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

Why DotA 2 MM System Fails - High Level Perspective

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    OP makes really good points. Good read.
    Karnivool's new album is DOPE!

    Comment


    • #17
      Originally posted by abc0815 View Post
      But its not the reality. Ontop he makes the same complains on the LoL forum while praising Dota2. Textbook Hypocrite
      Nope, I praise Dota's gameplay elements but hate what Valve did to it. LoL needs help in the gameplay department but Riot knows how to do everything else.

      Imagine that, life is more complicated than "Game A rules, Game B sucks!".

      Comment


      • #18
        Originally posted by Shaella View Post
        Welcome to pub games. They're pubs.
        It's not that we are trying to make poop smell like flowers.. This kind of statements you make does not bring anything new. 2 years people complain about MM, some of them crap some of them valid, how many more years we have to wait it to be fixed properly?
        Pub will be pubs i know but still it has to be more balanced, not like this!! For now i play this game just for experimenting, it's no fun without pre-made team. I should be playing this game SOLO.

        Comment


        • #19
          I totally agree with OP; the issue he described is also analogous to the issue with other skill levels of MM.

          There's no 'simple' solution to this issue, but two things came to my mind.

          In higher level games we could have pre-game lobbies where each of the teams would be in a chat lobby with a hero-picker and an extended hero-suggestion screen that would include individual players' preference for picks in this match (if any), their 'top performing heroes' and roles suggested by an algorithm and the current picks and lanes draft as planned by the team. Each team's lobby would be separate and invisible to the other team. The max time spent in pre-game lobby would be 60-90s, could be shortened by everyone pressing 'ready'. After that everyone is headed to the standard hero picker.
          The advantage of such a lobby would be establishing a sense of team unity and communication within the team before the match. It would also allow everyone to additionally draft their picks+lanes and see how would it work out, decreasing the amount of insane drafts. This feature could be combined with visible stats/ranking/ELO; as a player chooses his preferred pick his stats with that hero are being displayed etc.



          The other 'solution' to the issues of the current MM system is... well, a complete overhaul if I understand the current system properly. Looking at various posts from fletcher I've noticed that a player's skill is defined purely by one number, the MMR...
          I'm a tad of a perfectionist when it comes to such things, but I'd scrap MMR and implement Match Making Profiling ie. each player would have a profile to be rated by, not a single value. A profile would consist of multiple values that determine the player's skill in specific performance-related game aspects. 'Farm rate' would determine a players consistency in last-hitting when playing a hero that should be doing so, 'hero-role execution' would determine a player's ability to do what he's supposed to do with the selected hero (carry not dying early game, supports warding/roaming/ganking, OD basically winning lane every time etc.), ability to offlane could even be a boolean value that when true enables a new set of ratings determining player's effectiveness in off-laning.

          I understand that such a system would be really hard to create and implement as it would have to both collect and process a lot of data, but it is by no means impossible. Such a profiling system would also allow for various 'challenges' to be implemented in the game such as 'get XX lasthits as a solo-offlane WR without dying'. These challenges would guide various players in learning the ways of DOTA beyond the tutorial basics. The profiling could reallly find it's uses in various other parts of DOTA2.

          Now, how would the MM be done with profiles instead of single values? Honestly I can't say what the best option would be. A MMR could be maintained simultaneously with the MMProfile or could even be derived from it instead of being what it is now, but I have no clue about the effectiveness of this.

          No matter how impossible the above idea may seem I will hold onto this statement: a player's skill would be better determined by profiling than by a single value.

          Comment


          • #20
            OP's suggestions are only applicable to like 0.5% of playerbase, if that. I think that Valve should just make a "Grandmaster league" and put people who know how to play the game in there, separate from 99% that will never attain skills for competitive. Dota 2 is a popular game, but out of 4'000'000 monthly players, how many thousands actually can play in high level in-house games?

