Results 1 to 10 of 1488

Thread: General Matchmaking Feedback Thread [READ OP]

Threaded View

Previous Post Previous Post   Next Post Next Post
  1. #1
    Volunteer Moderator slowreflex's Avatar
    Join Date
    Dec 2011
    Location
    United Kingdom
    Posts
    3,065

    General Matchmaking Feedback Thread [READ OP]

    After some feedback from the devs, we are attempting to make feedback more useful for them. Please provide your feedback regarding matchmaking in this thread only. We only appreciate precise, objective, constructive, and to-the-point feedback. Simply saying, "I like/hate matchmaking" does not help the devs to improve it. When posting please post in the following format:

    LIKES

    - I like x aspect of matchmaking.
    <some comments if you have further suggestions to improve this aspect> (OPTIONAL)
    - I like y aspect of matchmaking.
    <some comments if you have further suggestions to improve this aspect> (OPTIONAL)
    - and so on

    DISLIKES

    - I dislike x aspect of matchmaking.
    Explain why you do not like it. Give examples (with matchID if possible).
    Finally, you must tell us what we can do to improve this aspect. (MANDATORY)
    - I dislike y aspect of matchmaking.
    Explain why you do not like it. Give examples (with matchID if possible).
    Finally, you must tell us what we can do to improve this aspect. (MANDATORY)
    - and so on

    If you do not follow the proper format in this thread (esp giving your reasons) or post off topic, your post(s) will be deleted. Do not reply to others posts in this thread. If you keep posting again without following the proper format, you may have your posting privileges revoked.


    General Information about Matchmaking (note that the quotes link to fruitful topics that you might want to read too)

    Vlog by Purge summarizing many aspects of the current matchmaking

    Quote Originally Posted by fletcher View Post
    we use a Elo-type system. We have data that shows that it correlates well with a number of quality metrics.
    Quote Originally Posted by EricTams View Post
    As far as Matchmaking is concerned none of those buckets [low, medium, high] exist - We match on a continuous scale.
    Quote Originally Posted by EricTams View Post
    the easiest way to think about it is as a percentile rank so 100 levels.

    There is a maximum range that we match to that decreases based on your party size. This is why when you have a stack of 5 you end up taking longer to match since we can only match you to a few percent of the population.
    Quote Originally Posted by EricTams View Post
    Typically the trade off in matchmaking is wait time versus match quality. We could always try and slow down the match time in an attempt to provide higher accuracy matches but this will get better naturally as we add more players so I'd rather keep match times snappy.
    Quote Originally Posted by EricTams View Post
    - High skill games are definitely an area that will benefit as we get more players. It's also an area has had issues with match times in the past so I'm sensitive to how long this group needs to wait.

    - There can be no guarantee of how a game will play out even with perfect matchmaking. Hopefully someone can help me out with the details of that recent tourney where the winner of the winners bracket was in the finals and got destroyed in one game then destroyed the other team in the next. I think it's actually a good thing that the outcome of a match isn't set from the beginning but depends on the decisions that individuals are making during the course of play.

    - Building on the previous point, even with good matchmaking there are plenty of things that can go wrong:
    * Being conterpicked or having a bad team comp (overdependence on farm, no heroes that can stop a push, lack of DPS, bad teamfight, etc.)
    * Allowing teammates to lose a lane early
    * Luck
    * Natural variations in individual performance
    * People trying a new hero or strategy

    - We don't enforce a 50-50 win rate (Maybe we should?) instead we always try to make the best match possible. I'm always happy when I see someone with a win rate close to 50% though since it means that we haven't been skewing one way or the other in giving matches that were too hard or too easy. I totally agree with your desire to have fewer stomps though.

    - We don't try to match high skill players with low skill players instead we try to get 10 people that fit together the best. Parties can sometimes create a very big gap in skill on a team though and are generally the reason for anything that seems really wacky. Remember too that people people will always have good games and bad games so that person that seems terrible might just be having an off day instead.

    - The outcome of a games is determined more by the best person on a team rather than the worst player. My assumption was originally the same as yours but I had to change my mind once I saw the stats showing otherwise.
    Battle Points and Battle Level are not used in matchmaking.

    Performance Bars are not used in matchmaking.


    Q: What makes a "high quality" match?
    Quote Originally Posted by blash365 View Post
    a perfect match in a perfect world is where everybody has the same (correctly measured skill).
    a "high quality" match tries to get as close as possible to that result.
    Quote Originally Posted by fletcher View Post
    Consider a 4-stack with a big skill differential between the players. What we try to do is match them such that the teams are as equal as possible, player-by-player. (Sort the players on each team by Elo, and then examine the difference in Elo of corresponding opponents. Ideally the best playes on each team are evenly matched, the 2nd best are evenly matched, etc.) So that is pretty straightforward for figuring out the ideal thing to put on the other team, and notice that average Elo is not necessary. But what sort of a party should we use to fill the 5th slot as his ally? Would it be a good idea to find a single guy who matches the lowest skilled player on the 4-stack? The highest?How would that guy appraise the match we made for him, being the best player in the game playing with a bunch of players far worse than he is, or the worst player, with players much better than he is? What if we cannot find a 4 stack that matches him well? Would it be a good idea instead to take two 2-stacks, one that match the top two players, and one that matches the bottom two? What is going to be their reaction to being put in such a match?

    What would you say is the optimal way to match a stack with a large spread in skill?


    Q: Do you add the percentile points for each team member and swap players until both teams are even, or something more complex?
    Quote Originally Posted by EricTams View Post
    We have two stages to create a game from Matchmaking

    1) Try to find 10 players that all match well to each other

    2) Try to split these 10 players into two groups of 5 that will create a balanced game
    Q: Why can't i see my Matchmaking Rating?
    Quote Originally Posted by fletcher View Post
    Showing players their Elo has proven to cause a lot of unwanted behaviour. (The same problems exist in Chess.)

    We know that there really is value in knowing how the match was formed [...] It helps to analyze your own skill and try to understand how to get better. We are actively working to try to figure out what data we CAN provide to players [...] without causing too many of the undesirable effects
    Quote Originally Posted by fletcher View Post
    Elo is a TERRIBLE way to give players a sense of "progress." Many (most?) people reach a plateau, and their Elo stabilizes. It is simply not mathematically possible for Elo to keep increasing in general for players indefinitely as more and more games are played.

    Given this reality, if players used Elo to measure "progress", we would constantly be reminding them that they are NOT making any. That would be really bad.


    Q: What about if someone who reported someone else, are they ever matched again?
    Q: Does matchmaking take players who only play one or two heroes into consideration?
    Quote Originally Posted by blash365 View Post
    No. At least not yet.
    Q: Does matchmaking take the type of hero a player likes to play (support, pusher, carry, random...) into consideration?
    Quote Originally Posted by blash365
    No. At least not yet. But there are threads about it: [Suggestion] Pick role before sign up for a game!
    Q. Does the commend/report system affect ones percentile?
    Quote Originally Posted by blash365 View Post
    No.
    General advice:
    Quote Originally Posted by blash365
    the longer you search for a match, the broader the search range gets. unbalanced matches might be results of a broad search range. either change to a better populated server (in case you are playing at odd hours) or simply reset the search to restart at a finer search range.
    Last edited by blash365; 07-19-2013 at 08:19 AM.
    Please read Common Discussions, Common Suggestions, Common Technical Issues, Common Bugs, and do a website search prior to making any new thread.
    Use the report () button to report flaming, illegal discussions, offensive content, excessive spam, trolling, advertising, or duplicate threads

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •