Guild Wars 2

So with each win im getting -1 rating, down to 19 rating for the 4th or 5th win in a row. How is this possible?
Also i have a 18/6 ratio nwith a buddy and we’re climbing like turtles because of that while i see people in top 250 with a 50% ratio.
Bug or just bad matchmaking?

So we can expect to see MMR hell again this season huh?

This reminds me of when I did guild pvp when it was a thing in other seasons. We would have like 8 games played, 5 win 3 loss, and have higher rating than someone who played like 32 games with 25 wins. (exaggeration but it felt this way).

Azukas.1426:

So we can expect to see MMR hell again this season huh?

Since you can drop divisions now, it will be better than before. Because the guyz that made that MMR hell are playing in the bronze division and will go back there if they pass to the silver and are not ready to stand their own in it… and so on.

What Evan didn’t tell to the OP is, it’s not just “accurate” MMR but much more about who you win and loose than how much you win and loose.

If you win 10 matches in a row versus low MMR players you will have a MMR lower than a guy that win 5 matches and lose 5 matches versus high MMR players.

Jourdelune.7456:

Azukas.1426:

So we can expect to see MMR hell again this season huh?

If you win 10 matches in a row versus low MMR players you will have a MMR lower than a guy that win 5 matches and lose 5 matches versus high MMR players.

So, now the question is, why is he matched against lower MMR players? As he said he is climbing very slowly but having high win ratio in same time, that doesn’t make any sence. Shouldn’t he have higher MMR? His MMR should rise until his win/lose ratio is 50/50.

If it goes much lower than 20 it is way too drastic. RNG invalidates win streaks.

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

Question, if we generally have a 60% win rate but happen to have a sheer run of bad luck in our placement matches, our mmr wouldn’t settle until we’re at a level closer to 50% win rate correct?

So if I generally lose 4 games out of 10, my mmr wont “settle” until it’s closer to 5/5?

Saiyan.1704:

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

Question, if we generally have a 60% win rate but happen to have a sheer run of bad luck in our placement matches, our mmr wouldn’t settle until we’re at a level closer to 50% win rate correct?

So if I generally lose 4 games out of 10, my mmr wont “settle” until it’s closer to 5/5?

Depends on the mmr of the othe team, but once the system got enough players around your mmr group, yes

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

Are you implying that since X got their placements done early…they placed high because the system didn’t have a playerbase to work off of yet? But now that there are players, those who placed high will just stay where they are because the “quota” is filled for initial “high MMR”?

Or is there a better explanation to how Pro A (playing early) goes 9-1 and gets 1.8k rating after placements while Pro B (playing later) goes 9-1 and gets 1.2k rating after placements?

And how come EVERYONE so far has at least 1 loss. Has anyone gone 10-0 so far?

You will have some outlier matches every so often. Either you would play teams that you could easily smoke or you would play vs. better teams. Your performance in those matches will affect your MMR the most.

Faux Play.6104:

You will have some outlier matches every so often. Either you would play teams that you could easily smoke or you would play vs. better teams. Your performance in those matches will affect your MMR the most.

Not sure how true this is… I do believe this SHOULD be the case but from recent experience…

Versed someone roughly +200 rating above. Won = +19. Lost = -19.
Versed someone even rating. Won = +19. Loss = -19

Seems to be no difference no matter how many times you verse the outlier. The leaderboard is there, it’s easy to find people you verse (if not just asking them).

LUST.7241:

Faux Play.6104:

You will have some outlier matches every so often. Either you would play teams that you could easily smoke or you would play vs. better teams. Your performance in those matches will affect your MMR the most.

Not sure how true this is… I do believe this SHOULD be the case but from recent experience…

Versed someone roughly +200 rating above. Won = +19. Lost = -19.
Versed someone even rating. Won = +19. Loss = -19

Seems to be no difference no matter how many times you verse the outlier. The leaderboard is there, it’s easy to find people you verse (if not just asking them).

Did you look up the entire team you had vs theirs?

Faux Play.6104:

You will have some outlier matches every so often. Either you would play teams that you could easily smoke or you would play vs. better teams. Your performance in those matches will affect your MMR the most.

That last bit isn’t true though. There is no in game performance measurement (not even a smidge) so even if I do get those matches that tests me, i’m still entirely dependant on my team; whether or not that 1 player on my team is running some yoloQ glasscannon psycho build, it’ll be impossible to win when the going gets tough.

Saiyan.1704:

Faux Play.6104:

You will have some outlier matches every so often. Either you would play teams that you could easily smoke or you would play vs. better teams. Your performance in those matches will affect your MMR the most.

That last bit isn’t true though. There is no in game performance measurement (not even a smidge) so even if I do get those matches that tests me, i’m still entirely dependant on my team; whether or not that 1 player on my team is running some yoloQ glasscannon psycho build, it’ll be impossible to win when the going gets tough.

So is everyone else, the removal of groups makes it fair for all players, because all got the same disadvantage and have to deal with the same randoms.

