Guild Wars 2

Asked everyone in /m chat to read their last 10 matches of gains and losses this ranked season. I received replies such as: +13, -12, +12, -13 or +10, +9, -10, -9 and otherwise balanced gains and loss evidence. It would seem the algorithm is working for most players but apparently I am meant to win 5 for every 1 loss if I want to progress. What the hell is going on here? Anyone else experiencing this garbage?

what’s your mmr?

You’re probably the top player or one of the top players on the teams you’re getting.

Unless things have changed, the problem is that it bases your loss or gain on your MMR versus the average MMR of the opposing team, not on your team’s average MMR versus the opposing team’s average MMR. This means that, if you’re consistently matched against weaker players, you stand to lose a larger amount of MMR if you lose while only having a small gain if you win… even if a comparison of the average MMR would reveal that the odds were stacked against you.

Meanwhile, the weaker players on your team are getting big gains if you win and small losses if you lose (because their personal MMR is low compared to the average MMR on the opposing team).

I have the same, currently rank 50 something on leaderboards.

Enjoy the side effect of wonky MM – you face low ranks for the sake of faster queues, you also might have lower ranks on your team. The rating gain/loss system however only takes MMR of enemy into account, it doesn’t acknowledge that you might be carrying silver players vs golds. Has been like this last season as well (my team is gold, enemy team is top 10, i loss crap load of rating because my MMR was somewhere close to top 10, the system didn’t take into account that i had teammates almost 2 divisions below the enemy). Made thread about it last season as well – per usual, fell on deaf ears.

Loop.8106:

I have the same, currently rank 50 something on leaderboards.

That’s pretty much normal then. I remember the same happening in hero battles in Guild Wars 1 : when i was #20 or so, i would wait quite long in order to get an opponent, and most of the time i would either get +2 points for a win/ – 15 points for a lose. The same happened in GvG, and i hardly recall many people complaining at the system.

Well I guess I won’t be reaching out of gold 3 then. At this rate it requires 4 wins to 1 loss to stay in the same position so it requires 5 wins to 1 loss to progress. This means that I must carry an 80% win rate to be able to “slowly” progress at +5 rating for every 6 games played. Yes, it means that to go from 1,550 gold 3 to 1,650 platinum 1, it would take me 120 games with an 80% win rate while playing those games. Other players I know gain strong division progression with 5 wins to 1 loss and an 80% win rate.

This is a stronger numbers related view of MMR Hell than any of the previous seasons were able to provide. May have to follow this for awhile.

I mean seriously, if my MMR is that much higher than my teammates or opponents, to be able to provide these negligent gains and detrimental losses, then why am I not being placed against better opponents? Doesn’t really make sense…

Trevor Boyer.6524:

Well I guess I won’t be reaching out of gold 3 then.

I was playing up at the top of Gold 3 (before a loss streak put down in gold 2) and was getting between 10 and 15 for a win and -10 to -15 for a loss. However I was also playing gold 3 and plat players.

It all depends on who you are playing against.

If you are playing a team that is at your level you will get the same amount of points for a win as a loss. if you are only getting 5 points for a win and 20 for a loss you are playing a team well below your rating and should easily be able to win. That is why you have to win 4 out of 5 matches just to stay even. If you can’t the math says you are rated too high.

Faux Play.6104:

If you are playing a team that is at your level you will get the same amount of points for a win as a loss. if you are only getting 5 points for a win and 20 for a loss you are playing a team well below your rating and should easily be able to win. That is why you have to win 4 out of 5 matches just to stay even. If you can’t the math says you are rated too high.

The problem is that the algorithm doesn’t take into account your teammates.

Whether your rating is higher than the other team’s average rating is actually pretty much irrelevant to how easy or hard the game is for you to win. The matchmaker tries to form even teams, so if you’re higher rated than the enemy team, you’re almost certainly having to make up for weaker teammates. To take a less extreme example, consider if you have 1500 rating, your allies all have 500, and your enemies all have 1000. This is not in any way a fair match – you’re going to have to carry hard to have any chance of winning – but if you lose, you’re going to get penalised hard, and if you win, you’re going to get virtually nothing.

From a less extreme perspective, assuming that the game did give fair matchups (the same average MMR for each team) every time, it actually creates the perverse result where the hardest games for you are the ones that penalise you the most if you lose and reward you the least if you win. If you’re the lowest MMR on your own team, then it doesn’t matter if you’re not playing at your best since your team-mates have a good chance at carrying you anyway – but if you win you get a big benefit and if you lose you don’t lose much. Conversely, if you’re the top of your team you have to be at peak performance because you’re the one who needs to do the carrying, and then you get virtually nothing if you win and penalised hard if you lose.

The OP’s results are probably indicative that they should be at a higher MMR, but they’re just having the bad luck of consistently being put in teams where they’re expecting to carry… and five times out of seven, they’ve successfully done so (possibly five times out of six – the 13 point loss may have been one where the OP was around the team average rather than being expected to carry). I’d be curious to see if the same names are popping up, in fact: those last six games are all in one day, if it was a non-peak time it may have been the same group of players, and if the OP happened to be the highest-ranked player queuing at the time…

wait is this at Eu or NA? your at gold 3 and you only gain that much per win? something not right. you playing odd hours or something? because you are being force to fight lower MMr players. like way LOW. so if your team screwed up and you lose you get the big hit (SS of T3 gold mmr gains)

draxynnic.3719:

Faux Play.6104:

If you are playing a team that is at your level you will get the same amount of points for a win as a loss. if you are only getting 5 points for a win and 20 for a loss you are playing a team well below your rating and should easily be able to win. That is why you have to win 4 out of 5 matches just to stay even. If you can’t the math says you are rated too high.

The problem is that the algorithm doesn’t take into account your teammates.

Whether your rating is higher than the other team’s average rating is actually pretty much irrelevant to how easy or hard the game is for you to win. The matchmaker tries to form even teams, so if you’re higher rated than the enemy team, you’re almost certainly having to make up for weaker teammates. To take a less extreme example, consider if you have 1500 rating, your allies all have 500, and your enemies all have 1000. This is not in any way a fair match – you’re going to have to carry hard to have any chance of winning – but if you lose, you’re going to get penalised hard, and if you win, you’re going to get virtually nothing.

From a less extreme perspective, assuming that the game did give fair matchups (the same average MMR for each team) every time, it actually creates the perverse result where the hardest games for you are the ones that penalise you the most if you lose and reward you the least if you win. If you’re the lowest MMR on your own team, then it doesn’t matter if you’re not playing at your best since your team-mates have a good chance at carrying you anyway – but if you win you get a big benefit and if you lose you don’t lose much. Conversely, if you’re the top of your team you have to be at peak performance because you’re the one who needs to do the carrying, and then you get virtually nothing if you win and penalised hard if you lose.

The OP’s results are probably indicative that they should be at a higher MMR, but they’re just having the bad luck of consistently being put in teams where they’re expecting to carry… and five times out of seven, they’ve successfully done so (possibly five times out of six – the 13 point loss may have been one where the OP was around the team average rather than being expected to carry). I’d be curious to see if the same names are popping up, in fact: those last six games are all in one day, if it was a non-peak time it may have been the same group of players, and if the OP happened to be the highest-ranked player queuing at the time…

No. The matchmaker puts similar players on the same team. on average your team’s rating is going to be higher than your opponents if your ratios are 5 to 20. the only way you would have to carry is if you duo queued with someone that had a much lower glicko rating.

Lord Hammer Hand.4815:

wait is this at Eu or NA? your at gold 3 and you only gain that much per win? something not right. you playing odd hours or something? because you are being force to fight lower MMr players. like way LOW. so if your team screwed up and you lose you get the big hit (SS of T3 gold mmr gains)

I’m currently t2 gold in NA. normal matches are 12 pts. I won as much as 18 and as little as 7. the outliers are around midnight pacific time when the population is low.

Faux Play.6104:

No. The matchmaker puts similar players on the same team. on average your team’s rating is going to be higher than your opponents if your ratios are 5 to 20. the only way you would have to carry is if you duo queued with someone that had a much lower glicko rating.

I used an extreme example for illustrative purposes, but it’s perfectly reasonable to get something like 1600 1500 1500 1500 1400 against a like team (people have reported greater variances, such as top 250 player matched with or against high gold players). The 1600 players, in this case, stand to lose more and gain less because the game compares their rating of 1600 against the enemy average of 1500. Meanwhile, the 1400 players stand to gain more and lose less, even though there’s a good chance they’re being carried.

This gets particularly iniquitous when you get a 1600 1500 1500 1450 1400 team against the team above – also well within the error margin of what the matchmaker can throw up, particularly at lower population times. In this case, one team is weaker than the other, but the 1600 player still gets heavily penalised if they lose, because the algorithm doesn’t care that they were in the weaker team.

Note that this thread went up during an off peak period, so it’s entirely likely that the OP’s game were played off peak.

What the system should do is compare your team’s average rating against the average rating of the enemy team, and adjust based on that. So if you have one team with an average rating of 1520 against another team with an average rating of 1480, say, everybody goes up and down based on the difference in the average ratings, regardless of whether their own rating was higher than, lower than, or equal to the average.

draxynnic.3719:

Faux Play.6104:

No. The matchmaker puts similar players on the same team. on average your team’s rating is going to be higher than your opponents if your ratios are 5 to 20. the only way you would have to carry is if you duo queued with someone that had a much lower glicko rating.

I used an extreme example for illustrative purposes, but it’s perfectly reasonable to get something like 1600 1500 1500 1500 1400 against a like team (people have reported greater variances, such as top 250 player matched with or against high gold players). The 1600 players, in this case, stand to lose more and gain less because the game compares their rating of 1600 against the enemy average of 1500. Meanwhile, the 1400 players stand to gain more and lose less, even though there’s a good chance they’re being carried.

This gets particularly iniquitous when you get a 1600 1500 1500 1450 1400 team against the team above – also well within the error margin of what the matchmaker can throw up, particularly at lower population times. In this case, one team is weaker than the other, but the 1600 player still gets heavily penalised if they lose, because the algorithm doesn’t care that they were in the weaker team.

Note that this thread went up during an off peak period, so it’s entirely likely that the OP’s game were played off peak.

What the system should do is compare your team’s average rating against the average rating of the enemy team, and adjust based on that. So if you have one team with an average rating of 1520 against another team with an average rating of 1480, say, everybody goes up and down based on the difference in the average ratings, regardless of whether their own rating was higher than, lower than, or equal to the average.

I attached a table from a previous post: https://forum-en.gw2archive.eu/forum/game/pvp/Suggestion-for-ranking-next-season/first#post6502151

What I did was simulated a large number of games between a player starting at 1850 rating and always playing another player with an 1850 rating. I varied the win ratio and looked at the effect on point differential and the players deviation. In order to get the numbers the OP showed, it means they are playing a team with an average glicko score more than 200 gliko points below them. Unless they at queuing with a bronze player, they should not have to “carry” their team. The more likely scenario is they are playing at low population times, and/or queue dodging players at and above their rating.

Considering the system tended to place people low this season and make them climb, it is very likely that there are still people that are several hundred points below where they belong and are still working their way up. I’ve had a couple 18+ point wins in low gold this season, as well as some that are around 7.

My advice for the OP is to watch people that are at and above their rating and queue when they queue if they want to gain more points from their wins.

Everything draxynnic said is spot on. Agree completely.

