View Full Version : S2 Ranked Matchmaking Discussion

02-07-2017, 07:43 PM
There have been a few players expressing concern about the matchmaking in Season 2 Ranked. We wanted to clarify a bit of the ranked experience below.

We are using the same matchmaking system as we did for Season 1. We use the S1 Match Making Rating (MMR) values as starting values - you are not matched based on your visual rank (bronze, gold, etc) - you are matched by your MMR which is separate and calculated exactly as it was S1. We then seeded you in S2 with a soft reset of your S1 MMR. A soft reset is when we moved everybody's MMR a percentage closer to the average MMR. As an arbitrary example, say you had 1850 MMR in S1, your MMR at the start of S2 would be moved down 50 points towards the average of 1500 MMR. If you had 1400 MMR, your MMR would be moved up to 1425 (again, numbers are just examples). There is only a small difference between your MMR last season and the MMR used to seed you this season.

Example - I start in Bronze V, like everybody does this season, but I keep my MMR from last season which is ~1800. When a match is made I am not matched with other people in Bronze V. Instead, I am matched using my ~1800 MMR. We compensate for this by also giving you bonus points if your MMR is above the target MMR for the division that you're currently in. So instead of getting 50 points per win in bronze, you get 75 points so you can get to the your appropriate division faster.

If you are seeing worse matches this season, it is because there are a lot of new people playing ranked and their MMR is unknown because of lack of games they play, which causes matchmaking to be less accurate. Once these new players play more games the matches will get better. In addition, on average, all the new players coming in will be of lower skill because in most cases they have played the game (and ranked) far less than players who played in S1. But, you never know. Sometimes the game just clicks for people, so it's best not to go into a match with any preconceived notions of someone else's skill level.

Using some arbitrary smaller numbers to show the issue - Say there are 10 really skilled S1 palyers online, at roughly the same time, but they queue up as 3 duos and 2 solos. They queue up 5 minutes apart because of timing reasons and there are 100 new people in the ranked queue. There aren't enough of the high MMR S1 players to make a full, even game because of the 3 duos, so we bring in some of the 100 new players into the game. We could set the wait time longer to achieve a better match-up, but how long is too long? 5 minutes? 10 minutes? Some people say they are willing to wait 10 minutes but that doesn't represent the larger view. After enough games the 100 new players will be spread out enough that they will be matched properly. Until then, there will be a rough patch during the influx of new people.

We are currently using the exact same parameters for making a match as we did in S1. We can tighten those parameters more, and that is what we will try, but I think with time the situation will improve.

All that being said, bugs do happen so I have asked the designers and engineers to double check all the match making configs and code for ranked and also look at the metrics for the matches made to see if we might have messed up a config file somewhere. Everything so far seems to be in order, but we do periodic checks of the data just to be sure.

Here is a video of Pju discussing this very same topic:

02-07-2017, 08:00 PM
I don't think anyone thinks this was a BAD idea. It's frustrating dealing with the obvious difference skill gaps when you're hoping to rank up quickly, but you're correct: In order to balance out where people should (most likely) be, it's necessary to deal with it at the beginning.

My suggestion was that now that this style was implemented for S2 already, S3 should be somewhat different. If the idea is that our MMR plays into effect more than the actual rank does, then placing tiers on starting ranks shouldn't make too much of a difference. For example, those who finish in Bronze/Silver this season should start at Bronze in S3. Those in Gold/Platinum should start at Gold. Those in Diamond/Master should start at Diamond. Contenders should have the option of bumping down, if they'd like, to keep them from being bored. Although, if rank isn't as important as MMR, then Contender probably doesn't matter.

Personally, I know I'd feel better if I ended Plat or Diamond last season and I was being matched with people of similar skill. I think the frustration (especially in Solo Q) is that people are initially grouped with people that may not be as skillful, or people that are new are being destroyed by "higher MMR" without even realizing it. And an unlucky string of these matches can affect the motivation of some players who may have been much higher before.

02-08-2017, 02:10 AM
The key problem is that even though these players are distributed evenly to keep the ranked system fair, that doesn't make it fun. I think this game suffers largely from the weaker members of the team have a far larger negative impact than a good player can have positive as the game is really limited on how hard a single person can carry. This is fair because it happens to you as many times as it happens to the other team, but it hardly feels worth playing the game if the deciding factor is almost always who has the worst player on their team.

The only realistic way I can see a fix to this is setting a level on the amount of experience a player needs to play ranked at a high amount of game. This amount is something i don't really know, I personaly don't think I was remotely good untill I had maybe 1000 games under my belt, and I understand other people don't care as much about joining the queue whilst being 'bad' at the game and they just want to play; but, they ruin the entire experience for 7 other people in the game if they're terrible. Is it worth having 1 extra person play ranked when all it does is make the other 7 peopel in the game not want to play?

02-08-2017, 11:40 AM
Oh.. I sure wonder what would be something extremely easy to do that would fix at least some of the negative impact of inexperienced/weaker players...


update: To anyone else reading this I actually got a dev response on my post, so ignore the initial salt in this comment now, I'm happy to have finally gotten a reply on at least one of the discussions I've brought up.

02-11-2017, 10:23 AM
Hossy you're right that the worst player has a great impact, and that's why it's up to you not to be that player.

02-12-2017, 02:27 PM
Here's a random new thought: Add something that shows if a player doesn't have all the lancers. Tried to save a lancer pick for someone and told him to pick the one I wanted. Then when it's time for him to pick my lancer, he tells me he doesn't have them. At the rate that people speak in team chat, that's insanely annoying.

02-14-2017, 02:12 AM
In spite of the fact that, if rank isn't as essential as MMR, then Contender presumably doesn't make a difference.

02-14-2017, 11:21 AM
Diamond and Master still require either a high MMR or consistently winning greater than 50% to get to. And once you are a master, your raw MMR is displayed on the scoreboard. It's a hybrid system where MMR matters more the higher rank you are.

06-26-2017, 01:21 PM
There is something wrong with the match making when you get 2 top 20 players in the one team, and 2 unranked people in the other team (rest rank 5-15). Are we quite sure match making is working as intended this season (S3)? Because that felt like an absolute nobrainer match and ended in like ten turns (go figure!)