Guild Wars 2

Right now there is a guy who is 1026/508 at rank 7, but rank 4 is a guy who is 8/4. 12 games seems like too small of a number to be that highly ranked to me. Can anyone explain to me how the system works?

It is a very simple MMR based system. You start off at a certain MMR, probably around the middle. Your team has an MMR, and the enemy team has an MMR. Whoever wins gains MMR according to what the team MMRs said “should” have happened (equal MMR = 50% chance to win every game), and essentially your MMR is continuously changed with every game, trying to find that perfect number where every game you play in has an exactly 50% chance of you winning against a team of equal rating.

The reason some people get high so fast is that ANet has an MMR system that has very wide swings on wins/losses. In a game like LoL, you only gain or lose like 10 MMR (average MMR is around 1200 in that game), but here it seems like you can gain hundreds per game. So if you get a lucky win streak you can gain massive rating very quickly.

“How do the leaderboards work?”

They don’t.

They really should redo the leader boards x__x
Team Arena should be based on win % since you’re suppose to be pre-made.

Solo Arena should be based on average personal points earned per game (permitting that anet adds the ability to earn points for defending/holding points). You really can’t control the people you play with in solo arena, so you shouldn’t be punished for other peoples inexperience.

XGhoul.7426:

“How to the leaderboards work?”

They don’t.

You beat me to it

Bhawb.7408:

It is a very simple MMR based system. You start off at a certain MMR, probably around the middle. Your team has an MMR, and the enemy team has an MMR. Whoever wins gains MMR according to what the team MMRs said “should” have happened (equal MMR = 50% chance to win every game), and essentially your MMR is continuously changed with every game, trying to find that perfect number where every game you play in has an exactly 50% chance of you winning against a team of equal rating.

The reason some people get high so fast is that ANet has an MMR system that has very wide swings on wins/losses. In a game like LoL, you only gain or lose like 10 MMR (average MMR is around 1200 in that game), but here it seems like you can gain hundreds per game. So if you get a lucky win streak you can gain massive rating very quickly.

Thanks for the answer. So the guy in my initial post just went on a 6-8 game win streak. Seems kinda lame to me as that makes movement up and down the ladder far more susceptible to luck.

sounds very Fair!
But what if every cap only 1 point and stay there as it gives points for defending in soloQ?

Wolfey.3407:

They really should redo the leader boards x__x
Team Arena should be based on win % since you’re suppose to be pre-made.

Solo Arena should be based on average personal points earned per game (permitting that anet adds the ability to earn points for defending/holding points). You really can’t control the people you play with in solo arena, so you shouldn’t be punished for other peoples inexperience.

I always wonder about SoloQ. What if you get a bad streak of nubs on your team yet you keep getting 150+ points in the each match, do they have any effect are your ranking?

If not then the leader boards are quite useless.

The leaderboards use a glicko2 system based on win/loss. Basically, if you beat very high-ranked players, your rating will go up a lot.

The player who is 8/4 probably got the four losses before the leaderboard reset back in November. Those matches still display on his record, but have no effect on his Glicko2 rating.

So he’s likely won 8 matches in a row. Maybe he was just lucky, in which case his rating will quickly decay down and he’ll lose matches if he tries to play again. It’s not anything to worry about either way. Either he’s just that good and he’ll stay there as he plays more, or he got lucky and he’ll be at the top for a few days at most.

Julie Yann.5379:

I always wonder about SoloQ. What if you get a bad streak of nubs on your team yet you keep getting 150+ points in the each match, do they have any effect are your ranking?

If not then the leader boards are quite useless.

The leaderboards shouldn’t be taken too seriously because there’s just no way to know exactly how good you are. But win/loss is a perfectly acceptable parameter. Good players cause their teams—even bad teams—to win more often. The better the player, the larger the effect on the win rate.

There is an old saying garbage in → garbage out I think the problem lies with the used data of the MM-algorithm.

