Enough with the legendary MMR?

Posts: 76
Seriously, when will we say enough to Blizzard with "MMR"? Is this some generic excuse nowadays for them to use when they mess up ladder? It seems so. It's like the Matrix. "Unfortunately, no one can be told what the Matrix is.". Replace it with "MMR" and there you go.
Reply Quote
Posts: 178
ello baby and like copy the chess brackets. only 200 elo range per bracket. than I can see if I am average or really suck.
Reply Quote
Posts: 1,665
Because then you get the inevitable "but my MMR is marginally higher than this guy's and he is in _______ league and I'm not" argument. It's just way simpler to say "you are there because I said so." There is a lot of math that goes into matching players and making those numbers public will only increase the amount of arguing over them.
Reply Quote
Posts: 1,144
MMR should stay because of how effective, relatively speaking, it is at managing a massive number of players. Elo ratings work great for medium or small groups and really MMR is a more complex Elo rating. Having played a lot of other RTS games online, SC2's MMR is the best match making I've found. Not perfect, but I have yet to see an actual solution outside of "just get rid of it!"
Reply Quote
Posts: 209
05/01/2014 07:13 PMPosted by BallsOfSteel
There is a lot of math that goes into matching players


You crack me up - there's no evidence of anything like that in the leagues, they're just filling up slots randomly. I know they say otherwise - they're lying.
Reply Quote
Posts: 890
05/04/2014 11:26 PMPosted by Dracul
05/01/2014 07:13 PMPosted by BallsOfSteel
There is a lot of math that goes into matching players


You crack me up - there's no evidence of anything like that in the leagues, they're just filling up slots randomly. I know they say otherwise - they're lying.


Translation: "I choose to reject your conjecture and replace it with my own conjecture!"
Reply Quote
Posts: 396
05/04/2014 11:26 PMPosted by Dracul
05/01/2014 07:13 PMPosted by BallsOfSteel
There is a lot of math that goes into matching players


You crack me up - there's no evidence of anything like that in the leagues, they're just filling up slots randomly. I know they say otherwise - they're lying.

even randomly filled slots would require a lot of math...
Reply Quote
Posts: 742
see the problem already?

already some of you are arguing over it. Already some of you are accusing blizzard of lying.

So given that - why in god's name would you ever think that exposing the MMR would help things, when no matter what Blizzard does, you'll just call them a liar?

Might as well not do anything then. If I was them, well then just forget it, if you have that kind of attitude.

Remember - 100 people shake your hand at the door, but you will always remember the guy who instead, slapped you in the face.
Reply Quote
Posts: 178
05/01/2014 07:42 PMPosted by CaptRaynor
MMR should stay because of how effective, relatively speaking, it is at managing a massive number of players. Elo ratings work great for medium or small groups and really MMR is a more complex Elo rating. Having played a lot of other RTS games online, SC2's MMR is the best match making I've found. Not perfect, but I have yet to see an actual solution outside of "just get rid of it!"
so far I not seen it. I feel elo is better until I see the mmr system. league fill by rank and not mmr it feels like. most game feel like luck. I constant feel Like match with some with 200 more elo than me in same league. if one league above it more like 400 to 800 elo difference.
Edited by spark on 5/6/2014 8:53 AM PDT
Reply Quote
MVP - StarCraft
Posts: 2,338
This thread is predicated on MMR being some arcane, unknown mechanism when it's actually quite simple and well-understood. It's functionally the same thing as Elo, but with leagues added on top, and leagues very simply encompass fixed rating ranges. Most leagues are about 250-300 rating wide, and Diamond's a little wider.
Reply Quote
Posts: 396
05/06/2014 02:01 PMPosted by ExcaliburZ
Most leagues are about 250-300 rating wide, and Diamond's a little wider.

Are these numbers connected to the points a player earns for winning? i.e. average game winning about 10-14 points means 18-30 games more won than lost up to next level?
Reply Quote
MVP - StarCraft
Posts: 2,338
05/09/2014 12:15 AMPosted by Plasma
05/06/2014 02:01 PMPosted by ExcaliburZ
Most leagues are about 250-300 rating wide, and Diamond's a little wider.

