|
On August 12 2010 02:40 Excalibur_Z wrote:Show nested quote +On August 12 2010 02:31 Shadowed wrote:On August 12 2010 01:59 Excalibur_Z wrote:On August 11 2010 21:17 paralleluniverse wrote:The top 200 players are determined across divisions by comparing their relative rankings and skill, while meeting certain requirements, such as ensuring that they’re active. We can all see that the top 200 is NOT sorted by points, and different from the rankings shown at www.sc2ranks.com. For example Dayvie is ranked 49 in the official top 200, but has always been in the top 10 in terms of points. This shows that the ladder ranks that the game uses based on points is nonsense. Either whatever method was used to calculate this top 200 is correct, or ranking based on points is. They can't both be right. If points are not the optimal way to rank players, why is Blizzard using it to rank in the game? Why not use this new method to rank? Or make points converge to the results given by this new method? Basically, Blizzard is admitting their points system for ranking is wrong, making the ladder rankings in the game a charade. Yeah you're right. Their ranking has to be based on MMR (because it's clearly not points), but may be influenced by ladder activity as well (they mentioned that has an impact, but to what degree we're not sure). Points need to be the sole ranking factor, or at least something else that's equally transparent. Lists like this one only serve to create confusion. My guess would be that this linear ranking is one that uses MMR - sigma*3, but then that wouldn't explain the 7-1 guy at the bottom whose sigma would probably be enormous (unless sigma rapidly changes after each match, far more than we estimated). Points are also absolutely comparable across divisions because the players all share competition, for those who are saying they aren't. Well I'm glad someone pointed that out. But I doubt that InSTinK is on the list cause of his platinum team, they said it's listing people across brackets so in all likely hood it's the 2v2 diamond team he's on: http://sc2ranks.com/team/131796The mistake people keep making is they are looking at divisions as anything except a wrapper around your rank. It's the equivalent to taking say, diamonds of varying qualities and putting them into separate boxes based on quality. You haven't made any of the diamonds more or less valuable, you just isolated them from the rest. The only part that could make points an inaccurate measure is the bonus pool not being totaled up consistently across leagues/players. While that would do well to explain it, it doesn't make sense. Bashiok said yesterday that the rankings are based off 1v1. It doesn't follow that team games would have any impact when historically in War3 and WoW 2v2 and 3v3 have always had completely separate MMRs and ratings. It wouldn't be fair if I went 99-1 in 2v2 to start matching me against 3000 MMR players in 1v1 from the start because 2v2 says nothing about my solo performance.
I think he might be correct, heres some evidence that 2v2 might affect your 1v1 MMR,
I played this guy a while back (I'm diamond with, at the time, about 60 games played)
![[image loading]](http://i927.photobucket.com/albums/ad120/Chriamon/Screenshot2010-08-0212_39_14.jpg)
|
Honestly it doesn't take a genius to show that Blizzard's Top 200 is inaccurate. No matter how someone may defend it - putting players who have played less than 20 games on top of 10,000s of players who have played much more is just stupid.
Someone said in this thread that the inaccuracy was due to the fact that the game has only been out for 2 weeks. I think of the opposite - thanks to that fact that we have extreme cases of ranked players with less than 20 games which are most obvious evidence that the ranking is broken. We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking.
Blizzard really needs to do something now.
|
So sad that in every rank site I 'm 150-170 ranked @ Eu server, but today blizzard disappointed me....
http://eu.battle.net/sc2/en/blog/348087
Btw I'm 700 points @ diamond and the last @ top 200 was only 517(well he got high ratio coz we all know that 5-0 placement is izi,and u have huge pool like 250-300,it's so izi to get to 500 points even jump from platim into diamond need 150-200points discount,but anyway 74% ratio is much izier than get a 700+ one...)
|
On August 12 2010 09:29 mrdx wrote:
We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking.
Why?
|
Perhaps they don't factor in points from the bonus pool? There could be other factors too, including calculation of opponent strength, etc.
|
On August 12 2010 09:32 Pyrthas wrote:Show nested quote +On August 12 2010 09:29 mrdx wrote:
We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking. Why?
well for example~
if u played only 10 games, u win each game u get a vey high increase on ur % ratio~
but if u plaed 1000, that cost u play 100 games each to get the same ratio...that's why
|
On August 12 2010 09:32 Pyrthas wrote:Show nested quote +On August 12 2010 09:29 mrdx wrote:
We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking. Why?
If you read the thread, those who defended Blizzard's top 200 have used unverifiable factors like MMR and such - which are hard to argue against because it's "hidden".
