Announcement

Collapse

Forum Rules

  • No flaming or derogatory remarks, directly or through insinuation.
  • No discussion, sharing or referencing illegal software such as hacks, keygen, cracks and pirated software.
  • No offensive contents, including but not limited to, racism, gore or pornography.
  • No excessive spam/meme, i.e. copious one liners in a short period of time, typing with all caps or posting meme responses (text/image).
  • No trolling, including but not limited to, flame incitation, user provocation or false information distribution.
  • No link spamming or signature advertisements for content not specific to Dota 2.
  • No Dota 2 key requests, sell, trade etc.
  • You may not create multiple accounts for any purpose, including ban evasion, unless expressly permitted by a moderator.

  • Please search before posting. One thread per issue. Do not create another thread if there is an existing one already.
  • Before posting anything, make sure you check out all sticky threads (e.g., this). Do not create new threads about closed ones.
  • It is extremely important that you post in correct forum section.

  • Balance discussion only in Misc.
  • All art related (such as hero model) feedbacks go to Art Feedback Forum.
  • All matchmaking feedback should go here: Matchmaking Feedback
  • All report/low priority issues should go here: Commend/Report/Ban Feedback
  • No specific workshop item feedback. These should go to workshop page of that item.
  • When posting in non-bugs section (such as this), use [Bugs], [Discussion] or [Suggestion] prefix in your thread name.



In case you object some action by a moderator, please contact him directly through PM and explain your concerns politely. If you are still unable to resolve the issue, contact an administrator. Do not drag these issues in public.



All rules are meant to augment common sense, please use them when not conflicted with aforementioned policies.
See more
See less

To Valve, Suggestions to make MMR better

Collapse
This topic is closed.
X
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • VVitcher
    replied
    Originally posted by StingBull View Post
    You must be new around here. Let me fill you in. You don't suggest anything here. Valve won't read it and any other creature that hangs around here will just tell you to suck it.
    +
    Still if many users come up with an official proposal for 'Improvement of dota 2' thing,valve may finally do some shit.

    Leave a comment:


  • Dugachki
    replied
    Originally posted by trilogique View Post
    Valve's MM does its job fairly well because people at their true MMR have ~50% winrate. That's because equally skilled teams have a 50/50 shot of winning. That's what the system is trying to do: put you in matches of overall equal skill. It isn't perfect and it never can be, but the easiest way to tell if MM is doing its job is if the majority of people have a ~50% winrate. Most people do because the system works.

    You belong where you're rated (so long as you have a large sample size of games). No one is lower or higher than they should be.
    Completely untrue in my opinion. 90% of my games, I am paired with noobs or I play against noobs. I can tell after 5 minutes if we are going to win or lose. Sometimes I can see my teammates and me are being so much better then the opposite team that the game is boring. We just run around and stomp them.

    After 5 minutes, I'm able to say:
    - this is one of those games where we will lose no matter what I do. And there were games where I had 25 4 16 score but we still lost.
    - or this is one of those games where I can pick anything, and do any build I want and we will still win.

    If you have that kind of matchmaking, having 50% win rate is the easiest thing in the world. Either your team is OP or theirs is.
    And that's exactly the matchmaking we have.

    Originally posted by FunkDrSpot View Post
    3. For players that abandon within the first 10-15 min, their team should not be penalized for MMR loss but that player should get hit for 3x the normal loss.

    4. If 1 or 2 players are playing above the level of their team, they shouldn't be penalized the full amount of MMR for the loss. Anywhere from half to a quarter of the full 25 mmr loss depending on HD, HH, TD, KDA and other factors.

    3. I agree, If player leaves within first, lets say, 20 minutes, you should not lose 25 mmr.
    -lets say there was a calculation based on his GPM, KDA and amount of time he spent in the game of how much his influence on the game was. If he leaves and you lose the game, your team will lose 10, 15 or 20 mmr each.

    4. I also agree. If you are having an amazing game, you gank, kill, farm, but that bottom lane is just feeding, why should you lose 25 points just like them?
    I agree this is very difficult for system to observe. But lets say we have millions of games on our disposal (and we do).
    We can see best, average and worst:
    GPM, LHD, KDA for every hero, every role and every outcome of the game.
    So if you, indeed played a great game and still lost, it can be because:
    -your team lost a close game, enemy team was slightly better.
    In this case all of your team loses 25 mmr. But how does the system know this? Well, score is close and both teams have similar xp and gold graphs, until they won a huge team fight and it all went downhill for your team.

    or


    - you lost the game because 1 or more players had a terrible game with score like 1 10 4. I would suggest that if your play was in top 30% of what is best with that hero, and there are players on your team that played in bottom 30% with their hero, then you should lose less mmr then them. (they should still lose 25)
    Last edited by Kryil-; 07-09-2015, 10:21 AM.

    Leave a comment:


  • StingBull
    replied
    You must be new around here. Let me fill you in. You don't suggest anything here. Valve won't read it and any other creature that hangs around here will just tell you to suck it.

