General Discussion

General DiscussionSerious Question about MMR...

Serious Question about MMR... in General Discussion
Zenoth

    @ Marlan

    in the system you mentioned, high rated players would no longer be "handicapped" but instead get +5 for every win and -45 for every loss. Anyone on their teams wouldn't be too happy with that either.

    taking stats into account would open it to abusers heavily.

    case study: calibration - someone discovered hero damage is heavily weighted for mmr changes during calibration. Cue thousands of 5-6k mmr accounts suddenly on sale from abusers, some even going as far as to sandbox feeders on the other team to get 7k rating, ruining thousands of games in the process.

    Este comentário foi editado
    Ye Season

      1. Learn to fucking read for the 3rd time (yes I'm keeping count of how many times you prove yourself to be a dumbass)

      This is what I wrote:

      "ELO in Dota 1 was way more accurate, and would calibrate how many points you gained/lost based on who was in the game."

      There are TWO "count em" TWO points in that statement:

      1. ELO is more accurate
      2. It calibrates how many points you gained/lost based on who was in the game - Notice the "YOU" in that. It calibrated on an individual basis. But since you are already so much of an idiot, I'll just give you the benefit of the doubt and say it was ambiguous.

      2. Once again, it was more accurate. I don't know what your statement that "Dota 1 matchmaking systems were a general cesspool of filth" has anything to do with.

      3. See 2

      I see now that I just have to break things down so much that a fucking retard can understand it for you.

      Zenoth

        you failed to support any of your points

        your argument is "it was more accurate"

        how the fuck is that an argument, it is simply a contentious statement

        do you even english

        Este comentário foi editado
        Ye Season

          Uh, it worked differently because it did. That's just a plain fact. ELO gains took your individual score into account - not just your team's average.

          Water is wet.

          ^ Your dumbass needs support for that point too?

          It's more accurate because the MMR gain apparently only changes very slightly for any one point of differential between teams, while ELO gains were way more sensitive (e.g. accurate) to whatever differential may have existed.

          (I forgot I had to break shit down for you so much)

          Este comentário foi editado
          MARLAN

            I find Ursa not as effective in pub games.

            Meepo is a lot better. He's also like 70-80% right now. It only shows 60% because of when I was shit at him I was like 40-45% with him.

            Harjanaz

              i lost a game worth +/- 5 points! So i only lost 5 points it was pretty cool

              wraithseeker

                actually 4500 4500 4500 4500 4500 vs 3k 3k 3k 3k 6k has a much easier time than the 6k player really

                Trodlabundin

                  ^This

                  Also there isn't anyone counting up mmrs in end of game to figure how much mmr to give out...
                  Obviously it's a computer/bot, and those aren't wrong.

                  I'm on Valvebots>You

                  MARLAN

                    That's what I meant...

                    It's in favor of the 4.5k team even though valvo thinks it is balanced.

                    Trodlabundin

                      It is balanced in a way. The 6k player has a hard task carrying the 3k players, but if it wasn't like this, why would he be 6,5k?

                      He's 6,5k because he's good enough to do this.

                      He's far superiour compared to 4,5k thus he can punish their mistakes heavily.
                      However, matchmaking would be more of quality if there was a bounch of 6k playters that could play versus eachother, however it's supposed to be harder and harder as you climb. Thus, a 6,5k player having to carry vs bounch of 4,5k players is fair imho.

                      sleave

                        its unfair for the 4.5k players to lane against a 6.5k player

                        King of Low Prio

                          How cute, people still believe dota 2 lanes are static

                          sleave

                            yea because the 6.5k guy cant rotate and dominate other lanes

                            Julien

                              So I was playing yesterday.
                              In my team we were all around 5000-5500 rank.
                              In the enemy team there was Sneyking with 6800 rank and some random guys with 4500-4800 rank.
                              Do you think the game was even?
                              Yes it was. Quite close game and all got a change of 25.
                              http://www.dotabuff.com/matches/831729793
                              My point is, the way matchmaking handles very high ranked players and arranges the teams worked fine in that case. And, from my experience, it works fine in general.

                              eXcel

                                'while ELO gains were way more sensitive (e.g. accurate)' sensitive =/= accurate

                                Some times having a less sensitive reading will give you a more accurate response, particularly regarding something like MMR where there are so many factors that can influence a game, for example, I played a CM game today where our captain was our lowest ranked player and first picked tinker, no one on the team could play tinker so I was forced to play it, we lost the game and I was highest MMR on the map, I lost 24 points, did I deserve the loss? not particularly, tinker is a hero I have no interest in playing and I was forced onto it. With your 'sensitive' argument I should have lost a lot more than 24 points (being the highest rated) but the loss was because our 'captain' drafted badly for our players and didn't communicate. (he also rage quit during the game)

                                Este comentário foi editado
                                Quick maffs

                                  Well lets be honest here, if it ever comes to the situation of a 6.5k player lane against a 4k player IN MID the game is over man .....

                                  The 6.5k player sidelane needs to feed A LOT for him to lose the game.

                                  Guts

                                    Pick void mmr goes up and only up ^^.

                                    Ye Season

                                      That's one particular anecdote - that's not how statistics work. That loss is balanced out over time by "normal" games which should happen far more regularly on average (since those games are ALSO using a higher sensitivity) than your abnormal games (e.g. where someone ragequits etc).

                                      "Accuracy" here just means that every point difference is being taken into account, rather than just ranges (or that the ranges are smaller) - which is how MMR seems to work. Think of it like this: rating something on a scale from 1-100 is more accurate than rating something from 1-5. Saying that something is a 93/100 is more accurate than saying that same thing is a 5/5.

                                      Whether or not a particular win/loss was reflective of your actual skill is a different matter.