So how do I get the “garbage” in in the first place?
Let me give u a small example:

In games like Dota or HON u start with a fix MMR like 1500. If u win a game you will get like 20 points max so u have 1520 after the match.

In Guildwars you will get way more points in the first games, the reason for this is that players will faster play with other in his league, at first this is a good thing but it poses a loot of problems.
>>players with 15 games in the top 25
>>but the more drastic problem with this system is, if you are “lucky” and get 7 good teams in a row and win. You will be ranked very high even if it is not your “real” MMR

So obvious you don’t need to play yourself “up” and would hit a point where you loose as many as as u win and so play with others in your league.
You rather get put some where in the leaderboard that is based on 10 games which is a pretty low quantity and therefore not really precise, since the accuracy of your MMR increases with the number of Games you play.

The result is that the “lucky” players poison the leaderboards and even at high ranks u will still play with new players.

So how can the algorithm get more precise?
>> let the PL play themselves up since more games make the MMR accurate
>> since we have a 500 score for each match use this score if a team wins 500:100 they will get more ranks them if they are playing 500:499 this also increases teamplay and even the loosing team will still try to get as many points as possible.
>> don’t count unfair matches (4 vs 5) just remake them or add a 10th person

BTW: the used Glicko2 alo was designed for chess a 1 vs 1 game with win lose and draw :-) and not a 5 vs 5 player game

Pollution.8703:

Bhawb.7408:

It is a very simple MMR based system. You start off at a certain MMR, probably around the middle. Your team has an MMR, and the enemy team has an MMR. Whoever wins gains MMR according to what the team MMRs said “should” have happened (equal MMR = 50% chance to win every game), and essentially your MMR is continuously changed with every game, trying to find that perfect number where every game you play in has an exactly 50% chance of you winning against a team of equal rating.

The reason some people get high so fast is that ANet has an MMR system that has very wide swings on wins/losses. In a game like LoL, you only gain or lose like 10 MMR (average MMR is around 1200 in that game), but here it seems like you can gain hundreds per game. So if you get a lucky win streak you can gain massive rating very quickly.

Thanks for the answer. So the guy in my initial post just went on a 6-8 game win streak. Seems kinda lame to me as that makes movement up and down the ladder far more susceptible to luck.

He went on a win streak but his team still had to beat some highly ranked comps to end up that high. I had an even luckier streak to end up something like 8-3 but I was only in the 400’s or so upon being ranked for the first time.

That being said, it should take a lot more than 11 games to get such a high rank.

I am 9/3 I think right now, and hit 60 (before decay). If I’m not actually that good, then over time as I played more and more games, that rating will balance itself out, since I should lose more games than I win if I am above where I should be, which will drop my rating.

They basically have to do this because while it makes leaderboards a bit “strange”, it speeds up the process of getting you around the correct skill level, and then it should fix itself to as close to perfect as possible over time. If it was slower, then an improper rating would stick for much longer, meaning players who are rated way above what they should be would be ruining games for their teams much longer (whereas now their ranking will drop like a rock in a few games), or players who are way below where they should be would be destroying players they shouldn’t even be against.

There are good and bad things about it.

Bhawb.7408:

I am 9/3 I think right now, and hit 60 (before decay). If I’m not actually that good, then over time as I played more and more games, that rating will balance itself out, since I should lose more games than I win if I am above where I should be, which will drop my rating.

They basically have to do this because while it makes leaderboards a bit “strange”, it speeds up the process of getting you around the correct skill level, and then it should fix itself to as close to perfect as possible over time. If it was slower, then an improper rating would stick for much longer, meaning players who are rated way above what they should be would be ruining games for their teams much longer (whereas now their ranking will drop like a rock in a few games), or players who are way below where they should be would be destroying players they shouldn’t even be against.

There are good and bad things about it.

In all honesty. I havent met a single “9-3 person” that could carry his load as a full fledged top 100 player. Which isnt weird because you only played like 10 games.