    Leave a comment:


  • trilogique
    replied
    Originally posted by FunkDrSpot View Post
    MMR does mean nothing as far as determining the quality of the player but at the same time it's the system that is used for matchmaking so it affects a lot.

    If I was valve I would try to use all the metrics at my disposal. KDA, HD, TD, HH, Kill participation, supporting efforts, etc while trying to factor in support & carry roles.
    That's worthless because statistics are misleading. Imagine you're 20-0-15. You'd think, "Damn that guy is good." In reality, it might just mean that guy ran around focusing on kills instead of objectives. When they needed him in a team fight, he wasn't there because he didn't wanna hurt his KDA. What if he was awful at last hitting and farm optimization? What if he was just good at carrying, but he was bad at every role? What if his team fight ability was awful? What if someone just had a bad game? Is there an objective metric for being a good team mate? How exactly do you rank this guy and who is the judge of it? Can an automated process actually determine that? If not, who do we determine as the arbitrator of Dota skill and why do we trust him?

    There are a million different factors that need to be accounted for and all of them can be misleading. And then trying to give all of those factors a weight? Who decides that and again why do we trust him? It's just his opinion, not an objective analysis.

    Valve's MM does its job fairly well because people at their true MMR have ~50% winrate. That's because equally skilled teams have a 50/50 shot of winning. That's what the system is trying to do: put you in matches of overall equal skill. It isn't perfect and it never can be, but the easiest way to tell if MM is doing its job is if the majority of people have a ~50% winrate. Most people do because the system works.

    You belong where you're rated (so long as you have a large sample size of games). No one is lower or higher than they should be.

    Leave a comment:


  • Export
    replied
    double post

    Leave a comment:


  • Export
    replied
    Originally posted by FunkDrSpot View Post
    MMR does mean nothing as far as determining the quality of the player but at the same time it's the system that is used for matchmaking so it affects a lot.

    If I was valve I would try to use all the metrics at my disposal. KDA, HD, TD, HH, Kill participation, supporting efforts, etc while trying to factor in support & carry roles.
    How do you determine who is what role? And how should the KDA HD TD HH KP look like?
    Should a SF that buys a mek be penalized? Should a mid that buys wards be penalized?
    Should a support that sacrifices his life to safe a carry be penalized?
    Last edited by Export; 07-02-2015, 01:21 AM.

    Leave a comment:


  • rockonrob
    replied
    Originally posted by FunkDrSpot View Post
    1. TBD means they've gotten their 'unranked' mmr through playing unranked games where people generally play lower quality games. It means they're not used to ranked games being a team function, counterpicking, etc. Like the examples I gave of real games, of the 3 TBD players I recently had, all 3 fed. Of the 2 that played on the other team, those 2 fed. That's 5 for 5.

    3. Players that abandon within the first 10-15min do not cause their teammates to lose MMR for losing the game but instead lose 3 times the amount they would have lost. This is to avoid feeder quits, rage quits and mid or feed quits. They could also avoid exploits to this suggestion by disabling it for people the abandoner is partied with.

    4. There actually are multiple ways to measure contribution to the team in a much more meaningful way than just looking at wins/losses. I'm sure no system will be 100% perfect but there should be SOMETHING beyond just wins/losses.
    You MMR suggestions are very VERY skill dependent (in terms of MMR). For example, at ~4k, most TBD players aren't feeding, they do have some skill and aren't usually reasonable far off from where they belong (also note that as you go higher in MMR, it becomes easier to see where certain players "belong." It isn't a thing about skill either, most 3k players could tell the difference at 4.5k, than maybe a 4.5k could tell at 3k).

    Also I have been seeing less feed and leave players as well, most players have been trying for a while and then give up later on rather than at the start. I don't think this is due to any sort of punishment system either, just how the player base is maturing (slightly). I would say that at about 3k, where is where i would guess your skill level is at, you are going to see this a lot more because there are just so many more players at that level and all for different reasons (like they feed and leave). But I dont think punishing people who happen to DC 2x more than the usual is really ok, they are already punish with MMR and usually LPQ for having a bad net. Not to mention bugs from Valve's servers.
    Plus, just because someone RQ early does not mean it is their fault that you would have lost and sometimes it is even beneficial.

    As far as #4, w/l is theoretically perfect, i mean it takes into account all factors such as KDA, GPM, XPM, etc etc, and even weights them for you so you don't need to do any work. The only problem is it's accuracy at any given point as well as crazy variance in player skill. KDA and all those factors don't mean anything, all you have to do is win and whatever that takes. In such a complex system, the people are the best way to judge who did well and who did poorly, except that not all players are capable of honestly assessing skill, and very few are any good at accurately assessing skill. Plus it would be abused and yadda yadda yadda, it just won't ever happen. I do think MMR could use some work, certainly, but either Valve need to think about this a lot and try some new things or nothing will ever happen. There have been a few really good suggestions about this and they were discussed but they were never looked at so nothing changed.