Also faux, few things to clear up questions pertaining to this:

• I don’t que dodge
• I do play at off hours from time to time but usually play around 8:00pm eastern to like midnight eastern. Sometimes wake up and can’t sleep so hop on for a few games around 3:00am eastern until I’m tired again. Sometimes play randomly through day.
• I que with lots of different people anywhere between silver/platinum and none of us have ever had this problem before. At the most someone may complain about +11 gains vs. -14 loss or something like that. But these numbers shown lately in my play are extreme differences in gains and losses. I can assure you that I do not possess an actual MMR similar to a pro player and that I am not que’ing with bronze 1 players.
• I can say that none of my matches feel balanced lately, like within the past week when this started happening. Enemies are easy to mulch, 1v1s, 1v2s and sometimes even 1v3s. I often have matches where I never die despite being left mid or far after a team wipe and finish cleaning up 2-3 enemies lingering. I find myself needing to /t explain to players what they should be doing more often than I ever have in the past. No I’m not trying to make myself out to be a carry hero. Just saying that the matches definitely feel like I am the senior player present in the game between both teams. This is like this lately no matter what time of day I que.
• I never see other older players that I know in matches anymore. It just never happens. I can see that they are que’ing in my contacts but I never meet them in matches with or against them. These guys are all slightly under me or above me in rating but I never meet them in matches despite the diminished community? Something is wrong with that. The last player I recognized in a match was the other day when I had to fight Wakkey and that was the -13 points you saw in the bottom of the screen shot.

Most importantly, remember you’re examining a guy who is playing between gold 2 and gold 3 right now, hovering around 1500 – 1550 who can’t progress due to this problem. Most of the players I duo with or slightly bellow my rating or slightly above. The lowest rated player I duo with is ranked 1200 something, which isn’t that large of a numeric margin and is it even possible for that duo to produce numbers like +4 on win but -19 on loss? That’s crazy extreme. The algorithm would need completely deny me access to play with or against other players near my rating for numbers like this to occur. It would actually have to go out of it’s way, to make sure I am only with or against well bellow 900 rating, all the time, to be able to produce extreme numbers like +4 and -19. Something is wrong with that. Draxynnic perfectly explains why match making like that is perverse, in his first post.

Well, if I can get my game dvr record to work “if windows has patched it back yet”, I’ll record some of these strange matches I am getting and post them.

I’m not sure how your table is relevant, Faux Play?

Here’s the problem. Let’s say you have two players, Alice and Bob, both rated at 1850. Let’s also say that 1850 is an accurate reflection of their actual skill. Let’s also assume that, somehow, every player that Alice and Bob fight alongside and against have an MMR that accurately reflects their actual skill, and the matchmaker is able to form teams that genuinely have an equal chance of going either way.

Through some quirk of the matchmaking, however, Alice keeps getting put as the highest-ranking member of teams that have an average MMR lower than her own. She plays at her best and does her best to carry (and has all the stresses of needing to carry), but the matchmaker has taken this into account, so her winrate is about 50%. However, because her MMR is higher than the average of the teams she’s fighting against, she loses more on the losses and gains less on the wins (the exact amount she gains and loses is irrelevant), so she’s going to slide steadily downwards. As she does so, her MMR is increasingly going to be mismatched against her actual skill, so her winrate will slide up (the MM thinks it’s making evenly balanced games, but it’s not because it’s underestimating Alice), until eventually she reaches an equilibrium point where she’s still losing more on a loss than she gains on a win, but her winrate is high enough to make up for it. Alice knows full well that she’s mismatched, but as long as she keeps getting those matches, she’s going to be stuck. It’s not fun for her, and it’s not fun for the genuinely lower-ranked players who are stuck playing against her.

Meanwhile, Bob has the opposite experience. He’s regularly the weakest member of the team, and gains more from a win than he loses on a loss. As his MMR is increasingly mismatched against his actual skill, Bob starts being a burden on his team more and more, until eventually he reaches a similar equilibrium to Alice: he’s losing more games than he wins, but he gains more on a win than he loses on a loss. In the meantime, he’s probably not having much fun being curbstomped by stronger players all the time, and he’s a curse for those he’s teamed with.