The players with ridiculously small numbers of games played in this first ever 200 ranking list are a clear cut evidence that the ranking is flawed. In a few months' time when everyone has played hundreds of games or more, there won't be such extreme cases, and it will be nearly impossible to verify if the top 200 ranking is correct or not. People will just conveniently throw the big word 'MMR' to shut up everyone who raise any doubt.
|
On August 12 2010 10:13 mrdx wrote:Show nested quote +On August 12 2010 09:32 Pyrthas wrote:On August 12 2010 09:29 mrdx wrote:
We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking. Why? If you read the thread, those who defended Blizzard's top 200 have used unverifiable factors like MMR and such - which are hard to argue against because it's "hidden". The players with ridiculously small numbers of games played in this first ever 200 ranking list are a clear cut evidence that the ranking is flawed. In a few months' time when everyone has played hundreds of games or more, there won't be such extreme cases, and it will be nearly impossible to verify if the top 200 ranking is correct or not. People will just conveniently throw the big word 'MMR' to shut up everyone who raise any doubt.
Conveniently, MMR is the most likely explanation. Blizzard have always have access to the real MMR numbers, and nobody in their right mind thinks that displayed rating is actually more accurate than MMR rankings.
Small numbers of games don't mean much except that the confidence the system has in the MMR is lower. The only way your argument even makes sense is if skill is only impacted by games played and starts off at the same level for everyone at release.
|
On August 12 2010 10:13 mrdx wrote:Show nested quote +On August 12 2010 09:32 Pyrthas wrote:On August 12 2010 09:29 mrdx wrote:
We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking. Why? If you read the thread, those who defended Blizzard's top 200 have used unverifiable factors like MMR and such - which are hard to argue against because it's "hidden". The players with ridiculously small numbers of games played in this first ever 200 ranking list are a clear cut evidence that the ranking is flawed. In a few months' time when everyone has played hundreds of games or more, there won't be such extreme cases, and it will be nearly impossible to verify if the top 200 ranking is correct or not. People will just conveniently throw the big word 'MMR' to shut up everyone who raise any doubt.
Their point has been that the ranking system is designed so that it works pretty well when dealing with people with lots of games, and has some bizarre consequences when dealing with people with only a few games. That is, their point has been that this behavior is expected, and that the system, while imperfect, works well in the long run, after it's been running for a while.
I'm not saying I agree--I honestly don't know enough about the system to say--but testing systems in extreme cases is not always good engineering, especially when there is not much riding on getting absolutely every case perfectly correct. (I don't even know what a perfect ranking system would be--I imagine there's bound to be substantial disagreement there.)
|
InStink is ranked 200th in the US by Blizzard, his rec is 7-1, here's his matchlist record:
1st game: win against Mewtwo 496 point Diamond
2nd game: win against Toosneaky 590 point Diamond
3rd game: win against Drone 580 point Diamond
4th game: win against Mercurio 663 point Diamond
5th game: win against Tozar 791 point Diamond
6th game: win against Hezzerboy 481 point Diamond
7th game: loss against Idra 1009 point Diamond
8th game: win against Foo 720 point Diamond
|
Conveniently, MMR is the most likely explanation. Blizzard have always have access to the real MMR numbers, and nobody in their right mind thinks that displayed rating is actually more accurate than MMR rankings.
I disagree with seeing 'MMR rankings' as a somewhat superior, more accurate ranking system. Assuming that Blizzard is using MMR for the top 200 ranking - a decent player can create a new account to start off from scratch, play only 20 games and have a very good MMR (yet low confidence score) to sneak in. Which is fine - I can't care less about that except if Blizzard uses their top 200 ranking list to invite people to their tournaments and people abuse this to get in.
Personally I trust sc2ranks much more than the official one because it's based on something verifiable 
|
On August 12 2010 10:37 mrdx wrote:Show nested quote +Conveniently, MMR is the most likely explanation. Blizzard have always have access to the real MMR numbers, and nobody in their right mind thinks that displayed rating is actually more accurate than MMR rankings. I disagree with seeing 'MMR rankings' as a somewhat superior, more accurate ranking system. Assuming that Blizzard is using MMR for the top 200 ranking - a decent player can create a new account to start off from scratch, play only 20 games and have a very good MMR (yet low confidence score) to sneak in. Which is fine - I can't care less about that except if Blizzard uses their top 200 ranking list to invite people to their tournaments and people abuse this to get in. Personally I trust sc2ranks much more than the official one because it's based on something verifiable
A simple fix for Blizzard would be to limit the list to only those with more than 50 or 100 games.
|
On August 12 2010 10:37 mrdx wrote:Personally I trust sc2ranks much more than the official one because it's based on something verifiable
Only in the sense that we can check the points. What is up for discussion is whether the points are actually a reliable way, or a better way than MMR or whatever, to compare players, especially across divisions. Verifiability is only valuable if what we're verifying is actually important.
|
On August 11 2010 21:17 paralleluniverse wrote:
This shows that the ladder ranks that the game uses based on points is nonsense. Either whatever method was used to calculate this top 200 is correct, or ranking based on points is. They can't both be right.
