. But if the punishment is too harsh and goes on legit people… it’s become bad.

2- Rating need to get explain to the players. We can get from -20 to +20 rating, I want to know why i have that, so i can get better at the system.

3- Rating need to be positive on a match lost at 400+. This was in season 1 and prevent or help the AFK, the rage quitters or the not so competitive players to finish the game. This was a key success to the PIP system at that time.

4- Divisions and tier should matter. The best case scenario would be to have two teams with a full crew of the same division and tiers. The winner will raise his rating until he goes to other tier to never see the losers again. The loser are lowered and will have to beat lower rated team to raise again.

If that could not be obtain, never go more than +1 or -1 tier. It would be better to offer a bot system aka GW1 where you put equally a place holder bot on that game. Could be some kind of bruiser sustain build to hold home for both team, for example. I know population is low, but well… having unfair matches where I am Gold tier 3, that I have gold tier 1 and silver in my team to face Platinum full crew, is DISTASTEFUL and I prefer to play unranked for this big joke of a match-making.

5- New players should play together for longer than 20 rank.

As a veteran of spvp, I think that a baseline for good games is 3000 games. Anyone under this don’t understand all classes/meta/role and rotations. Having 12 000 games played don’t give me the ability to carry 2 new players with under 200 game played versus 3k+ players.

To enter GOLD divisions, I think most player should have played 1000 RANKED games. This would be some kind of a big step for the new pop, but can be achieve if GOLD is a goal in mind for the new spvp players. This would make impossible the cheating with new account that intentionally match-manipulation the outcome of a game for throwing. Smurfing from GOLD and + would not be possible.

GOLD is a mark of skills. Many players don’t get there. It should now be a mark that say: “To get there, you need to be a veteran player”.

But… this would not be enough. Many veteran that have multi-account already have those 1000 games. But… not the win ratio with it. If you scan the rating of most second account players, there is a strong disjunction between the win ratio. The main account is 60%+, while the second account could be 45% or less.

In that Regard, Gold Players should have 1000 games + 48% and more win ratio.

Those numbers are placeholder, ANET need to scan all account and put their own numbers into this, to prevent the cheating properly. It could be more simple to fix the account to 1 MAC ADDRESS, so to have a second account would mean to play on a second computer. This would let family members to share spvp game without being penalize.

But my biggest frustration in this matchmaking, is that Anet make me play with people who don’t know what they are doing.

And the biggest frustration of the new player, is to get farm by players that know what they do.

So, with all honesty we need a big league NHL like and a kid league like AHL one (hockey). So if you want to play with the best guy, you need to pass the first league that is the new guyz.

In 2013-2014, I did considered Ranked players from the top 1000 be in the big league. I then considered the 90%+ (and top 1000) to be in the big league. Because rating can goes down for some bad luck but will goes back up, cause you know… skills.

The big league could be the top 5000 account of spvp, the rest is the entry league. Having two league would help tons for the match quality. The new players would enjy playing with new players and farmed each other… and the veteran would enjoy having great matches without having to cope for bad new players decisions or builds.

Regards,

Ensign.2189:

Faux Play.6104:

The calculation should be against your rating not the teams. However, there shouldn’t be huge variations in the rating of team members. The issue is with the matchmaking not glicko. Right now we are sacrificing short queue time for poor quality matches.

The matchmaker cannot possibly provide high quality matches with garbage ratings being fed into it. Ratings, and rating differences are supposed to represent likelihoods of outcomes, and when the matchmaker is paired with a consistent rating system the ratings will converge to make the predictor true.

But that doesn’t happen because the matchmaker and the rating systems are inconsistent.

For instance, the matchmaker has baked into it some logic that says that if you put high platinum player into a mid gold game (that is, a player with a +200 rating difference), that it would expect the platinum player’s team to win 57% of the time. That’s a bit lopsided, but not so unacceptable that it can’t fire the match off. Unfortunately, that assumption isn’t true – because the player rating update is inconsistent, the platinum player’s team actually has a 76% chance of winning the match.