            Comment


            • #21
              Originally posted by allbusiness View Post
              Here is exactly why your DotA 2 MM system fails Valve, it's pretty simple. This is coming from someone who plays on the first 10 pages regularly against all sorts of nonsensical players like Clairvoyance, Merlini, Mikey, Blitz, LGD.Int players on USW, various other DotA pros/semi-pros/personalities.


              The Dota 2 MM system is built around the fact that players will be playing their hardest, and as such are playing optimally. If both said teams are playing optimally, then you're going to have an even 50% chance for both teams to win. Here's where the MM system fails; it's not really the MM systems fault at all. The reason why you have such one sided games, is predominantly because of the fact that weaker players attempt to take high impact roles, forcing either bad team compositions, or forcing a stronger player into playing a role that isn't as important in public games.

              First and foremost, let me preface that supports are extremely important in high level/competitive games. At the IHL level, if you don't have good supports, your team will basically get utterly crushed due to lack of wards, dynamic movement for ganks, TP support, pooling of regen, etc. However, supports are only important if the high impact roles of mid/carry/offlaner are played properly and optimally. These roles are extremely important to be played properly, because should any of these roles be played bad, it can be cascaded into a very bad game. As such, in most pub games of League, HoN, etc. the higher rated players tend to take the high impact roles, especially if there is a big discrepancy in ratings. The reason why is because these roles require strong mechanics, especially the mid lane. In pub games, your mid/carry has to be strong. if he is not and gets fed on, it is extremely easy for you to just lose the game and get snowballed on. This is exactly why various players have complained about bad matches; lower skilled players will take high impact positions, then lose when the MM system is expecting them to play optimally. You aren't playing optimally though if you are sending weaker players out onto mechanically demanding and high skill positions.

              Mid is pretty much the easiest example of what I'm talking about. If you have a discrepancy in skill levels between the two mid players, it's going to be real hard for the weaker mid to do anything, and as such he will likely not only lose the lane, but he will also make multiple mistakes and get fed on. The stronger player now not only has an immense experience advantage, he also has complete control of the runes. Even if you deny him rune access, he still has 2-3 levels on the side lanes, and as such can still easily gank and control the map, choking the opposing team from gold/exp. This is especially true when the weaker mid is losing a match-up he should win, for example he's losing to Pudge as Silencer; if he dies to Pudge (which he shouldn't at all), he is basically going to get fed on completely the entire time. This in turns allows Pudge a quick level 7 + rune control, which means he's free to run around and camp other lanes now.


              The biggest offense is when you have a dual que paired up with a 3 stack; now you need to figure out who is going where, but there's going to obviously be arguments because one player wants to play one hero, and the other player wants this hero. So now you're gonna run into a situation where you likely have optimal picks, but do not have the correct assignments due to lack of knowing who really should be playing on what position. The MM system did the correct thing; it paired up the correct players against each other. However, the players are assigning themselves sub-optimal positions most of the time, resulting in the crappy MM games you have currently.
              Hey man, think you nailed it, for example I had a low lvl, low win (300 wins) amount player take mid as opposed to 700 + wins (going bot).
              The mid guy totally believed that he could stand a chance, but in the end, did no ganking, no map awareness.

              Comment


              • #22
                Originally posted by ArchieChoke View Post
                Hey man, think you nailed it, for example I had a low lvl, low win (300 wins) amount player take mid as opposed to 700 + wins (going bot).
                The mid guy totally believed that he could stand a chance, but in the end, did no ganking, no map awareness.
                please, total amount of wins do not tell a persons skill level.

                The new match making is vastly better than the old one. This is information that valve can use, as well as what OP writes, that some times team composition in pub games as well as playing unusual roles can lead to stomps.

                After the patch though stomps no longer happen consistantly because of the skill, but because of the reasons stated by OP.

                Comment


                • #23
                  Originally posted by Blazingfist View Post
                  please, total amount of wins do not tell a persons skill level.