Both are equally skilled players, but one ends up significantly higher than the other simply because of the way the algorithm works. Theoretically, you could say that someone should have an equal chance of being the highest or the lowest… but this might not be the case. If you’re a relatively high-ranking player during a low population time, for instance, you’re likely to be the strongest player in the match on a regular basis… in which case, you’d better be good at carrying or you’re going to end up like Alice.

Furthermore, the effects of random chance have a stronger effect the fewer games played. For somebody who’s played 30 games, for instance, it’s probably not ridiculously improbable for them to have had 20 games where they’re higher-ranked than the team average and 10 where they’re the lower – putting them, broadly speaking, in Alice’s position. Or, alternatively, the reverse might happen. Over the rest of the season, this might even out… but with volatility, it will take more games to get back to where you should be than it took for that initial run to put you in the wrong place to begin with.

Now, let’s say the system instead based MMR gains or losses based simply on comparing the team average MMR. If the matchmaker is genuinely throwing 50/50 games at you, then you’ll oscillate around your proper MMR, regardless of whether you happen to be the stronger or weaker player on the teams you’re getting. If you get a 55/45 game, then there’s a smaller change if the 55 team wins, and a larger change if the 45 team wins to represent the upset. Doesn’t this sound like a fairer system overall?

Sooooo. The amount of points you lose vs how many points you gain for winning, even on win streak, is not good.

Check it out, played 9 more games today. The bottom match entry in this .jpg of 3/22/2017 at 10m 50s and 448r to 500b is the same match as in the previous .jpg that was the most recent match. In other words, I’ve skipped no records. This .jpg picks up where we left off.

On a side note, I did want to link a reference to an old thread of mine that dealt with the same exact problem that we are seeing here but it was in older format from previous seasons/eras. -> https://forum-en.gw2archive.eu/forum/game/pvp/High-MMR-is-punished-for-solo-que/first This thread is still very relevant towards the problem being discussed.

Got my win10 game dvr working so I recorded a match earlier. This match ended around 4:45pm eastern time zone on 3/23/2017. I’ve had better matches as well as worse matches but I chose this video for two reasons:

• First – This is a very typical match for me. My standard match outcomes usually look very similar to this as well as my play.
• Second – Due to flow of enemies, I ended up staying far the entire match. This is actually good for the example this video was meant to display, concerning the discussion in this thread. Due to my positioning, it makes it very easy to compare my progress in 1v2s vs. the progress of my team in 4v3s and sometimes 4v2s. Very easy to track this match, in other words.

I apologize for cheesy music. Had to cover up TS discussion in background.

~ Feel free to comment on the video. I’d like to hear what others have to say.

Trevor Boyer.6524:

Got my win10 game dvr working so I recorded a match earlier. This match ended around 4:45pm eastern time zone on 3/23/2017. I’ve had better matches as well as worse matches but I chose this video for two reasons:

• First – This is a very typical match for me. My standard match outcomes usually look very similar to this as well as my play.
• Second – Due to flow of enemies, I ended up staying far the entire match. This is actually good for the example this video was meant to display, concerning the discussion in this thread. Due to my positioning, it makes it very easy to compare my progress in 1v2s vs. the progress of my team in 4v3s and sometimes 4v2s. Very easy to track this match, in other words.

http://www.youtube.com/watch?v=6DGRFVbvCnk

I apologize for cheesy music. Had to cover up TS discussion in background.

~ Feel free to comment on the video. I’d like to hear what others have to say.

dunno why you getting match against lower mmr players than yours but i just wanna point out since i see you play symbolic DH, swap your off-hands. sword/focuss then scepter/shield (so you can push ele off point by shield 5 then shield 3 for easy decap) also focus 5 does explode so if youre at close range with your enemy they get damage and its a blast finisher. also avoid clicking your skills get some nice mmo mouse w/ lots of button and bind them.

You are looking at yourself only and forgetting about the big picture.

Engineer and Elementalist agaisnt 2 or 3 necros is very risky play. Necro hard counters engi. Also, with that amount of necros, counters the ele as well.

You would have had better outcome if left the engi at far fighting against the DH’s and you at mid fighting off those necros. I’ve noticed times that your ele had to defend home solo, if he was a support build, it ends up hurting your team.

