|
On February 15 2012 03:35 [F_]aths wrote:![[image loading]](http://i.imgur.com/0L9Ag.jpg) In this posting I attempt to explain why it's not a good idea to display MMR in the player profiles, and why it is okay to match random teams versus arranged teams. The match making rating (MMR) was developed by Blizzard to match players so that the winning likelihood of either one is 50% (or close to 50%, if no exact match is found.) This has several implications: • MMR is not what many guys consider skill: The ability to get ranked high in tournaments. The MMR is blind to the causes of your ladder wins. Some players who have the MMR to get matched versus actual pros, will never win a tournament unless they are going pro, too. "Winning likelihood on ladder" does relate to "skill", but it does not translate 1:1. Displaying MMR would give many guys the false impression that they finally can compare their true skill. • MMR considers the winning likelihood relative to the current ladder pool. That means, a ranking of 1000 MMR points today relates to much higher skill than lets say four weeks after the release of the game. It's good that without the display of the MMR, no one can brag around with his peak MMR he could got at any point in the past. • Again, MMR was developed only to get you an opponent which has the same chance of winning the game as you have right now. When win or loss streaks occur, the match maker uses a mechanic to consider short-term slumps or performance explosions without too large impact on the MMR while it still tries to get you opponents close to your current winning likelihood. To do this, the match maker uses a confidence value is widened or narrowed depending on how close you fulfill the match maker's prediction of the outcome of the game. This mechanic uses arbitrary values which are intended to reflect common casual fluctuations of your performance. MMR is not your true skill. Because of this, the game can match random teams versus arranged teams. If you are a random team player and if you play enough games, your winning ratio will be around 50%. It doesn't matter if the opponent arranged team uses Skype. You are matched against them because your random team still has about 50% to beat the other team. If a team just began to use voice chat and therefore wins some games, their winning streak will be disrupteded when they get too strong opponents. Then this team will lose some games until the MMR reflects the correct winning likelihood. So yes, you can get an opponent which is stronger than his MMR reflects, but you have the same chance to get an opponent who is weaker than his MMR suggests. The MMR is constantly changing. The top-ranked arranged teams have a very high winning ratio (70% and higher.) This is because there are not enough teams of the same MMR online. Since the top arranged teams have a higher MMR compared to any other team online (regardless if arranged or random), very good arranged teams can sustain a ratio quite a bit above 50%. That also means, the team who is likely to lose (because of its lower MMR) will only lose few points while the arranged team, if it wins as predicted, gains very little points. While this is frustrating, it's still better than having no game at all. With separate AT / RT pools, the search time would be longer while the matching wouldn't be any closer. This issue only affects a very small ladder population, though. For any widely available MMR level, the shared AT / RT pools allows faster game search and closer winning likelihood matching. Recommended read: Excalibur_Z's ladder guide.
the problem with team games have never really been the Arrange team vs Random team issue ,though i do hate it because it makes for really bad games ( either you roll a team effortlessly or you get rolled effortlessly) but whats more of an issue is Partial Arrange teams competing in Random team games.
For instance , if you play a 4v4 Random team game and you are in party of yourself + 2 other people. That means 3 out of 4 people are Arranged, which means its a 75% arranged team and you are getting ranked within random team ladders. Blizzard has put in a system which slightly takes into consideration full blown arrange teams vs random teams to even up things a bit. But they have put absolutely no system in place for Partial arrange team abuse. Who cares for instance if one of your allys are really bad in 4v4 rt ( the randomly given ally), the Arranged 3 players can do very strong cheeses to eliminate none arranged team players out of the game quickly because they dont have the team coordination, they arent talking on vent, they arent saying their builds, its just a fucking mess.
Do me a favor and check all the masters random team profiles you come across and look specifically for the ones that are doing the best. From my experience the guys with the best random team records are the guys playing with the same friends all the time and accepting 1 randomly given ally.
Anyway, i dont really want to get into this anymore as i have said enough about it at the blizzard forums, there are better ways to find matches faster than to encourage arrange team vs random team. but for the very least blizzards needs to stop having "player parties" destroy random team games.
|
i disagree with the AT/RT thing. trust me, its so enfuriating to play vs arranged teams; its practically auto-loss. i dont understand what blizzard was thinking.
|
What are you all afraid of that will happen when displaying the mmr?
--------------------------------------------------------------------------------
Blizzard's reasoning for not showing the MMR is that it's noisy and doesn't change much for most players over the long run. Their concern is that it would be a disincentive for playing for most players.