Its far more common that the guys that got catapulted to the top, ruin higher MMR games due to their lack of experience, awareness or mechanical skill. Its actually so common thats its one of the main reasons why some very specific higher rated ppl rage in chat.

So i think it would better to let their MMR rise at a slower pace….like in other games. Let them learn the gameplay and pvp mode specific tricks and rotations slowly but steadily.

I agree it should be a bit slower, but having it too slow ruins low games as well. Its simply a matter of which side do you want to screw up more. Also, lets say I am bad since I’m using myself for the example (and frankly I’m probably not top 100 anyway); with the huge swings one game will also drop me way out of the top brackets. So while I might screw up one game, it is only one game, and then I’m down in the top 500ish.

Also realize that you could have a player with 2k games in soloQ or teamQ go into the opposite game mode (obviously an extreme example). If they do significantly slow the progress, they need to introduce ways for people who get placed far outside (above or below) their actual deserved rating to be quickly shifted. In League they have ways to skip divisions, both up and down, due to how slowly you can rise or fall compared to how much better or worse you might be.

Not saying that the system is perfectly, but the wild swings, in my opinion, allow for them to have such a simple system without too badly ruining games over a long period of time since things adjust so quickly.

Bhawb.7408:

I agree it should be a bit slower, but having it too slow ruins low games as well. Its simply a matter of which side do you want to screw up more. Also, lets say I am bad since I’m using myself for the example (and frankly I’m probably not top 100 anyway); with the huge swings one game will also drop me way out of the top brackets. So while I might screw up one game, it is only one game, and then I’m down in the top 500ish.

Also realize that you could have a player with 2k games in soloQ or teamQ go into the opposite game mode (obviously an extreme example). If they do significantly slow the progress, they need to introduce ways for people who get placed far outside (above or below) their actual deserved rating to be quickly shifted. In League they have ways to skip divisions, both up and down, due to how slowly you can rise or fall compared to how much better or worse you might be.

Not saying that the system is perfectly, but the wild swings, in my opinion, allow for them to have such a simple system without too badly ruining games over a long period of time since things adjust so quickly.

The entire point of an MMR system is too seperate new players from experienced veterans. Yet i see guys who face rank 1s on the leaderboard during their first few games. And it doesnt happen once…Its every single f ing game.

I see one experienced guy give up after the other. Its far to frustrating due to the horrible matchmaking system.

Bhawb.7408:

It is a very simple MMR based system. You start off at a certain MMR, probably around the middle. Your team has an MMR, and the enemy team has an MMR. Whoever wins gains MMR according to what the team MMRs said “should” have happened (equal MMR = 50% chance to win every game), and essentially your MMR is continuously changed with every game, trying to find that perfect number where every game you play in has an exactly 50% chance of you winning against a team of equal rating.

The reason some people get high so fast is that ANet has an MMR system that has very wide swings on wins/losses. In a game like LoL, you only gain or lose like 10 MMR (average MMR is around 1200 in that game), but here it seems like you can gain hundreds per game. So if you get a lucky win streak you can gain massive rating very quickly.

This is a good explanation, only thing I have to add is the reason that new players can shoot to the top of the LB is because volatility is very high when people don’t have a long record of past games. Someone like me with 300 team queue games can win 5 games in a row and gain some ranks, someone just starting out with a high volatility will get into the top 10. The same thing happened to me when I first started playing solo queue, now when I have a winning streak it barely changes my rank.

Julie Yann.5379:

I always wonder about SoloQ. What if you get a bad streak of nubs on your team yet you keep getting 150+ points in the each match, do they have any effect are your ranking?

If not then the leader boards are quite useless.

Win loss ratio only. A dev answered this a while ago.

Wolfey.3407:

They really should redo the leader boards x__x
Team Arena should be based on win % since you’re suppose to be pre-made.

No. Then people can just farm easy wins during offpeak hours and get a crazy 90% win ratio.

Ryan.9387:

Julie Yann.5379:

I always wonder about SoloQ. What if you get a bad streak of nubs on your team yet you keep getting 150+ points in the each match, do they have any effect are your ranking?