I guess that means, this season it will be true… ‘learn to carry’ seems to be the way raising your MMR, until you end in an MMR where you can’t carry anymore, because all got roughly the same skill

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

How is it getting more accurate? We’re stomping most of our games and have a 75% winratio. Aren’t we supposed to get more rating for each consecutive win like a winstreak?
We just had bad placements ( 6/4 i believe) after that we won 12 and lost 2 but got les rating for each win.
I think this is not how its supposed to be.

Matches in the first few days are going to be very volatile. Just because you have played more than 20 matches and would like to see your MMR settle, you have to realize there are people who may have just started playing and so the system doesn’t have a accurate read for them. I would expect matches to be better and the resultant MMR gain/loss to improve from next Monday (you need the weekend players to have played their 10 matches).

Volatility probably increases when you winstreak, but is counteracted by the volatility decrease from the sample size of your games increasing.

Don’t worry much, keep a high win rate and you will cruise up. The players up top won’t be going anywhere.

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

But ain’t this contraproductive? I mean, how you want to motivate ppl to play when they can’t progress? I for example had 6 games win streak and for each next win I got less and less points, shouldn’t it be opposite that players get actually rewarded for the win streak?

Also, yesterday I had a match against Helseth for example, which is basically top 10 and my team won, still I got only 20 rank points. So even against the highest ranked ppl the points gain is very low. With this ratio I would have to beat the top ppl like 10 times in row ( or even more since the point gain would be less and less for each next win) to reach their rank. Not even saying that the game doesn’t even give me the oportunity that often to play against the top players (even I am not that far behind, currently rank 70)

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

I am sorry but this makes absolutely no sense at all.

If some one is on a winning streak, how can a stagnant MMR help him get to where his real skill rating is.

IMHO, MMR volatility should be used to determine match making. If someone’s MMR is volatile, he has large probability to match against players further away from his MMR. So he can move towards his real skill rating quicker.

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

So a player who plays 10 games gets 200 mmr for each win…yet someone with the same ratio that plays 100 games gets 10. How does this make sense?

Evan Lesh.3295:

JMcore.4652:

Bug or just bad matchmaking?

Neither. This is how MMR works. The more accurate it gets, the less it moves.

How can you “move” your MMR? What are average rates? From my experience so far it takes at least 100 of games to see a difference.

@Evan: should MMR work solely how it should work? What I mean is the following: after winning 5 games on a row(half of what was defined as a placement sample) players are getting less points for wins. That means they are close to their own average and going against players that are close in rating.

This is not what happens with most competitive games however, is it? After winning a streak, players are “given” a trial match against higher rated players, therefore actually increasing gains after winning while minimizing losses. Here we are more or less doomed into our first ratings regardless of skill, because moving from any average requires hundreds of deviation samples. Note “more or less”, I know exceptional people can climb, but what about “gold” folks that started on bronze?

maxwelgm.4315:

@Evan: should MMR work solely how it should work? What I mean is the following: after winning 5 games on a row(half of what was defined as a placement sample) players are getting less points for wins. That means they are close to their own average and going against players that are close in rating.

This is not what happens with most competitive games however, is it? After winning a streak, players are “given” a trial match against higher rated players, therefore actually increasing gains after winning while minimizing losses. Here we are more or less doomed into our first ratings regardless of skill, because moving from any average requires hundreds of deviation samples. Note “more or less”, I know exceptional people can climb, but what about “gold” folks that started on bronze?

Was afraid of the same thing… I don’t think I belong anywhere higher than gold, (given I used to be a ruby player), but I played my placements on the first day and they were incredibly random, frequent blowouts especially the first few that placed me very low, and then tried to improve the situation in a less-than-ideal, angry state resulted in a screwed up few games right after the placements (which is obviously my fault, should have stopped there and not try but anyway), now I am just trying to play catchups, was winning all yesterday, but it is very slow

^Do realize that having a slow climb is not a problem in itself; you might have a win rate that averages to 50% but is slightly higher than that and your climb is going to be at a correct rate of less than 1%. The problem arises when you are always matched to the mean no matter how many successive victories you have. In this case, you will always face people that are behind your “instantaneous” rating, therefore diminishing returns on your victories.

What I suggest is that the system is indeed working as intended, but it should actually be tweaked to generate “instantaneous” higher rating matches as you get more and more wins on a single streak, giving a short burst of qualification matches for you to place yourself wherever you belong in a certain sense. It is, I believe, a good choice for a game with relatively short competitive seasons, whereas in games with year long PvP seasons, they give plenty of time for players to compensate the difficulty in causing a deviation of the mean distribution of their rating.

I’m sorry – if the system is working as intended you’re generating a major grind for those on lower ranks.. I’m getting a 60% win ratio.. and its moving so slowly.. it needs to be higher that +MMR adjustment.

We’ve all had the fun of the mess of placement matches (3 different matches – 3 people went AFK) – but getting a 80% win ratio on last 20 matches and a slow climb is not fun. The ability of the players that I’m fighting against is honestly sub-par what I would usually play.