    Leave a comment:


  • Redthirst
    replied
    Originally posted by FunkDrSpot View Post
    If I was valve I would try to use all the metrics at my disposal. KDA, HD, TD, HH, Kill participation, supporting efforts, etc while trying to factor in support & carry roles.
    All of those mean nothing. You can pick Zeus and just spam your ult on CD to get high HD, does it mean you're good? No. K/D/A only matters something for carries, and even for them it's not accurate. HH is useless because only a handful of supports can heal. "Supporting efforts" like wards placed can and will be abused.

    Leave a comment:


  • FunkDrSpot
    replied
    Originally posted by pent1um View Post
    You better than anyone should know that MMR means nothing. Saw shitty players at 4k and awesome players at 2k.

    Out of curiosity. How would you implement point 4?
    MMR does mean nothing as far as determining the quality of the player but at the same time it's the system that is used for matchmaking so it affects a lot.

    If I was valve I would try to use all the metrics at my disposal. KDA, HD, TD, HH, Kill participation, supporting efforts, etc while trying to factor in support & carry roles.

    Leave a comment:


  • FunkDrSpot
    replied
    Originally posted by katsaroulhs View Post
    1. TBD does NOT mean low MMR. It means that this player hasn't finished calibrating his account yet. Those TBD players might have higher MMR than you.
    2. This (and similar suggestions) has been discussed multiple times. I totally agree with it, but Valve has completely ignored it. They do not seem to care about this.
    3. What exactly do you mean by that?
    4. There is no way to accurately measure one's contribution. This would lead to several people losing more (or less) than they deserved.
    1. TBD means they've gotten their 'unranked' mmr through playing unranked games where people generally play lower quality games. It means they're not used to ranked games being a team function, counterpicking, etc. Like the examples I gave of real games, of the 3 TBD players I recently had, all 3 fed. Of the 2 that played on the other team, those 2 fed. That's 5 for 5.

    3. Players that abandon within the first 10-15min do not cause their teammates to lose MMR for losing the game but instead lose 3 times the amount they would have lost. This is to avoid feeder quits, rage quits and mid or feed quits. They could also avoid exploits to this suggestion by disabling it for people the abandoner is partied with.

    4. There actually are multiple ways to measure contribution to the team in a much more meaningful way than just looking at wins/losses. I'm sure no system will be 100% perfect but there should be SOMETHING beyond just wins/losses.

    Leave a comment:


  • pent1um
    replied
    You better than anyone should know that MMR means nothing. Saw shitty players at 4k and awesome players at 2k.

    Out of curiosity. How would you implement point 4?

    Leave a comment:


  • katsaroulhs
    replied
    1. TBD does NOT mean low MMR. It means that this player hasn't finished calibrating his account yet. Those TBD players might have higher MMR than you.
    2. This (and similar suggestions) has been discussed multiple times. I totally agree with it, but Valve has completely ignored it. They do not seem to care about this.
    3. What exactly do you mean by that?
    4. There is no way to accurately measure one's contribution. This would lead to several people losing more (or less) than they deserved.

    Leave a comment:


  • Aparat
    replied
    1. garbage
    2. garbage
    3. garbage
    4. garbage

    Leave a comment:


  • FunkDrSpot
    started a topic To Valve, Suggestions to make MMR better

    To Valve, Suggestions to make MMR better

    I thought I'd at least give it a try to have Valve hear out my suggestions, what with Reborn in beta and all. I've climbed from 1.4k MMR to 3.5k MMR on a 60% winrate (http://www.dotabuff.com/players/15335542) but I'm fresh off of losing 4 ranked games in my last 7 mostly due to:

    a) TBD players - http://www.dotabuff.com/matches/1586792363, http://www.dotabuff.com/matches/1597431641, http://www.dotabuff.com/matches/1594039460
    b) Foreigners that can't communicate - http://www.dotabuff.com/matches/1597431641
    c) Teammate abandons or foreigners with terrible connections - http://www.dotabuff.com/matches/1576064570, http://www.dotabuff.com/matches/1594039460
    d) 1-2 teammates feed - http://www.dotabuff.com/matches/1575765458, http://www.dotabuff.com/matches/1575471105, http://www.dotabuff.com/matches/1591708206, http://www.dotabuff.com/matches/1592255505


    So, to give some suggestions:

    1. This game needs to limit the MMR effect on established players playing with TBD players. 3 of my last 5 losses have been with TBD players that pick poorly and feed even worse.

    2. I like the theory that Valve lets us play in any region with any type of people BUT too many people abuse this function, cant speak the languages they selected so they get through the queue for a game faster, and have a terrible connection that has the whole team playing at a disadvantage. Valve should give an option to avoid playing with players outside of your region if you so choose.

    3. For players that abandon within the first 10-15 min, their team should not be penalized for MMR loss but that player should get hit for 3x the normal loss.

    4. If 1 or 2 players are playing above the level of their team, they shouldn't be penalized the full amount of MMR for the loss. Anywhere from half to a quarter of the full 25 mmr loss depending on HD, HH, TD, KDA and other factors.


    I understand full well that no system is perfect but DOTA 2's matchmaking could be so much better. I know a lot of players have too much time invested with this game to be unbiased so instead of all of us arguing, how about if you disagree with a suggestion, you post YOUR idea on what would work better?
Working...
X