If not then the leader boards are quite useless.

Win loss ratio only. A dev answered this a while ago.

So they are pointless cause they aren’t a reflection of anybodies skill.

Julie Yann.5379:

So they are pointless cause they aren’t a reflection of anybodies skill.

Win/loss is the best parameter for measuring skill that I can think of. Skilled players tend to help their teams win more often than not. Sure, there’s a margin of error—players who get lucky and land on great teams their first ten games tend to shoot up the leaderboard, then quickly settle back down. Some players get unlucky and land on awful teams or 4v5s multiple times in a row, and fall down lower than they rightfully should be. But over time, the extent to which you help your team win is the best tangible way to measure skill.

Edit: I don’t mean to say that the leaderboards are a perfect reflection of skill, because they’re not. Just saying that win/loss is an acceptable parameter.

NevirSayDie.6235:

Julie Yann.5379:

So they are pointless cause they aren’t a reflection of anybodies skill.

Win/loss is the best parameter for measuring skill that I can think of. Skilled players tend to help their teams win more often than not. Sure, there’s a margin of error—players who get lucky and land on great teams their first ten games tend to shoot up the leaderboard, then quickly settle back down. Some players get unlucky and land on awful teams or 4v5s multiple times in a row, and fall down lower than they rightfully should be. But over time, the extent to which you help your team win is the best tangible way to measure skill.

Edit: I don’t mean to say that the leaderboards are a perfect reflection of skill, because they’re not. Just saying that win/loss is an acceptable parameter.

I can see this working for Tpvp where people usually create structured teams with proper comps and such. SoloQ however is all up to chance, you basically get lucky enough to have a decent comp or not. However, if they included a variable for how many points you make in any giving match it would offset the randomness of the horrible Solo match maker. You could lose a match because of a bad team make up but still not lose any points because of what you contributed to the fight. On the flip side, you could be lucky a get a spot on the winning team but your contribution is so low that you don’t gain any points. There would have to be ways for counting bunkers who defend nodes but don’t get a lot of kills and such but that is the general idea and would be way more accurate than the current system.

Just saying that there is a better way for determining ranks on the leaderboard. Though I doubt we will ever see anything that sophisticated.

Julie Yann.5379:

Ryan.9387:

Julie Yann.5379:

I always wonder about SoloQ. What if you get a bad streak of nubs on your team yet you keep getting 150+ points in the each match, do they have any effect are your ranking?

If not then the leader boards are quite useless.

Win loss ratio only. A dev answered this a while ago.

So they are pointless cause they aren’t a reflection of anybodies skill.

Assume that every game there is a player so bad his team will lose no matter what. 100% of the time you arent that player.

You will lose 44% of the time (4 / 9)
You will win 56% of the time (5 / 9)

This is a really simple answer but it shows that your mmr isn’t 100% dependent upon your allies. The leaderboards generally reflect this. Of the top 100 NA:

16 have under 56% win/loss
8 have under 53% win/loss
1 has under 50% win/loss

Julie Yann.5379:

I can see this working for Tpvp where people usually create structured teams with proper comps and such. SoloQ however is all up to chance, you basically get lucky enough to have a decent comp or not. However, if they included a variable for how many points you make in any giving match it would offset the randomness of the horrible Solo match maker. You could lose a match because of a bad team make up but still not lose any points because of what you contributed to the fight. On the flip side, you could be lucky a get a spot on the winning team but your contribution is so low that you don’t gain any points. There would have to be ways for counting bunkers who defend nodes but don’t get a lot of kills and such but that is the general idea and would be way more accurate than the current system.

Just saying that there is a better way for determining ranks on the leaderboard. Though I doubt we will ever see anything that sophisticated.

So if I double-capped home at start and got ten extra points, that would raise my MMR? Or if I survived a 1v3 on far point for 10 minutes but didn’t get any kills or caps, my MMR would go down?

Sometimes solo queue teammates feel random, but that doesn’t change the value of win/loss as a parameter for the leaderboard. Even if there were no matchmaking and every solo queue was totally random, the strongest players would win much more often than average players. Again, I’m not saying the leaderboards are great or super accurate; but on the whole, they’re probably more accurate than they would be using additional parameters like score.

NevirSayDie.6235:

So if I double-capped home at start and got ten extra points, that would raise my MMR? Or if I survived a 1v3 on far point for 10 minutes but didn’t get any kills or caps, my MMR would go down?

That’s why I said they would have to include some way of calculating bunkers that hold points. As for the randomness, yesterday I got put on a team with 4 rangers, not saying that rangers are bad (well the 2 Bear Bows were) , but it is a disadvantage when facing a well balanced team. I got 3 4v5 matchess, and was on the losing side of Skyhammer 4 times in a row. All that time I was top 2 scorer by a a considerable margin between me and the guy below me. My MMR dropped and there was nothing I could do to stop it. 1 or 2 people cannot carry a whole team.

That’s why I say that the current system is pointless. It doesn’t mean a thing when so many factors out of your control influence the results.

Julie Yann.5379:

NevirSayDie.6235:

So if I double-capped home at start and got ten extra points, that would raise my MMR? Or if I survived a 1v3 on far point for 10 minutes but didn’t get any kills or caps, my MMR would go down?

That’s why I said they would have to include some way of calculating bunkers that hold points. As for the randomness, yesterday I got put on a team with 4 rangers, not saying that rangers are bad (well the 2 Bear Bows were) , but it is a disadvantage when facing a well balanced team. I got 3 4v5 matchess, and was on the losing side of Skyhammer 4 times in a row. All that time I was top 2 scorer by a a considerable margin between me and the guy below me. My MMR dropped and there was nothing I could do to stop it. 1 or 2 people cannot carry a whole team.

That’s why I say that the current system is pointless. It doesn’t mean a thing when so many factors out of your control influence the results.

I agree that skyhammer and 4v5s are both massive problems that are killing solo queue right now. They’re the reasons I usually solo join team arenas instead of solo queuing.

Skyhammer is easy, just needs to be removed from the rotation. 4v5s are tougher and will likely never be completely eliminated, but simple things like an audible ping on queue pop and a 1-minute confirmation button required to enter the map would help.

So yes, there are issues like this making the leaderboards less meaningful, but I don’t think changing the parameters would be helpful.

leaderboards and mmr are a joke. What’s the point of being in the top 100 or w/e when you constantly get matched with people who have just started or aren’t even in top 1k?

Julie Yann.5379:

That’s why I said they would have to include some way of calculating bunkers that hold points. As for the randomness, yesterday I got put on a team with 4 rangers, not saying that rangers are bad (well the 2 Bear Bows were) , but it is a disadvantage when facing a well balanced team. I got 3 4v5 matchess, and was on the losing side of Skyhammer 4 times in a row. All that time I was top 2 scorer by a a considerable margin between me and the guy below me. My MMR dropped and there was nothing I could do to stop it. 1 or 2 people cannot carry a whole team.

That’s why I say that the current system is pointless. It doesn’t mean a thing when so many factors out of your control influence the results.

This. Out of 3 matchs today, 1 was lost because we were 4v5 for 1mn, second was was lost because it was 4v5 the whole match, and the third one was lost because people were fighting off point… Nothing i can do here..
The funniest part was when i checked leaderboards. Last week , i was having -150 ranks for 1 win 2 loses. Now i noticed i went up +2% for 3 loses..

At least, the system should be much less volatile. 4v5’s and bad match ups would be less annoying if you didn’t drop so much in the ranking( thus getting a worse mmr)

I as top 100-200 got an instant pop just now and got matched with ppl that dont even excist on the leaderboard yet. Thats how new they are.

Not in the top 1000 OR in the percentages.

never understood leaderbord, just shows some numbers but not what they stand for.