Last edit: 2012-02-15 10:52:11
They are dead wrong here, and also aiming for the wrong group of players.
With this they are trying to please the casual non competitive gamer while at the same time displeasing the competitive gamer who loves to know his elo/mmr despite it beeing "low"
Sc is to difficult for casual gamers and they wont be able to please thoose people annyway for a bit longer time.
They should aim to go for the bit more serious player, (more serious does not imply higher skill btw!) and the bit more serious player who likes to be competitive would love to see his elo/mmr
Even if this would show that he would not improve.
If look at the chessworld, peole LOVE their elo.
This not only goes for thoose with high elos, or thoose who are raising their elo fast.
This also goes for all lower levelled players, they all love their elo and their elo not improving is no deterent to play.
Now chess is in a huge slump and less and less people play it but this is NOT due to the elo beeing displayed.
I agree that the elo in sc wil be verry noisy,much more then it is in chess (where it is verry rare to drop more then 100 points on a "loosing streak") but this should not be a deterent,
it just means that there is alot more variance in sc, though most people already noticed this.
i dont mind playing arranged teams btw, i play lots of random 4x4 myself (over 200 games a season) and run into arranged teams or partially aranged teams (where 3 people are from same team) quiet often and it never botherd me.
If the mmr is the same i still have 50% change to win,mmr dont lie
|
I want to ask a question concerning MMR. Are 1v1, 2v2, 3v3 and 4v4 MMR completely independant from one another?
Meaning, can someone be Master 1v1, but insta leave a few games in 2v2, be placed in Bronze in 2v2, and then get extremely easy opponents in 2v2 from then on until he wins a lot of games.
Or are all the modes dependent on 1v1 MMR?
|
On February 16 2012 01:55 THM wrote:
I want to ask a question concerning MMR. Are 1v1, 2v2, 3v3 and 4v4 MMR completely independant from one another?
When a player places for the first time in a game type, the other MMRs are used as an initial estimate. After that, they are independent.
|
it is/was no secret that displayed ranks only work with a reference to a given time.
are you a blizzard guy or where do you get these information?
it sure is a nice article but there is no evidence your assumptions are correct or incorret.
since the article has some kind of semi-scientific form, i was expecting some serious references to proove your point.
no serious reference: maybe just your personal point of view?
|
On February 16 2012 02:23 eighteen8 wrote:
it is/was no secret that displayed ranks only work with a reference to a given time.
are you a blizzard guy or where do you get these information?
it sure is a nice article but there is no evidence your assumptions are correct or incorret.
since the article has some kind of semi-scientific form, i was expecting some serious references to proove your point.
no serious reference: maybe just your personal point of view?
While the exact MMR formula isn't known, the principle behind it is known. Game designer Sirlin was approached by Blizzard, he blogged about it: http://www.sirlin.net/blog/2010/7/24/analyzing-starcraft-2s-ranking-system.html
Excalibur_Z from the TL community got some clarifications from Blizzard during a Blizzcon. With all things we know, it's reasonable to assume that Blizzard uses a derivation of the Glicko system, which is described: http://www.glicko.net/glicko.html
|
On February 16 2012 01:55 THM wrote:
I want to ask a question concerning MMR. Are 1v1, 2v2, 3v3 and 4v4 MMR completely independant from one another?
Meaning, can someone be Master 1v1, but insta leave a few games in 2v2, be placed in Bronze in 2v2, and then get extremely easy opponents in 2v2 from then on until he wins a lot of games.
Or are all the modes dependent on 1v1 MMR?
Each X versus X random team mode has it's own MMR. That means, your 2v2 RT MMR is independent of your 3v3 RT MMR. This is also true for each arranged team. If you play with friend A a 2v2 team and with friend B, too, each team has its own MMR.
You need to play 5 placement matches, and for those matches the match maker considers the MMR from other leagues, if available. (Otherwise the system assumes a default MMR.)
|
United States12240 Posts
On February 16 2012 01:48 Rassy wrote:
What are you all afraid of that will happen when displaying the mmr?
--------------------------------------------------------------------------------
Blizzard's reasoning for not showing the MMR is that it's noisy and doesn't change much for most players over the long run. Their concern is that it would be a disincentive for playing for most players.
Last edit: 2012-02-15 10:52:11
They are dead wrong here, and also aiming for the wrong group of players.
With this they are trying to please the casual non competitive gamer while at the same time displeasing the competitive gamer who loves to know his elo/mmr despite it beeing "low"
Sc is to difficult for casual gamers and they wont be able to please thoose people annyway for a bit longer time.
They should aim to go for the bit more serious player, (more serious does not imply higher skill btw!) and the bit more serious player who likes to be competitive would love to see his elo/mmr
Even if this would show that he would not improve.
If look at the chessworld, peole LOVE their elo.
This not only goes for thoose with high elos, or thoose who are raising their elo fast.
This also goes for all lower levelled players, they all love their elo and their elo not improving is no deterent to play.
Now chess is in a huge slump and less and less people play it but this is NOT due to the elo beeing displayed.
I agree that the elo in sc wil be verry noisy,much more then it is in chess (where it is verry rare to drop more then 100 points on a "loosing streak") but this should not be a deterent,
it just means that there is alot more variance in sc, though most people already noticed this.
i dont mind playing arranged teams btw, i play lots of random 4x4 myself (over 200 games a season) and run into arranged teams or partially aranged teams (where 3 people are from same team) quiet often and it never botherd me.
If the mmr is the same i still have 50% change to win,mmr dont lie
You're correct that they're aiming for the more casual player, and Blizzard can afford to do this because they know that the more competitive players will play the game anyway and they can draw inferences just as some of us have done with the mechanics of the ladder system. If you understand how the system matches, what the division tiers represent, how to read opponent match histories, and how to read point gains and losses, you can get a pretty good idea via looking at trends where you stand as a percentage of the greater population. That's a far cry from your chess Elo rating which you know and everyone else knows because it's visible and easily accessible. However, if you put in the work to solve the riddle of your own MMR, you can get close enough to draft a reasonable conclusion. Typically, that's not an issue at all because competitive players -- even if they want to know their exact skill rating and use it for comparisons -- won't stop playing the game over it.
Casual players I'd imagine see the game like this: place in a league, gain points, try to be the division #1, level out somewhere, play only when bonus pool has accumulated, season roll, repeat. Their points basically only go up, but the better players' points go up faster. Still, all that's visible are points so that's what the casual player has to gauge skill. For about 87.5% of a season, their points are going up via consuming bonus pool which creates an illusion of improvement. Then the season rolls over and they can do it all over again while wondering why they haven't been promoted despite being #1 in their lowest-tier Bronze division. Unlike the competitive player who is more focused on actually being better than his opponents, the casual player wants to witness progress which shows that he is better. Therefore, while a displayed MMR doesn't really affect the competitive player other than satisfying his curiosity, a displayed MMR which practically never changes adversely affects the casual player who would quickly notice that he's not actually improving from season to season.
|
I don't think anyone wants to specifically see their MMR, they just want a ranking system that they can use to compare against any player.
Right now, a Gold and Platinum player have no real way to tell how much of a difference their ranks have. If there was a separate rank that gave a placement as related to every active player, I think a lot of people would be happy.
|
I'm not sure if this has been answered already but does the MMR take the matchup in consideration when matching two opponents together? Because I've been matched against pretty fucking good zergs lately and TvZ is my worst matchup so I lose them all. I'm doing fine in other matchups though. I doubt it does I guess?
|
It's amazing that after what must have been more than a dozen posts in the other thread by now, you still haven't realized that the issue most of us have with AT vs. RT is the horrible quality of games and the massive potential for abuse. Just like in the other thread, you completely ignore this abuse, you refuse to accept the statistical evidence (teams with 90%+ or even 100% win rates, overall MUCH higher AT win rates compared to RT), and you have yet to admit that just like the rest of us, you're in the dark as to how exactly the MMR works.
I realize I had initially promised not to reply to your posts any longer; I make an exeption since you've even gone so far as to open a new thread on this subject. That you have the nerve to repost this without acknowledging any of the points brought up before is ... bold, to stay diplomatic.
|
On February 17 2012 03:27 Shockk wrote:
It's amazing that after what must have been more than a dozen posts in the other thread by now, you still haven't realized that the issue most of us have with AT vs. RT is the horrible quality of games and the massive potential for abuse. Just like in the other thread, you completely ignore this abuse, you refuse to accept the statistical evidence (teams with 90%+ or even 100% win rates, overall MUCH higher AT win rates compared to RT), and you have yet to admit that just like the rest of us, you're in the dark as to how exactly the MMR works.