                  The new match making is vastly better than the old one. This is information that valve can use, as well as what OP writes, that some times team composition in pub games as well as playing unusual roles can lead to stomps.

                  After the patch though stomps no longer happen consistantly because of the skill, but because of the reasons stated by OP.

                  And it will never stop because Valve has literally everyone in one que. That's an issue, and Valve needs to figure it out somehow. League has done a pretty good job at separating most of the casuals and the competitive/serious players from each other through things like Normals/Dominion/Ranked. There's no reason Valve cannot do the same.

                  Comment


                  • #24
                    Originally posted by Propkid View Post
                    I totally agree with OP; the issue he described is also analogous to the issue with other skill levels of MM.

                    There's no 'simple' solution to this issue, but two things came to my mind.

                    In higher level games we could have pre-game lobbies where each of the teams would be in a chat lobby with a hero-picker and an extended hero-suggestion screen that would include individual players' preference for picks in this match (if any), their 'top performing heroes' and roles suggested by an algorithm and the current picks and lanes draft as planned by the team. Each team's lobby would be separate and invisible to the other team. The max time spent in pre-game lobby would be 60-90s, could be shortened by everyone pressing 'ready'. After that everyone is headed to the standard hero picker.
                    The advantage of such a lobby would be establishing a sense of team unity and communication within the team before the match. It would also allow everyone to additionally draft their picks+lanes and see how would it work out, decreasing the amount of insane drafts. This feature could be combined with visible stats/ranking/ELO; as a player chooses his preferred pick his stats with that hero are being displayed etc.



                    The other 'solution' to the issues of the current MM system is... well, a complete overhaul if I understand the current system properly. Looking at various posts from fletcher I've noticed that a player's skill is defined purely by one number, the MMR...
                    I'm a tad of a perfectionist when it comes to such things, but I'd scrap MMR and implement Match Making Profiling ie. each player would have a profile to be rated by, not a single value. A profile would consist of multiple values that determine the player's skill in specific performance-related game aspects. 'Farm rate' would determine a players consistency in last-hitting when playing a hero that should be doing so, 'hero-role execution' would determine a player's ability to do what he's supposed to do with the selected hero (carry not dying early game, supports warding/roaming/ganking, OD basically winning lane every time etc.), ability to offlane could even be a boolean value that when true enables a new set of ratings determining player's effectiveness in off-laning.

                    I understand that such a system would be really hard to create and implement as it would have to both collect and process a lot of data, but it is by no means impossible. Such a profiling system would also allow for various 'challenges' to be implemented in the game such as 'get XX lasthits as a solo-offlane WR without dying'. These challenges would guide various players in learning the ways of DOTA beyond the tutorial basics. The profiling could reallly find it's uses in various other parts of DOTA2.

                    Now, how would the MM be done with profiles instead of single values? Honestly I can't say what the best option would be. A MMR could be maintained simultaneously with the MMProfile or could even be derived from it instead of being what it is now, but I have no clue about the effectiveness of this.

                    No matter how impossible the above idea may seem I will hold onto this statement: a player's skill would be better determined by profiling than by a single value.
                    I agree with this.

                    Valve will have to collect more in-game information than they have to date in order for any of these suggestions to make sense. I'm not sure how an algorithm would evaluate in-game effectiveness for any given hero against any given team comp (and it will vary from team comp to team comp), but I'm sure Valve can figure out a way. It would be rather complex if done right, however.

                    Comment


                    • #25
                      Originally posted by allbusiness View Post
                      Here is exactly why your DotA 2 MM system fails Valve, it's pretty simple. This is coming from someone who plays on the first 10 pages regularly against all sorts of nonsensical players like Clairvoyance, Merlini, Mikey, Blitz, LGD.Int players on USW, various other DotA pros/semi-pros/personalities.


