Dean Ayala responded to a thread on Twitter about matchmaking in Arena and was able to confirm some details on how the system works.
- Your record with your deck is assigned a value.
- The matchmaker attempts to match you with another deck with the same value, but possibly a different record.
- If a match can't be found after some time, it will slowly expand the search range until one is found.
- They have enough data to determine how different records do against each other.
Dean also went on to confirm an internal skill rating for players, though it isn't used in matchmaking.
- They track a skill rating for players but only use it for data tracking.
- Match outcomes are decided almost entirely by the skill level of the players involved. A fact they know due to the skill rating metric.
You can read on below for his full tweets.
Quote From Dean Ayala All records (0-2 –> 11-0) are assigned a value and are then matchmade by that value. This isn't exactly correct, but it looks something like this:
0-2 = -10
1-2 = -9
0-1 = -8
etc
etcThe order of records and distance between records is some complex table. Basically we have enough data at this point to understand how well a 2-0 performs vs a 2-1 or an 8-2 performs vs a 0-0 on average.
And just like other matchmade modes, the matchmaker looks for an opponent with an identical value, then after X time extends the search parameters a small amount and repeats that process until a match is found.
Contrary to popular belief, match outcomes are decided almost entirely by the skill level of the players involved. We know this because we track a skill rating as well we just don't use it for anything outside of data tracking.
![All The Great Dark Beyond Card Reveals for October 23 - Day 13 [Finished]](the-great-dark-beyond-day-13.webp)
Comments
I'm curious if the "skill rating" is decided solely by wins/win percentage, or if deck composition, and/or smart plays that result in swing-turns plays a role at all.
Am I understand this incorrectly?
They are giving scores to the user's current win/loss state, but not the deck's power. And they are saying since deck powers are equal, the game is decided by the player's skill levels!?
Let's say if a deck is very bad but a very skilled player is piloted it to 5-1 win/loss ratio. Another user which has a lower skill level has a very good deck and achieved a 5-1 ratio also. When these two players compete, how can we say that the player with a higher skill level will win?
I think they should add a multiplier for the deck's estimated power rating when calculating a player's power score.
The fact that you would pilot a bad deck to 5:1 is the exact definition of game outcome decided by the better player, given that it’s well above average already (witch would be 3:3).
You should see winrate more over large amounts of runs then one specific deck. Obviously not every deck can get to 12 wins and that’s perfectly normal. But on average, better player have significantly better results.
Bad Decks that managed to reach 5:1 have surpassed their expectations already and don’t need an additional benefit for even higher results
Partially correct. Dean is saying that match outcomes are decided almost entirely by the skill level of the players involved. That does not imply that the decks are equally strong. Take into account that the better player is also more likely to draft the better deck. Thus, not only does the better player make the better plays, but the better player is also likely to have a better deck. Apparently, situations like you are describing are just not that likely to occur.
So Dean just told me to Git Gud here. Well, nice.
well at least no hidden MMRs there...
Personally, I detest the use of hidden MMR in digital card games. It seem like most of them use hidden MMR, but I just cannot grasp why the MMR has to be hidden.
At bottom, these systems require us to just "trust" the company that their matchmaking is fair, because it's based on your MMR compared to your opponent's MMR.
I mean, if we could see the MMR's, then we could actually confirm that this is true. But they hide the MMR's, so we don't have any way of knowing if they are actually matching us against players with similar skill to us or not. We just have to trust them, in the end.
The whole thing just seems way shadier than it has to be. What would it hurt to reveal MMR?
(I feel like I must be missing some fundamental problem with revealing the MMR, since it's so widespread to use hidden MMR, but it just doesn't sit well with me at all.)
I believe the first time I saw it brought in (which I think may have been WoW Arena? But I forget) the argument went something like "we don't want players to be able to easily game this system to be able to grief noobs, plus the rating is unlikely to change much over time so wouldn't give very good feedback to players" or something like that.
I dread the day that hidden MMR and skillbased matchmaking finally takes over this last game mode.