It is not necessary to know the exact MMR formula because the principle behind MMR is known. We know what MMR can do and what it can't do. Not all details are known, for example how the average team MMR is calculated from all random team members. The arithmetic average would be a bad idea. I assume that Blizzard found a way to get to a somewhat reliable average calculation. It isn't hard to check the expected winning likelihood with the actual one and then finetune the calculation of the team average MMR. But even if Blizzard's average MMR caluclation is flawed, it still doesn't matter since then you will lose so many games to have an random MMR so low that you no longer get matched versus ATs which are too strong.
If there aren't enough top teams online to match them, they are set versus lesser-skilled teams since this is the only way to allow for a game. Your statistical evidence covers only a very small part of the population. The last time I looked at sc2ranks, I saw that many master ATs had just slightly above 50% win rates (about 55% wn rate.) Some arranged teams however were able to sustain win ratios above 70% for umpteen games, but as expected those teams are mostly found in the top ranks. Those win rates are only possible because they play teams with lesser MMR. Mixing AT and RT is a separate issue.
There is a conjunction though, since all top teams are arranged, not random teams. But putting AT only versus AT solves nothing because it's the MMR which should be equal. When there are no other teams of the same strength online, the stronger team will likely win their match even if they play a weaker AT.
You see that I am actively avoiding a discussion about "horrible quality" of games because this is a matter of taste. How can separate team player pools improve the game quality anyway?
Getting a "good" experience is difficult within a random team. I guess many players participate in team games because they want a laid-back experience. In my experience, they are not really willing to cooperate, but they demand protection from the team when they are attacked. How can it improve game quality if you match RT only versus RT? To save them from being stomped by an AT? When teams with same MMR (regardless if AT or RT) are matched, the winning likelyhood is even. I cannot see the abuse here.
With random teams you can be lucky to get paired with a good player who carries the team, or unlucky since you get a scrub who doesn't talk until he is attacked, then he demands help from the team only to quit 30 seconds later after he publicly declared his team is noob. This luck factor per single game even increases when RTs are only paired versus RTs because both sides are unstable. If you match an RT of stronger individual players versus an AT of weaker individual players, compensated by better player cooperation, at least one side isn't that much luck-based.
Unless I made a big logical blunder I dare to say that luck-based one-sided game results are an issue of random teaming, not of putting them against ATs.
|
On February 15 2012 03:38 Gheed wrote:
Nobody cares if the MMR makes AT vs RT "even," they care about the horrible games it produces. Why do you insist on posting this everywhere? The idea that nobody would queue for arranged 2s if it had its own bracket is also farcical. It did not take that long to find a game in WC3. In fact, you could argue that more people would play if there was an AT ladder, because the competition would be something other than RT noobs. You could argue that using the same "facts," you are, i.e. none.
AT vs RT isn't that bad. It really makes no difference in 2v2 until the highest of levels where you'll never be evenly matched with the best of the best 2v2 teams, unless you and your random partner are actually that much better to bridge the "mmr" gap. Random teammates at the top levels are also much less prone to be retarded and carried to that level. You might walk into one every 10-15 queues but I simply don't see them. All of my games as random are fairly even and I've enjoyed a 60-70% win ratio three seasons straight (barring this one cause my mmr tanked out of the top).
Anything, ANYTHING below the highest of the high in the region, let alone diamond is simply muddied with too many other variables that disallow an accreditation to AT vs RT. It's highly likely the AT had better players than you and your random partner, not your random partner alone. MMR depicts your current trends relative to other players around your rating, not your relative skill.
Keep in mind, below the top levels of MMR, in lets say diamond, the AT may be using skype and might be winning more often because of their coordination. Their mechanics/strategies are not the best and they WILL eventually be placed against random players who ARE better, and they WILL break even at 50% UNTIL they start queuing against the highest MMR's in the region. Until then, all of your complaints about RT's are delusional. Also, this assumes everyone in an AT (in 3v3/4v4) is suddenly competent and AT's are never carrying a weaker player beyond his rating through coordination via skype.
edit: Also the AT's with seemingly massive win rates are simply not playing enough games to be placed within the right MMR. 2v2 is a much more fleshed out league mmr wise and teams with massive win rates will break even very fast if they aren't destined to hit the ceiling of masters. 3v3/4v4 there are a boat load of teams that play less than 20-30 games and go like, 20-2. They aren't playing enough games to be moved to their MMR. It could be possible a lot of these diamond+ teams should be placed within the highest of masters. Not enough matches are played from any one team, and 3v3/4v4 in general is very murky when it comes to what level of MMR depicts a team's skill; It's the exact opposite of 1v1 which has a very clean, rising trend of MMR levels.
|
|
|
|
|
|
EPT
