Guild Wars 2

So, I was browsing the leaderboards today and I am literally laughing.

This guy has 50/50 w/l ratio and is there at 1849 rating.
http://i.imgur.com/tM3FetT.jpg

This guy on the other hand won 6, lost 4 and still he is placed at 1881.
http://i.imgur.com/AVJQWsX.jpg

So, I have won over a half of the matches I played and have only 953…

If you are worried about your rating being to low do not, anet simply doesn’t even want you to play with those people xD

Of course this is probably due to duoqueuing, but still, horrible system.

He’s playing harder opponents because he had high MMR prior to the start of the season.

winning over half your matches against bronze tier players should count for less than winning matches in the plat/legend division. That is reflected in the rankings.

Main account, I went 8/2 in placements, got placed at ~1420 (gold 1). 4-season legend.

Alt, non-hot account for ascended farm, I went 5/5, placed at ~1415. Never played ranked before Tuesday.

Both solo queuing. Truly weird system. I PM’d Evan about it because I was just genuinely curious, but he has declined to respond. I’m guessing it’s just the quality of the opponent in the amalgamated matches, but 80% win rate and 50% is pretty vastly different to end up in the same exact place.

Nonsense. Since your also teamed with utterly random bronze tier players its much harder to win, then if your teamed with folks you can moderately rely on to be competent

So I cannot raise my ranking because opponents I fight are supposed to be idiots in contrary to opponents in higher divisions? I think this makes no sense at all.

vove.2768:

So I cannot raise my ranking because opponents I fight are supposed to be idiots in contrary to opponents in higher divisions? I think this makes no sense at all.

I had proposed this problem to Lesh in another thread. That by bringing new players at an average MMR of 1200, if there was a steady influx of new players, it would be pretty random when you’d ever be able to break above the range that those players could be placed in and randomly tank you.

I suspect that getting through gold will be easier than getting through silver because of the randomness of player quality and the fact that you individually can’t make that much difference, so its sort of up to the gods of luck. Now those whom sailed through will say the system is fine and all their skill. And those whom have significant trouble will attribute it all to the fact that their teammates were poor. The answer will lie somewhere in the middle.

Also the problem is not the opponents you fight, its the folks you have fighting with you

I went 4-6 on my main account placements and got bottom of silver. Horrible teammates and literally had 2-3 thieves a game who all refused to switch. Since then I’m sitting on an 80% win rate and pushing into T2 gold.

On my alt EU account w/out HoT I went 6-4 and placed Mid Gold. Now 2 of those loses were due to people leaving the game so I would have prolly placed Plat on a non HoT account playing with across the ocean lag.

The problem with this season is the placement matches. The rest is awesome and I enjoy playing the season so far.

Azukas.1426:

I went 4-6 on my main account placements and got bottom of silver. Horrible teammates and literally had 2-3 thieves a game who all refused to switch. Since then I’m sitting on an 80% win rate and pushing into T2 gold.

On my alt EU account w/out HoT I went 6-4 and placed Mid Gold. Now 2 of those loses were due to people leaving the game so I would have prolly placed Plat on a non HoT account playing with across the ocean lag.

The problem with this season is the placement matches. The rest is awesome and I enjoy playing the season so far.

Clearly all those whom argue your ranking is an accurate reflection of your skill would simply tell you that EU is clearly much easier to play in than NA

It is exactly because this sort of thing happens that I’d argue that a lot of your ranking is up to dumb luck of the draw. You can’t argue that the rankings are an accurate determination of skill, and explain how the same player can end up completely different on two accounts.

I’m sure that with theoretically infinite games you’ll eventually end up where your supposed to be, but that thinking is flawed in that the season would end before that ever happened. The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

Azukas.1426:

The problem with this season is the placement matches. The rest is awesome and I enjoy playing the season so far.

I’m just thinking this out myself mostly, but maybe that’s not such a horrible thing? Obviously there will be growing pains, integrating players into this new system from the old one, of straight division-grinding until you level out somewhere.

Going forward, the system will know your previous seasons’ MMR, and take that into account as the data comes in with the w/l records of your placements. The system can then account for whether it is a normal or abnormal result and weight it appropriately.

Maybe that will be how it works, maybe not. Maybe we’re just seeing the forest right now, and not the trees. Which sucks to have to deal with, but adaptation to new systems isn’t usually easy either.

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

You make a straw man argument. I am not saying that Gliko is not a good rating system, just that the application here is not the intended use case.

Incidentally your need to resort to Ad Hominems doesn’t improve anyone’s already questionable opinion of you.