Trevor Boyer.6524:

Most importantly, remember you’re examining a guy who is playing between gold 2 and gold 3 right now, hovering around 1500 – 1550 who can’t progress due to this problem. Most of the players I duo with or slightly bellow my rating or slightly above. The lowest rated player I duo with is ranked 1200 something, which isn’t that large of a numeric margin and is it even possible for that duo to produce numbers like +4 on win but -19 on loss? That’s crazy extreme.

That is exactly your problem. Queue solo and you will be getting matches with people that are closer to your skill level, or queue with someone that is rated higher than yourself. The math for glicko 2 will require you to win 80% of your matches against people that are ~200 rating points below yourself. On top of that you are making it hard for the match maker to put together a quality match by queuing with someone that doesn’t have a similar rating to yourself.

draxynnic.3719:

I’m not sure how your table is relevant, Faux Play?

I created it using the glicko 2 algorithm. It shows what win ratio someone needs to maintain over their opponents in order to maintain a rating. If you are playing people at your level, it is 50/50. The win ratio the OP was showing suggested they are playing teams with an average rating 200+ rating points below them.

Assuming the OP is in NA their rating of 1500 puts them close to the top 250. That will make it harder for the match maker to find opponents that are at the player’s skill level. Duo queuing with people below your rating compound the problem.

Faux Play.6104:

draxynnic.3719:

I’m not sure how your table is relevant, Faux Play?

I created it using the glicko 2 algorithm. It shows what win ratio someone needs to maintain over their opponents in order to maintain a rating. If you are playing people at your level, it is 50/50. The win ratio the OP was showing suggested they are playing teams with an average rating 200+ rating points below them.

This is exactly the problem, however. A 50/50 winrate only suffices to keep you at your rating if you happen to be close to the average of the team. If you’re at the top, you need… well, look at the OP. With the numbers they’re getting, they need to win two or three games for every loss to maintain their position.

The problem being… the MM is trying to make balanced matches. So someone who is at the correct MMR should be getting a 50/50 win rate. If someone’s MMR is not reflective of their actual skill, this should reflect in their winrate until they settle to their correct point. However, with the system as it currently is, someone can rise to higher than they deserve if they have more matches where they’re the weaker player, or fall further than they deserve if they have more matched where they’re the stronger player.

Like I said, the exact numbers don’t matter. The problem is a qualitative one – that the system makes MMR adjustments based on the player’s individual MMR versus the enemy team’s average. This can mean that players can get a big reward for winning an easy game – if they got carried by stronger players and their team had a higher average MMR than the enemy team, as long as their own MMR was lower – or a harsh penalty for losing a very hard game where they were being expected to carry a team of no-hopers, where their team’s average MMR was lower than that of the enemy team.

Instead, it should give gains/losses based on your team’s average against the enemy team’s average, since this is a better reflection of when you won despite a handicap or lost what should have been an easy win. (However hard the MM tries, it’s just not going to get perfect 50/50 matches every time.)

draxynnic.3719:

Faux Play.6104:

draxynnic.3719:

I’m not sure how your table is relevant, Faux Play?

I created it using the glicko 2 algorithm. It shows what win ratio someone needs to maintain over their opponents in order to maintain a rating. If you are playing people at your level, it is 50/50. The win ratio the OP was showing suggested they are playing teams with an average rating 200+ rating points below them.

This is exactly the problem, however. A 50/50 winrate only suffices to keep you at your rating if you happen to be close to the average of the team. If you’re at the top, you need… well, look at the OP. With the numbers they’re getting, they need to win two or three games for every loss to maintain their position.

The problem being… the MM is trying to make balanced matches. So someone who is at the correct MMR should be getting a 50/50 win rate. If someone’s MMR is not reflective of their actual skill, this should reflect in their winrate until they settle to their correct point. However, with the system as it currently is, someone can rise to higher than they deserve if they have more matches where they’re the weaker player, or fall further than they deserve if they have more matched where they’re the stronger player.