If points are not the optimal way to rank players, why is Blizzard using it to rank in the game? Why not use this new method to rank? Or make points converge to the results given by this new method?
Basically, Blizzard is admitting their points system for ranking is wrong, making the ladder rankings in the game a charade.
So much wrong that I dont know where to start
1.) The ladder is a ladder. It is based on points. NO WHERE did it say "oh look here, rank 1 is most skilled and rank 100 is the least skilled". It is incredibly obvious that the ladder is based on points and points alone. Your just be very ignorant and ASSUMING that rank = skill.
2.) What do you consider to be the "optimal way" to rank players? How can we truley realize the skill of a player? We can't. Points seem fine. You people just take it way to seriously. If you really want to prove yourself, GO TO TOURNAMENTS, ladder means nothing.
|
On August 12 2010 09:29 mrdx wrote:
Honestly it doesn't take a genius to show that Blizzard's Top 200 is inaccurate. No matter how someone may defend it - putting players who have played less than 20 games on top of 10,000s of players who have played much more is just stupid.
Someone said in this thread that the inaccuracy was due to the fact that the game has only been out for 2 weeks. I think of the opposite - thanks to that fact that we have extreme cases of ranked players with less than 20 games which are most obvious evidence that the ranking is broken. We won't have this chance again because when everyone has hundred of games in their history, it will be harder to validate the ranking.
Blizzard really needs to do something now.
right and you know the formula of how these results were derived?
uh huh so shut it.
|
On August 12 2010 10:37 mrdx wrote:Show nested quote +Conveniently, MMR is the most likely explanation. Blizzard have always have access to the real MMR numbers, and nobody in their right mind thinks that displayed rating is actually more accurate than MMR rankings. I disagree with seeing 'MMR rankings' as a somewhat superior, more accurate ranking system. Assuming that Blizzard is using MMR for the top 200 ranking - a decent player can create a new account to start off from scratch, play only 20 games and have a very good MMR (yet low confidence score) to sneak in. Which is fine - I can't care less about that except if Blizzard uses their top 200 ranking list to invite people to their tournaments and people abuse this to get in. Personally I trust sc2ranks much more than the official one because it's based on something verifiable
You just said that the player is decent.
You cant break the top 200 in MMR without playing at the very least the top 400, and if you beat players from 200-400 100% of the time you probably are top 200.
It doesn't mean anything that sc2ranks is verifiable, its based on something that explicitly is not supposed to accurately represent a player's skill level.
|
You cannot compare cross divisions using points. You can only compare using match making rating. Only Blizzard knows the match making rating so only Blizzard can give us an accurate global ranking of the top players. So yes, in summary, ranking sites that use points are wrong and do not rank based on player skill but instead on an arbitrary value that is only weakly linked to skill.
|
On August 12 2010 10:33 Dionyseus wrote:
InStink is ranked 200th in the US by Blizzard, his rec is 7-1, here's his matchlist record:
1st game: win against Mewtwo 496 point Diamond
2nd game: win against Toosneaky 590 point Diamond
3rd game: win against Drone 580 point Diamond
4th game: win against Mercurio 663 point Diamond
5th game: win against Tozar 791 point Diamond
6th game: win against Hezzerboy 481 point Diamond
7th game: loss against Idra 1009 point Diamond
8th game: win against Foo 720 point Diamond
Although he haven't played a lot of games yet, those 8 opponents are quality opponents. So I can see why he's ranked so high on Blizzard's own rankings.
|
On August 11 2010 21:30 paralleluniverse wrote:Show nested quote +On August 11 2010 21:24 Hanno wrote:
it sounds like someone doesn't understand MMR I have a perfect understanding of MMR. If MMR gives the correct rank and points don't: then stop using points to rank and start using MMR. Alternatively, make points converge to MMR, so when several dozen games are played, they are essentially equal. Note that points in WoW do converge to MMR. But if this top 200 is ranked by MMR (it's probably some combination of points and MMR and possibly other factors), then they've shown that points don't converge to MMR, again making points worthless.
It's pretty clear you don't. The difference between the top 200 and sc2ranks is pretty minimal. They will likely start looking alike over time though as people settle into their ranks assuming no one improves or gets worse skillwise.
|
Its obvious to anyone (I hope) that the bonus pool alone makes displayed rating an inaccurate skill metric.
I could have told you that in beta, without waiting for Blizzard to show anything.
|
|
|
|
|
|
EPT