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

The system may not get you to the perfect spot your first match but after enough matches you will get there. The placement and soft reset provide enough volatility for people that have significantly improved over a season the opportunity to quickly jump their rating. Someone that deserves to be in gold will be able to win their way out of bronze. Yes you have bronze tier players on your team but so does the other team, and since you should be significantly better on average you will be able to win enough matches to get out of that division. Yes you will get some bad teammates you can’t overcome, but since there are no gates to your MMR, the bad players will continue to fall and the good players will rise.

Glicko is fine for the most part, even with the modifications for 5v5.

The problem, I think, is that the volatility goes away too quickly. Placement matches are intentionally volatile and random. Players placed in the right division from placements will remain there. In theory, players placed either too low or too high will either outplay the other team, or be outplayed, and go up or down.

In practice, that doesn’t appear to be happening. Players placed too low and go on a win-streak where every match is won by more than 150 points aren’t gaining enough points to break them out of a low division. I won more than 80% of the matches immediately post-placement, but my rating changes quickly stabilized to +10-15 points per win. I don’t think that should have happened, as something like that should have pushed me higher (and if it turned out I really don’t belong in that high division, I would expect to go on a losing streak).

I’d say keep the extra volatility for longer. 10 Placement matches, plus at least 10-20 more. The extra matches post-placement will give more data to determine where a player really begins.

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

Starting NFL lineups change all the time through injuries, trades, and throughout the game as you give non-starters snaps.

CS:GO is a team game as well. As is League.

If people don’t adapt before match start to get a better comp, they’re likely not doing everything in their power to win anyway, and you can see how that plays out. At the higher levels people switch all the time. Class MMR has been separate for a long time as well. Maybe you multi classed for your overall MMR.

You still haven’t come up with a better system.

Faux Play.6104:

The system may not get you to the perfect spot your first match but after enough matches you will get there. The placement and soft reset provide enough volatility for people that have significantly improved over a season the opportunity to quickly jump their rating. Someone that deserves to be in gold will be able to win their way out of bronze. Yes you have bronze tier players on your team but so does the other team, and since you should be significantly better on average you will be able to win enough matches to get out of that division. Yes you will get some bad teammates you can’t overcome, but since there are no gates to your MMR, the bad players will continue to fall and the good players will rise.

The problem is your speaking of the average average situation. In a system like this with enough players you’ll hit a good number of outliers whom simply get enough bad match ups for long enough that their MMR will settle, and that if they did get to the 1200 mark again would randomly draw the new player card on their team.

It might help if people with highly random MMR’s were more likely to face highly random folks, this way having slowly worked your way up from 900 to the 1200 range wouldn’t put you into the “did I get a new player lottery” So basically including the confidence level in the match making.

And yes, you all played forever you’d probably get there….. but that doesn’t happen. And might not be able to if you look at a lack of class stacking understanding from the algorithm and playing classes for acheivements that folks haven’t been rated with. You can build pendantic cases where you’d simply never get there.

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

Faux Play.6104:

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

I’m actually arguing that their skill will likely be different in each of those classes, and the match making should be based on the class their playing to have more accurate results.

shion.2084:

Faux Play.6104:

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

I’m actually arguing that their skill will likely be different in each of those classes, and the match making should be based on the class their playing to have more accurate results.

I just switched to Rev at the end of blow out games to get the achievements. So I have like a 95% winrate on it with whatever MMR came from that. The system probably thinks I’m a god at rev, it has no other way to tell at the moment. Let’s just say thank god it doesn’t.

Now, if you want to argue they should keep track of which class you spent most of the time in the match on then you could reasonably die 8 times and switch classes after every death and that adds too much more unpredictability which we don’t want.

shion.2084:

Faux Play.6104:

The system may not get you to the perfect spot your first match but after enough matches you will get there. The placement and soft reset provide enough volatility for people that have significantly improved over a season the opportunity to quickly jump their rating. Someone that deserves to be in gold will be able to win their way out of bronze. Yes you have bronze tier players on your team but so does the other team, and since you should be significantly better on average you will be able to win enough matches to get out of that division. Yes you will get some bad teammates you can’t overcome, but since there are no gates to your MMR, the bad players will continue to fall and the good players will rise.

The problem is your speaking of the average average situation. In a system like this with enough players you’ll hit a good number of outliers whom simply get enough bad match ups for long enough that their MMR will settle, and that if they did get to the 1200 mark again would randomly draw the new player card on their team.

It might help if people with highly random MMR’s were more likely to face highly random folks, this way having slowly worked your way up from 900 to the 1200 range wouldn’t put you into the “did I get a new player lottery” So basically including the confidence level in the match making.