                      The Dota 2 MM system is built around the fact that players will be playing their hardest, and as such are playing optimally. If both said teams are playing optimally, then you're going to have an even 50% chance for both teams to win. Here's where the MM system fails; it's not really the MM systems fault at all. The reason why you have such one sided games, is predominantly because of the fact that weaker players attempt to take high impact roles, forcing either bad team compositions, or forcing a stronger player into playing a role that isn't as important in public games.

                      First and foremost, let me preface that supports are extremely important in high level/competitive games. At the IHL level, if you don't have good supports, your team will basically get utterly crushed due to lack of wards, dynamic movement for ganks, TP support, pooling of regen, etc. However, supports are only important if the high impact roles of mid/carry/offlaner are played properly and optimally. These roles are extremely important to be played properly, because should any of these roles be played bad, it can be cascaded into a very bad game. As such, in most pub games of League, HoN, etc. the higher rated players tend to take the high impact roles, especially if there is a big discrepancy in ratings. The reason why is because these roles require strong mechanics, especially the mid lane. In pub games, your mid/carry has to be strong. if he is not and gets fed on, it is extremely easy for you to just lose the game and get snowballed on. This is exactly why various players have complained about bad matches; lower skilled players will take high impact positions, then lose when the MM system is expecting them to play optimally. You aren't playing optimally though if you are sending weaker players out onto mechanically demanding and high skill positions.

                      Mid is pretty much the easiest example of what I'm talking about. If you have a discrepancy in skill levels between the two mid players, it's going to be real hard for the weaker mid to do anything, and as such he will likely not only lose the lane, but he will also make multiple mistakes and get fed on. The stronger player now not only has an immense experience advantage, he also has complete control of the runes. Even if you deny him rune access, he still has 2-3 levels on the side lanes, and as such can still easily gank and control the map, choking the opposing team from gold/exp. This is especially true when the weaker mid is losing a match-up he should win, for example he's losing to Pudge as Silencer; if he dies to Pudge (which he shouldn't at all), he is basically going to get fed on completely the entire time. This in turns allows Pudge a quick level 7 + rune control, which means he's free to run around and camp other lanes now.


                      The biggest offense is when you have a dual que paired up with a 3 stack; now you need to figure out who is going where, but there's going to obviously be arguments because one player wants to play one hero, and the other player wants this hero. So now you're gonna run into a situation where you likely have optimal picks, but do not have the correct assignments due to lack of knowing who really should be playing on what position. The MM system did the correct thing; it paired up the correct players against each other. However, the players are assigning themselves sub-optimal positions most of the time, resulting in the crappy MM games you have currently.
                      100% correct.
                      finally someone who knows what he's talking about!
                      one-sided matches are usually caused by this, and NOT by a failure of the matchmaker.

                      Comment


                      • #26
                        Originally posted by Dreams-Visions View Post
                        I agree with this.

                        Valve will have to collect more in-game information than they have to date in order for any of these suggestions to make sense. I'm not sure how an algorithm would evaluate in-game effectiveness for any given hero against any given team comp (and it will vary from team comp to team comp), but I'm sure Valve can figure out a way. It would be rather complex if done right, however.
                        Yeah, there's just too many variables in DOTA to easily say how such algorithms would work. For the case you mentioned (measuring performance of hero vs opponent picks if I understood right): the overall performance rating wouldn't have to consider enemy picks too much as knowing when to pick the hero or not (counters, synergies etc.) is a part of knowing how to play the hero.

                        Comment


                        • #27
                          The fact that I get matched people with 900 more matches then me, I would call the matchmaker a failure.
                          Even had a player playing his first game on my team...

                          Comment


                          • #28
                            Originally posted by Phoenixclaw View Post
                            100% correct.
                            finally someone who knows what he's talking about!
                            one-sided matches are usually caused by this, and NOT by a failure of the matchmaker.
                            Spacecowpoop.
                            You cannot tell me that im on the same level as a person who doesnt know howto use armlet. I refuse to believe that.

                            the mm system is broken bcause it pairs noobs with good players to get a avg. for a team, thats where it fails completely.

                            Comment

                            Working...
                            X