Like I said, the exact numbers don’t matter. The problem is a qualitative one – that the system makes MMR adjustments based on the player’s individual MMR versus the enemy team’s average. This can mean that players can get a big reward for winning an easy game – if they got carried by stronger players and their team had a higher average MMR than the enemy team, as long as their own MMR was lower – or a harsh penalty for losing a very hard game where they were being expected to carry a team of no-hopers, where their team’s average MMR was lower than that of the enemy team.

Instead, it should give gains/losses based on your team’s average against the enemy team’s average, since this is a better reflection of when you won despite a handicap or lost what should have been an easy win. (However hard the MM tries, it’s just not going to get perfect 50/50 matches every time.)

Then people would just duo with a smurf account to knock their team MMR down so they could get easier games. The system doesn’t prevent you from doing that , but you are discouraged from doing it because you gain very little rating and you risk taking a big hit if you lose.

Oh there is definitely smurfing going on. I even did it myself to help a buddy this season. He had two accounts, one a main of course and then one alt in which he barely ever plays and I signed on that fresh alt.

Every match was easy until I brought that alt account to around gold 3. Things started evening out at that point. But moral of the story is that we went on a smurf adventure 20+ games deep in to a win streak for him. Jumped him up an entire division from it.

Notice each season there are these mysterious weird accounts that place in top 50 that no one knows and has never seen before. I’m sure its pro and semi-pro players taking turns on alts. How else you gonna 40 and 1 in a ranked season like this? I mean really.
Even look at actual ESL history or WTS history, No one pro team is so dominant that it has a 95% win rate against other pro teams. No, if the algorithm was working the ideal way, these guys should be losing some games against other pro players from time to time man. It’s almost as if they don’t run in to each other somehow. After what I’ve experienced with my match making, I would believe they rarely play against each other and that’s just all messed up ^^.

Trevor Boyer.6524:

No, if the algorithm was working the ideal way, these guys should be losing some games against other pro players from time to time man. It’s almost as if they don’t run in to each other somehow. After what I’ve experienced with my match making, I would believe they rarely play against each other and that’s just all messed up ^^.

Right now people can queue dodge so that increases their odds they play people rated lower than themselves. However, that increases their glicko volatility. Which means if they do lose they will lose more points. I made a post this off season showing how your rank should be based on the lower limit of your 95% confidence interval. That will encourage people to take matches that won’t increase their volatility, and it will prevent rank sniping because the only way you can get rank fast is to have a high volatility.

https://forum-en.gw2archive.eu/forum/game/pvp/Suggestion-for-ranking-next-season/first#post6502151

Faux Play.6104:

Then people would just duo with a smurf account to knock their team MMR down so they could get easier games. The system doesn’t prevent you from doing that , but you are discouraged from doing it because you gain very little rating and you risk taking a big hit if you lose.

There are other ways to discourage that.

For instance, you could make it so that if the gap in MMR between two players who are duo-ing is too large, the game is matched as if the lower player had the same MMR as (or one only slightly lower than) the higher player. Duoqueue smurfing of the kind you talk about then becomes impossible, and you don’t have the problems I’ve raised in this thread with the current system.

I think it would fix a lot if the algorithm would consider that no single player should encompass more than 1/3rd of his party’s average MMR weight, for the reason of how conquest is actually played. I’ll explain:

• Take a split of: 8, 2, 2, 2, 2 = 16 vs. 4, 4, 3, 3, 2 = 16. These numbers look good and even on paper but are bad and imbalanced in actual conquest application. Sure the 8 will be able to choose a node and defend it. He’ll rotate to help others and he may win that node as well but the node he was previously at will be back-capped because his 2s cannot win against the 4, 4, 3, 3, 2. In other words, the 8 can only win one node at a time while his 2s are being mulched by the other team. Even though the average team MMRs look balanced, the 4, 4, 3, 3, 2 has a far greater probability of being able to hold 2 nodes consistently through the match for the win. The problem here is how MMR weight is distributed on the three nodes during conquest play. The 4, 4, 3, 3, 2 is able to more evenly distribute it’s weight amongst three node play where as the 8 is worth 50% or 1/2 of his party’s MMR alone but since he can only be in one place at a given time, his team has a high likelihood of winning where the 8 is at but also a high likelihood of losing the other two nodes where he is not at. Too much of the party’s MMR weight is on the 8’s head, creating an imbalanced match up for how conquest three node hold is played.
• Now let’s look at a split like: 6, 3, 3, 3, 3 = 18 vs. 4, 4, 4, 4, 4 = 20. Even though the team of 4s has a higher average party MMR, this is actually a more balanced match for how conquest is played than the first example. The 3s are close enough in skill to the 4s where they should be able to survive long enough for the 6 to rotate and assist them, despite the 6’s team having the lower party MMR. Again, this is more balanced because of how the MMR weight is distributed amongst the three nodes. In this case, the 6 is only worth 1/3rd of his party’s MMR weight. So again, despite total 18 vs. total 20, this is more about the distribution of a party’s weight amongst three nodes more so than it is having perfect average MMR party totals.
• An extreme example would be 10, 1, 1, 1, 1 = 14 vs. 3, 3, 3, 3, 2 = 14. The 10 is worth 2/3rds of his team’s MMR weight. In this match he’ll be expected to engage three of the 3s to be able to balance out the weight amongst three nodes. So this guy has to place himself in situations where he is 1v3ing the entire match and shouting at his 1s to leave him on point alone and go fight/cap around the remaining 3 and 2. G’luck with all those 1v3s for one thing and g’luck to your 1s as they fumble learning the game against the enemy 3s and 2. Winning for the 10 is improbable here. And of course there is that problem where he will be severely punished rating wise for losing this match, yet rewarded little for winning against impossible odds.

So we all know “the ideal match” looks like: 5, 5, 5, 5, 5 vs. 5, 5, 5, 5, 5 and that sometimes 10, 1, 1, 1, 1 vs. 10, 1, 1, 1, 1 does happen. But for those times when the 10, 2, 2, 2, 2 is vs. 4, 4, 4, 4, 4, there needs to be a limit to how much weight one man is expected to pull and that limit should be roughly around 1/3rd of his party’s total MMR weight.

Match making tonight after 3/29/2017 patch seemed very different. Points were pretty evened out for wins/loses but boy oh boy did the match making feel volatile tonight. Not sure if anything was changed for match making? But it felt like placement matches all night tonight. Either blow out win for me or blow out win for other team. Was like this in ranked all night tonight.

But yeah, w/e. Rating gains/loses seemed better.
Even when que’ing with people well above and bellow me.

all your math are viased, using a 1-10 scale for examples in the pvp scenario should mean10 as top 25 or 50 players and 1 as deep deep bronze ones, and gives a false perspective of abismal diference betwen players in a match
mmr scale is a thing betwen a theoric 0 and 2200-2300 or so and reported in forum xtreme uneven matches are of about 400 mmr diference, in a 1-10 proportional scaling 2 points

Check this out though. +7 on wins and -15 or more, for dealing with matches like this all day. is anyone else experiencing this crap? Been dealing with this all season and as you can see from my previous posts, this is a reoccurring pattern. Why are my matches always like this? For a person who hovers in the gold 3 tier, you’d think I’d get matches where I am the one being carried every once in awhile.

Probably a population issue. Matchmaking has been terribad since SAB returned.

Side effect of D E D G A E M

Gold t3 is like pretty high on NA, that’s like top 100 so probably rare to find people in game better than you

draxynnic.3719:

Probably a population issue. Matchmaking has been terribad since SAB returned.

I am sorry to say, but since SAB matchmaking is much more balance for me.

I don’t have to carry 4 guyz with less than 1000 games anymore. I have more Legend something in my teams and on the other side.

SAB bring back the bored PVE guyz that don’t play META in RANK and can’t rotate properly.