And yes, you all played forever you’d probably get there….. but that doesn’t happen. And might not be able to if you look at a lack of class stacking understanding from the algorithm and playing classes for acheivements that folks haven’t been rated with. You can build pendantic cases where you’d simply never get there.

I experimented with 2 f2p accounts, and the system seems to work fine. One I placed high in the other I did not. After playing both accounts about 25 matches they were approximately the same rating. I’m not a great player, but give me a bronze tier account and I can get it to upper tier silver/gold in less than 40 matches.

Honest John.4673:

shion.2084:

Faux Play.6104:

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

I’m actually arguing that their skill will likely be different in each of those classes, and the match making should be based on the class their playing to have more accurate results.

I just switched to Rev at the end of blow out games to get the achievements. So I have like a 95% winrate on it with whatever MMR came from that. The system probably thinks I’m a god at rev, it has no other way to tell at the moment. Let’s just say thank god it doesn’t.

I’m not sure that helps the argument that Gliko is a good choice when these player variables exist

shion.2084:

Faux Play.6104:

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

I’m actually arguing that their skill will likely be different in each of those classes, and the match making should be based on the class their playing to have more accurate results.

Then you are ignoring the rules of the game. The game allows you to change your build and class before the match starts. If you have enough knowledge of the game and ability to be able to pick a different build or class that will put your team in a better place then you deserve to be rated higher. Would you argue that if someone played and won without any runes or traits they should get more points than if they ran a good build? It will take more skill to win a match that way, but on average you will be hurting your chances of winning as well as your team’s.

shion.2084:

Honest John.4673:

shion.2084:

Faux Play.6104:

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

I’m actually arguing that their skill will likely be different in each of those classes, and the match making should be based on the class their playing to have more accurate results.

I just switched to Rev at the end of blow out games to get the achievements. So I have like a 95% winrate on it with whatever MMR came from that. The system probably thinks I’m a god at rev, it has no other way to tell at the moment. Let’s just say thank god it doesn’t.

I’m not sure that helps the argument that Gliko is a good choice when these player variables exist

They exist, did you miss the part where I said it’s separate? It’s not used in the calculation at all, they just track it.

vove.2768:

So, I was browsing the leaderboards today and I am literally laughing.

This guy has 50/50 w/l ratio and is there at 1849 rating.
http://i.imgur.com/tM3FetT.jpg

This guy on the other hand won 6, lost 4 and still he is placed at 1881.
http://i.imgur.com/AVJQWsX.jpg

So, I have won over a half of the matches I played and have only 953…

If you are worried about your rating being to low do not, anet simply doesn’t even want you to play with those people xD

Of course this is probably due to duoqueuing, but still, horrible system.

lol, what crap of post it is
your mmr is not a result of matches in this league, is a “skill” calculation based on your story in pvp, the best players in this league are the “historical” best players that plays that league not the ones whith best winratio in this league
the main fault of this league sistem is not set a timegate to join it after a few weeks dont leting enter more players ,and not set a minimum number of matches to qualify for being in rank, that causes that “historical best” players play few, just to mantain their status and not geting downed for not playing

megilandil.7506:

vove.2768:

So, I was browsing the leaderboards today and I am literally laughing.

This guy has 50/50 w/l ratio and is there at 1849 rating.
http://i.imgur.com/tM3FetT.jpg

This guy on the other hand won 6, lost 4 and still he is placed at 1881.
http://i.imgur.com/AVJQWsX.jpg

So, I have won over a half of the matches I played and have only 953…

If you are worried about your rating being to low do not, anet simply doesn’t even want you to play with those people xD

Of course this is probably due to duoqueuing, but still, horrible system.

lol, what crap of post it is
your mmr is not a result of matches in this league, is a “skill” calculation based on your story in pvp, the best players in this league are the “historical” best players that plays that league not the ones whith best winratio in this league
the main fault of this league sistem is not set a timegate to join it after a few weeks dont leting enter more players ,and not set a minimum number of matches to qualify for being in rank, that causes that “historical best” players play few, just to mantain their status and not geting downed for not playing

The system allows the top players to accumulate rating points and doesn’t force them to queue for hours to get a match. Whatever the current #1 rating is now will be less than the score of the number 1 player at the end of the season. As they get closer to the top, the rating points will get more and more valuable, while the max rating you can qualify with will be the same. The system will reward the people that can sustain a high win ratio.

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

Faux Play.6104:

shion.2084:

Honest John.4673:

shion.2084:

Honest John.4673:

shion.2084:

The problem with algorithms that seem logically reasonable but don’t account for practical reality failings.