Are these numbers connected to the points a player earns for winning? i.e. average game winning about 10-14 points means 18-30 games more won than lost up to next level?


Noooot exactly. I'll give you a rundown. If you have the same MMR as your opponent, that game is worth 16 rating for both players (+16 for win, -16 for loss). If you have the same MMR and the same amount of adjusted points (points excluding spent bonus pool), that game is worth 12 points for both players. So, even if everything is equal, there's a minor difference between what you'll earn in points versus rating. Now, beyond a certain your-points-to-opponent-rating difference, a game will be worth +13/-11 points for you, or +14/-10, or +15/-9. Those breakpoints for each step aren't linear. You can see what they are here: http://postimg.org/image/8bny2acav/

The point system is built on top of the rating system, which means Blizzard has more flexibility to do whatever they want with points. For example, each league (except Master and GM) has a certain MMR floor for purposes of earning points. Just to use quick made-up numbers, if Bronze encompasses a range of 0-300 rating and you draw an opponent with 50 or less rating, their rating for the purposes of point calculation would be 50. MMR can still go below that floor. I believe we last estimated it to be between 73 and 88 for each sub-Master league. This means that if your MMR is below the floor and you're playing against other below-the-floor players, both of you will earn +14/-10 points, which prevents your points from stagnating at zero. Before the floor was added, there used to be a big problem in low Bronze where players could earn +1 point for a win, win 24 times in a row, then lose -24 for their first loss.

In terms of knowing how close you are to promotion, you can still kind of guess based on your adjusted points. Even though the scales are different between rating and points, the fact remains that points chase rating, so they're not completely worthless. If we determine that a league's rating range is 300, then you're going to be pretty hard-pressed to find a player on SC2Ranks/Nios who has more than 300 adjusted points in that league (because they would have been promoted already). You'll hit +300 rating slightly faster than 300 adjusted points, but you can keep in the back of your mind that the closer you get to 300 adjusted points, the closer you are to promotion.

Sort of a disclaimer on that last point: this all assumes you haven't experienced MMR decay. MMR decay kicks in at 2 weeks/14 days/336 consecutive hours without playing a game mode. It starts at 0 on Day 14 and ramps up by -22.5 rating per day (-0.9375 per hour) linearly up to a cap of -315 at Day 28. Naturally even if you had 180 adjusted points and +240 rating, you could no longer rely on that 180 as a guide if you idled for a month and decayed by -315, bringing you to 180 adjusted points but -75 rating.
Reply Quote
Posts: 86
Hmm, maybe just publishing everyone's individual MMR rating and comparing it to good players, bad players, and average players per league. They do it for some standardized testing in schools that tells the student if they did above or below average, and in sports like baseball. Why not for SC? It will keep the average players from speculating of how it works.
Reply Quote
Posts: 396
05/09/2014 08:23 PMPosted by ExcaliburZ
05/09/2014 12:15 AMPosted by Plasma
...
Are these numbers connected to the points a player earns for winning? i.e. average game winning about 10-14 points means 18-30 games more won than lost up to next level?


Noooot exactly. I'll give you a rundown. If you have the same MMR as your opponent, that game is worth 16 rating for both players (+16 for win, -16 for loss). If you have the same MMR and the same amount of adjusted points (points excluding spent bonus pool), that game is worth 12 points for both players. So, even if everything is equal, there's a minor difference between what you'll earn in points versus rating. Now, beyond a certain your-points-to-opponent-rating difference, a game will be worth +13/-11 points for you, or +14/-10, or +15/-9. Those breakpoints for each step aren't linear. You can see what they are here: http://postimg.org/image/8bny2acav/

The point system is built on top of the rating system, which means Blizzard has more flexibility to do whatever they want with points. For example, each league (except Master and GM) has a certain MMR floor for purposes of earning points. Just to use quick made-up numbers, if Bronze encompasses a range of 0-300 rating and you draw an opponent with 50 or less rating, their rating for the purposes of point calculation would be 50. MMR can still go below that floor. I believe we last estimated it to be between 73 and 88 for each sub-Master league. This means that if your MMR is below the floor and you're playing against other below-the-floor players, both of you will earn +14/-10 points, which prevents your points from stagnating at zero. Before the floor was added, there used to be a big problem in low Bronze where players could earn +1 point for a win, win 24 times in a row, then lose -24 for their first loss.

In terms of knowing how close you are to promotion, you can still kind of guess based on your adjusted points. Even though the scales are different between rating and points, the fact remains that points chase rating, so they're not completely worthless. If we determine that a league's rating range is 300, then you're going to be pretty hard-pressed to find a player on SC2Ranks/Nios who has more than 300 adjusted points in that league (because they would have been promoted already). You'll hit +300 rating slightly faster than 300 adjusted points, but you can keep in the back of your mind that the closer you get to 300 adjusted points, the closer you are to promotion.

Sort of a disclaimer on that last point: this all assumes you haven't experienced MMR decay. MMR decay kicks in at 2 weeks/14 days/336 consecutive hours without playing a game mode. It starts at 0 on Day 14 and ramps up by -22.5 rating per day (-0.9375 per hour) linearly up to a cap of -315 at Day 28. Naturally even if you had 180 adjusted points and +240 rating, you could no longer rely on that 180 as a guide if you idled for a month and decayed by -315, bringing you to 180 adjusted points but -75 rating.

awesome, thx, like this
Reply Quote
Posts: 209
05/09/2014 08:23 PMPosted by ExcaliburZ
Sort of a disclaimer on that last point: this all assumes you haven't experienced MMR decay.


Or you're not playing 1v1 - for anything else the entire system fall apart. There's this wierd delusion in blizzards mind that a high ranked players 1v1 skill is completely lost when they play 4v4 , or as part of an arranged team (within a random game with more players as well). This makes it completely trivial for players to 'hide' their real mmr from the system and have themselves a fun time roflstomping noobs all night on the ladder. As long as they switch team-mates regularly they never have to risk their mmr growing.

Either that, or they just ignore mmr and use random numbers in team games - or first come first served. Certainly no evidence they attempt to balance team games.
Reply Quote
Posts: 2,960
05/16/2014 04:54 AMPosted by Dracul
05/09/2014 08:23 PMPosted by ExcaliburZ
Sort of a disclaimer on that last point: this all assumes you haven't experienced MMR decay.


Or you're not playing 1v1 - for anything else the entire system fall apart. There's this wierd delusion in blizzards mind that a high ranked players 1v1 skill is completely lost when they play 4v4 , or as part of an arranged team (within a random game with more players as well). This makes it completely trivial for players to 'hide' their real mmr from the system and have themselves a fun time roflstomping noobs all night on the ladder. As long as they switch team-mates regularly they never have to risk their mmr growing.

Either that, or they just ignore mmr and use random numbers in team games - or first come first served. Certainly no evidence they attempt to balance team games.


Or you just accept that it's plain not possible to balance around team games.

If I'm plat and everyone else on my team is bronze, does that equal out the fact that I'm faced off against the other team, which is all silvers?

Or I'm masters with all bronze partners, is it totally even if everyone else on the other side is gold?

See what I mean? How would you do it?

8 people, all of varying skill levels all queue up for a team game. How would you handle who goes on which side? Can you guarantee a game which will be a close one? You honestly can't.
Reply Quote

Please report any Code of Conduct violations, including:

Threats of violence. We take these seriously and will alert the proper authorities.

Posts containing personal information about other players. This includes physical addresses, e-mail addresses, phone numbers, and inappropriate photos and/or videos.

Harassing or discriminatory language. This will not be tolerated.

Forums Code of Conduct

Report Post # written by

Reason
Explain (256 characters max)

Reported!

[Close]