What is the matchmaker supposed to do when it can’t accurately evaluate the quality of a match?

The matchmaker can only create fair matchups if it can accurately predict the outcomes. If the ratings being given to it don’t mean what it thinks it does, and update in a way inconsistent with those assumptions, it’s going to keep spitting out crappy matches with no correction mechanism.

Faux Play.6104:

Also, we really aren’t using glicko it is more ELO. This season they disabled the key difference between the two systems which is increasing rating deviation due to inactivity.

They aren’t using either system – again, Elo is a consistent system, while A.Net’s special snowflake rating system is not.

A.Net’s special system does not find, or try to find, fair match-ups for its top players – it instead finds weak opponents to feed to those players in blowout after blowout. That is probably not the design intent of their rating system, but it’s what theirs does.

You are confusing the match maker with the rating system. They are different.

• The rating system does its work after the match is over. It assigns a rating and a deviation to the players. At the end of a match, it adjusts the rating and deviation based on whether the team won or lost. Anet didn’t make up their own system. It is based on glicko 2 and it is consistent.
• The match maker does its work before the match starts. Based on a set of criteria, it assigns players to teams. Its goal is to put players of similar skill levels on a team, and assign them opponents. However, it has to balance team consistency with queue times.

No match making system can provide good quality matches and short queue times when populations are low.

Currently the queue of rosters considered for a match can have as big of a gap of 1200 rating. In addition, it looks at the “effective” rating, and ignores the average rating. So when 4 top tier players are matched with a low gold/silver, their team will have a high rating deviation, and could end up matched with a team of all mid tier gold. (assuming N/A). The following is from the wiki relating to the ranked matchmaker:

   <Rating start="3m" end="10m" max="1200" min="25">

Filter/Rating/@padding
Padding is added every second you wait in the queue after Filter/Rating/@start has passed. This is an outlier fail-safe to ensure everyone gets a match.

   <Rank distance="0">
   <Rating distance="-10">

Scoring/Rank/@distance
Score added or removed based on the distance between the potential roster’s average rank and the average rank of all selected rosters, including both teams.

Scoring/Rating/@distance
Score added or removed based on the distance between the potential roster’s average effective rating (i.e. rating – deviation) and the average effective rating of all selected rosters, including both teams.

Jourdelune.7456:

….over 5000 char

1&3) You are leaving out the part where you can lose points for winning. One can also make the argument that in a close loss any one player could have made a difference but didn’t. In a blowout, you can’t say which player or players were responsible.

2) The wiki has a good breakdown of the system. It is based on glicko2. http://www.glicko.net/glicko/glicko2.pdf
https://wiki.guildwars2.com/wiki/PvP_Matchmaking_Algorithm
TLDR: you get same points for a loss as a win if the team you are playing is at the same rating. If you are rated higher than the team, a loss will cost you more points than you get for a win.

4) Would increase queue times, and people would have to be able to fall through tiers.

5) I think having divisions you can’t break through until you play a whole season is a good idea. For example, at the end of the season, the top 25% in bronze qualify for silver/gold, the bottom 25% in silver/gold get demoted to bronze. Same for the gap between silver/gold and plat/legend. But wow 1k matches? A good player will be able to get to gold level in well under that. Making them play against low tier people for that long is more likely to cause them to get bad habits.

Jourdelune.7456:

But my biggest frustration in this matchmaking, is that Anet make me play with people who don’t know what they are doing.

And the biggest frustration of the new player, is to get farm by players that know what they do.

Usually, opponents do not know what they are doing neither but got solid builds and that makes the difference.

Anyway, this situation is avoided by getting rid of placement games and seasons. Right now, the system is close to perfect, but we’re having 10 decisive games sending anyone at any rank. If we swapped players at 1200 rating and players at 1500 rating, i am not sure everyone would end up at his previous rank. Let’s make everyone start at 1000 rating and climb the way through months.