If you want to design, code and implement a better algorithm be my guest. But judging from your complete lack of awareness of even the most simple key concepts of ratings and matchmaking you’d get nowhere. ANet didn’t come up with these algorithms, they’ve been industry standard for years, been adapted to multiple team sports and have been improved upon and updated over the years.

But I’m so glad that we have people like you, through your unmatched genius to point out flaws in these systems that the Ph.D’s and statisticians somehow failed to spot. I’m sure Mark Glickman would love to hear your analysis.

I’m sure the Glickman algorithm is fine if you are running all 5 members of your team. I don’t think he’d support it in this instance. I mean just think about it.

Like I literally just said, it can and has been used and adapted to other team based games/sports like LoL, CS:GO, the NFL, etc.

EDIT: Please don’t attack the tone of my argument, use the fallacy fallacy and then do the thing you literally just said didn’t want to happen to you.

And I argue with the adaption. I maintain its reasonable for its intended original purpose of Chess or Go. If the team composition remains mostly consistent such as an NFL team, then you could treat that as one “person” as well. However to apply it to an ever changing grouping of 5 folks whom need not play the class their rating was based on etc. is obviously going to be fraught with problems. Then you add in class stacking. then, then…. you’ll see that these derivations from the original purpose although minor seeming will result in well…. unpredictability.

The fallacy fallacy would be to argue you were wrong in your assertion because you made an ad hominem. I simply indicated that you argued against the wrong point (straw man) and that your ad hominem is not relevant.

So are you arguing that if someone can be competitive at multiple classes they don’t deserver to be ranked higher than someone that is one dimensional?

I’m actually arguing that their skill will likely be different in each of those classes, and the match making should be based on the class their playing to have more accurate results.

I just switched to Rev at the end of blow out games to get the achievements. So I have like a 95% winrate on it with whatever MMR came from that. The system probably thinks I’m a god at rev, it has no other way to tell at the moment. Let’s just say thank god it doesn’t.

I’m not sure that helps the argument that Gliko is a good choice when these player variables exist

They exist, did you miss the part where I said it’s separate? It’s not used in the calculation at all, they just track it.

I wasn’t arguing if they exist or not, I was conceeding they do…. I don’t even know what you believe you are responding to.

I’ve explained why the Gliko is not appropriate to make reasonable matches in a situation where players can play classes that are not the ones they were rated for. I don’t think that anyone would argue the prediction is at all reliably accurate in these circumstances. Eg. I have played all season as an ele. I want my engi acheive so I swap classes, Gliko places me based on my skill rating I’d received as an ele.

I mean this is only a fringe example of why it doesn’t work. It also can’t work out that certain class compositions cause entirely different predicted results. etc. etc.

In your NFL scenario if the entire starting lineup was wiped out by plague, the odds makers wouldn’t give you the same odds on the game. And the Gliko prediciton would be off as well. Whenever you entirely assemble a completely new team, the synergies against a completely different opponent will be incredibly hard to predict. If it were the same 5 people playing on the team, it would be more reasonable to apply.

shion.2084:

It is exactly because this sort of thing happens that I’d argue that a lot of your ranking is up to dumb luck of the draw. You can’t argue that the rankings are an accurate determination of skill, and explain how the same player can end up completely different on two accounts.

Statistically speaking a sample of 10 is indeed dumb luck of the draw. Especially early on when rating volatility is greater. Try playing at least 30-50 games before making judgment.

Zenith.6403:

shion.2084:

It is exactly because this sort of thing happens that I’d argue that a lot of your ranking is up to dumb luck of the draw. You can’t argue that the rankings are an accurate determination of skill, and explain how the same player can end up completely different on two accounts.

Statistically speaking a sample of 10 is indeed dumb luck of the draw. Especially early on when rating volatility is greater. Try playing at least 30-50 games before making judgment.

Agreed, and if your the average average guy it might work out eventually… or the season might end, or you might keep hitting 1200 and getting whacked with new players by bad draw luck before things average out in the long run. You just never know.

shion.2084:

Zenith.6403:

shion.2084:

It is exactly because this sort of thing happens that I’d argue that a lot of your ranking is up to dumb luck of the draw. You can’t argue that the rankings are an accurate determination of skill, and explain how the same player can end up completely different on two accounts.

Statistically speaking a sample of 10 is indeed dumb luck of the draw. Especially early on when rating volatility is greater. Try playing at least 30-50 games before making judgment.

Agreed, and if your the average average guy it might work out eventually… or the season might end, or you might keep hitting 1200 and getting whacked with new players by bad draw luck before things average out in the long run. You just never know.

