Matchmaking mathematics: current system is fair but should be refined

Gamma
1 posts Member
CG is right to stand by skill points based matchmaking, because it is demonstrably fair.

If two players have been hovering around the same skill points level, then the probabilities of either winning are by construction 50% / 50%, even if there’s a big GP discrepancy => matchmaking is fair.

What the community feedback has been about is that we not only want matchmaking with 50/50 odds of winning, we also want 50/50 odds of winning CONDITIONALLY on both players participating.

Consider the situation where I played all my GAC rounds and had 50% win/loss ratio at my current skill points, and my opponent had 50% win/loss at same skill points, because they won 100% of the rounds they played but only played 50% of them.

My odds of winning are indeed 50%: the matchmaking was fair.

But the outcome depends entirely on whether my opponent will show up or not: there is no game element.

This screams for a simple conceptual fix: continue to match players based on skill points, but also match them by the probability they play a round based on their history of 0 banner rounds.


Replies

  • Gamma wrote: »
    CG is right to stand by skill points based matchmaking, because it is demonstrably fair.

    If two players have been hovering around the same skill points level, then the probabilities of either winning are by construction 50% / 50%, even if there’s a big GP discrepancy => matchmaking is fair.

    What the community feedback has been about is that we not only want matchmaking with 50/50 odds of winning, we also want 50/50 odds of winning CONDITIONALLY on both players participating.

    Consider the situation where I played all my GAC rounds and had 50% win/loss ratio at my current skill points, and my opponent had 50% win/loss at same skill points, because they won 100% of the rounds they played but only played 50% of them.

    My odds of winning are indeed 50%: the matchmaking was fair.

    But the outcome depends entirely on whether my opponent will show up or not: there is no game element.

    This screams for a simple conceptual fix: continue to match players based on skill points, but also match them by the probability they play a round based on their history of 0 banner rounds.

    What’s the probability that CG will be able to code such a system successfully and bug free, do you think?
  • CCyrilS
    6669 posts Member
    Gamma wrote: »
    CG is right to stand by skill points based matchmaking, because it is demonstrably fair.

    If two players have been hovering around the same skill points level, then the probabilities of either winning are by construction 50% / 50%, even if there’s a big GP discrepancy => matchmaking is fair.

    What the community feedback has been about is that we not only want matchmaking with 50/50 odds of winning, we also want 50/50 odds of winning CONDITIONALLY on both players participating.

    Consider the situation where I played all my GAC rounds and had 50% win/loss ratio at my current skill points, and my opponent had 50% win/loss at same skill points, because they won 100% of the rounds they played but only played 50% of them.

    My odds of winning are indeed 50%: the matchmaking was fair.

    But the outcome depends entirely on whether my opponent will show up or not: there is no game element.

    This screams for a simple conceptual fix: continue to match players based on skill points, but also match them by the probability they play a round based on their history of 0 banner rounds.

    What’s the probability that CG will be able to code such a system successfully and bug free, do you think?

    50/50
  • Gamma
    1 posts Member
    What’s the probability that CG will be able to code such a system successfully and bug free, do you think?

    This is a matchmaking post, not a CG/bug complaint post. How to balance thorough testing vs speed of new content release is unrelated to matchmaking.

    The idea is that they don’t need to revamp the new system to address a lot of concerns from the community, and that it’s something they should prioritise.
  • CCyrilS
    6669 posts Member
    Gamma wrote: »
    What’s the probability that CG will be able to code such a system successfully and bug free, do you think?

    This is a matchmaking post, not a CG/bug complaint post. How to balance thorough testing vs speed of new content release is unrelated to matchmaking.

    The idea is that they don’t need to revamp the new system to address a lot of concerns from the community, and that it’s something they should prioritise.

    Nobody is saying your idea doesn't have merit. Just pointing out a reality of implementation.
  • CCyrilS
    6669 posts Member
    But hey, maybe they'll use your idea for LOTR!
  • In my opinion, a way they could improve it would be a combination of the new system and old system as follows:

    - Skill rating is determined purely on win / loss as it is today, so no change here
    - Skill rating is what determines your placement in League / Division, so no change here
    - Matchmaking THEN looks at a range of Skill Rating, let's say 25 skill rating points, and groups the brackets by closest overall GP

    This would solve the issue of the show / no show players, but the downside is that smaller rosters could make it all the way to Kyber 1 as they would essentially always be battling other smaller rosters, climbing together. It would be more "fun" this way as the matches would actually be competitive, but maybe unfair to those 7M rosters who are long time players...
  • Gamma wrote: »
    What’s the probability that CG will be able to code such a system successfully and bug free, do you think?

    This is a matchmaking post, not a CG/bug complaint post. How to balance thorough testing vs speed of new content release is unrelated to matchmaking.

    The idea is that they don’t need to revamp the new system to address a lot of concerns from the community, and that it’s something they should prioritise.
    Don’t get me wrong - I do think something needs to be done about matchmaking. I’m just pouring on a dash of realism to your suggestion.

    CG have made a very odd decision in both GAC and TW matchmaking of late. They’ve focused on individuals and guilds getting a reasonably even spread of rewards from the 2 PvP modes, rather than ensuring all matches are “fair”.

    This is truly baffling to me, as I can’t imagine anyone is happy with a 6-6 record in a GAC campaign where the 6 wins were walkovers because the opponent didn’t play and the 6 losses were landslides because the opponent did play. Surely people would rather finish 3-9 where all 12 matches were close affairs that could have gone either way.

    Sadly, this has become the norm for the devs. Whenever they make a new kit, or tweak one with a new ability, or change the way a game mode operates, they fail to anticipate the implications time after time.
Sign In or Register to comment.