A new player isn’t going to play a sustained number of matches at 1200. Their volatility will be high. If they are good (someone’s alt) they will move up, if they are new to the game they will rapidly fall. Only if that is where they belong will they stay there. Only Anet has the data, but you can’t say you are more likely to get an experienced player’s alt or a brand new player coming in at a 1200 rating for their first match.

Faux Play.6104:

shion.2084:

Zenith.6403:

shion.2084:

It is exactly because this sort of thing happens that I’d argue that a lot of your ranking is up to dumb luck of the draw. You can’t argue that the rankings are an accurate determination of skill, and explain how the same player can end up completely different on two accounts.

Statistically speaking a sample of 10 is indeed dumb luck of the draw. Especially early on when rating volatility is greater. Try playing at least 30-50 games before making judgment.

Agreed, and if your the average average guy it might work out eventually… or the season might end, or you might keep hitting 1200 and getting whacked with new players by bad draw luck before things average out in the long run. You just never know.

A new player isn’t going to play a sustained number of matches at 1200. Their volatility will be high. If they are good (someone’s alt) they will move up, if they are new to the game they will rapidly fall. Only if that is where they belong will they stay there. Only Anet has the data, but you can’t say you are more likely to get an experienced player’s alt or a brand new player coming in at a 1200 rating for their first match.

Agreed, but I’m arguing what if there are a steady stream of new players being introduced. It’s not the same new player, its a different one each time… or the same… but you get what I mean.

shion.2084:

Faux Play.6104:

shion.2084:

Zenith.6403:

shion.2084:

It is exactly because this sort of thing happens that I’d argue that a lot of your ranking is up to dumb luck of the draw. You can’t argue that the rankings are an accurate determination of skill, and explain how the same player can end up completely different on two accounts.

Statistically speaking a sample of 10 is indeed dumb luck of the draw. Especially early on when rating volatility is greater. Try playing at least 30-50 games before making judgment.

Agreed, and if your the average average guy it might work out eventually… or the season might end, or you might keep hitting 1200 and getting whacked with new players by bad draw luck before things average out in the long run. You just never know.

A new player isn’t going to play a sustained number of matches at 1200. Their volatility will be high. If they are good (someone’s alt) they will move up, if they are new to the game they will rapidly fall. Only if that is where they belong will they stay there. Only Anet has the data, but you can’t say you are more likely to get an experienced player’s alt or a brand new player coming in at a 1200 rating for their first match.

Agreed, but I’m arguing what if there are a steady stream of new players being introduced. It’s not the same new player, its a different one each time… or the same… but you get what I mean.

1) You have the same chance to play against the inexperienced player as to play with them.
2) The same argument holds for experienced player’s alts.

That is why there is a high volatility at the start to quickly move the outliers.

Condensing the rating down to winning matches is the simplest and most elegant way of doing it. It eliminates a huge amount of variables and is really the only important metric. Someone could be very mechanically skilled, but make awful decisions about where to fight and cost his team the win. The metric is evaluating not only your mechanical skill, but your overall understanding of the game mode and your ability to maximize your team’s chances of winning.

shion.2084:

I’ve explained why the Gliko is not appropriate to make reasonable matches in a situation where players can play classes that are not the ones they were rated for. I don’t think that anyone would argue the prediction is at all reliably accurate in these circumstances.

I guess I just don’t understand why (apart from the AP example) anyone would swap to a class they are horrible at in. the. first. place.

First you say it should be based off class.

match making should be based on the class their playing to have more accurate results.

Then I tell you they aren’t even using it in the calculation. And for good reason, cause it’s a stupid thing to do. Then you somehow thought that proved your point from the start that they shouldn’t be using glicko in the first place. Then you say it shouldn’t be based off class using the exact same example I gave to you why they don’t use it in the calculation.

You’re a weird one, I’ll give you that.

You guys are missing a huge point. The starting MMR was (old invisible MMR +1200)*0.5 averaging .
So two different accounts can have the same w/l ratio in the placements and have different ratings.

Answer is because you started at different points to begin with.

shion.2084:

vove.2768:

So I cannot raise my ranking because opponents I fight are supposed to be idiots in contrary to opponents in higher divisions? I think this makes no sense at all.

I had proposed this problem to Lesh in another thread. That by bringing new players at an average MMR of 1200, if there was a steady influx of new players, it would be pretty random when you’d ever be able to break above the range that those players could be placed in and randomly tank you.

Yep, anyone around 1200 rating ends up in the dumping ground for placement matches. I’m pretty sure that’s why my placement matches on the first day were so much better than my placement matches for another account on the second day after everyone heard that PvP was easy way to get ascended armor.