Why so much Slayer in HCS?
The current representation has every possible HCS configuration once. It so happens that Slayer has the most configurations. If we duplicate CTF / Stronghold entries, then you will sometimes get those twice in a row. Right now we are erring on the side of not seeing the same entries multiple times in a row, but we are willing to reconsider.
Ideally we would just weight each configuration for a more even representation, but right now we can't do that. So we have to choose between duplication or over-representation.
What are the difference in matchmaking across the different modes?
We get a few questions about this, so I'll try and summarize.
They all work almost the same, the main exception being Firefight. Firefight matchmaking does its best to put everyone onto the best dedicated server it can, but that’s it.
Ranked and Social Arena, and also the non-Firefight Warzone playlists all use skill similarly, with these differences:
In order to help our very top players actually get into matches, we treat them as having lower MMRs than they actually have when they matchmake. Otherwise, they would often not get any matches at all. This results in high-skilled players getting the occasional rare very difficult match vs. top players.
How do you update maps for a playlist? Or just how do you update playlists in general?
The same way we update just about anything. A combination of
To give you an idea of how long it can take to do an evaluation, just the small changes we are currently making to the HCS preview list take about 1 week to get right. This includes:
This means we have to prioritize the playlists, often by a combination of popularity and health. This unfortunately means some playlists don’t get as much attention as others.
Why are my some of my matches unfair?
According to our stats, the great majority of matches are actually amazingly close. Both in terms of MMR and actual match outcomes.
Confusion happens because the only things we show are CSR and SR. Neither of these are used for matchmaking. We use MMR. We’ve actually run tests to look at CSR and SR and both are significantly worse than MMR for the purposes matchmaking. However, they are also both better than MMR for Ranking and Progression respectively.
So, for the most part, you are being put in the best matches we can.
But, we can do a better job at communicating that these matches are fair than we currently do. I’ve mentioned a few times that I’d like to call out the MMR of each team (not player) so you can see how close each match actually is. It’s not something we can do in the short term, but just so you know we are aware there is a good solution to this.
It almost always feels bad when you lose, but it’s often random performance differences in a given match that causes that, and not an actual discrepancy in skill. Sometimes players perform much better or worse than usual or expected. That leads to almost all of the “unfair feeling” cases we see.
Now, that all said, there are some fairly rare exceptions where our higher-skilled, though not top players can get into uneven matches. This happens because our top players would never get to play at all if we didn’t let them match against worse players. But, again, this is rare because the number of high-skilled players is much much larger than the number of top players. The high-skilled mostly just match other high-skilled players. It just “feels” memorably bad when those large gap matches do happen because the difference in skill is so obvious.
The current representation has every possible HCS configuration once. It so happens that Slayer has the most configurations. If we duplicate CTF / Stronghold entries, then you will sometimes get those twice in a row. Right now we are erring on the side of not seeing the same entries multiple times in a row, but we are willing to reconsider.
Ideally we would just weight each configuration for a more even representation, but right now we can't do that. So we have to choose between duplication or over-representation.
What are the difference in matchmaking across the different modes?
We get a few questions about this, so I'll try and summarize.
They all work almost the same, the main exception being Firefight. Firefight matchmaking does its best to put everyone onto the best dedicated server it can, but that’s it.
Ranked and Social Arena, and also the non-Firefight Warzone playlists all use skill similarly, with these differences:
- Ranked imposes a maximum MMR gap that can never be breached. This gap is mostly the same for every playlist---1 full Rank’s worth, which means mid-Plat can match up to mid-Diamond in MMR, for example---, though I’ve eased a few back because there were players unable to match in some lists. After 40 seconds of waiting, Ranked will allow anything within that maximum gap. In addition, if you don’t search focused, and someone else has already waited 40 seconds, that player can match with you instantly as long as they are within that gap.
- Social doesn’t have that gap. After 40 seconds of waiting, social allows any skill gap. Also, if you don’t search Focused, you can end up in a large gap match immediately if someone else in the match already waited over 40s
- Warzone is the same as social, except it only waits 30s because it has to gather three time more players.
In order to help our very top players actually get into matches, we treat them as having lower MMRs than they actually have when they matchmake. Otherwise, they would often not get any matches at all. This results in high-skilled players getting the occasional rare very difficult match vs. top players.
How do you update maps for a playlist? Or just how do you update playlists in general?
The same way we update just about anything. A combination of
- Design evaluation
- Playing the playlists
- Tracking stats (playlist health, map quit rates, etc.)
- Listening to community feedback
To give you an idea of how long it can take to do an evaluation, just the small changes we are currently making to the HCS preview list take about 1 week to get right. This includes:
- Making the changes in the map and playlist configurations
- Double checking all of the changes we intended to make happened, and none we didn’t.
- Testing those changes (usually requires two full teams)
- Doing any needed localization changes (languages) if there are new names
- Handing off the changes to the Production team that actually gets them uploaded and put into the live game. This process also has quality checks.
This means we have to prioritize the playlists, often by a combination of popularity and health. This unfortunately means some playlists don’t get as much attention as others.
Why are my some of my matches unfair?
According to our stats, the great majority of matches are actually amazingly close. Both in terms of MMR and actual match outcomes.
Confusion happens because the only things we show are CSR and SR. Neither of these are used for matchmaking. We use MMR. We’ve actually run tests to look at CSR and SR and both are significantly worse than MMR for the purposes matchmaking. However, they are also both better than MMR for Ranking and Progression respectively.
So, for the most part, you are being put in the best matches we can.
But, we can do a better job at communicating that these matches are fair than we currently do. I’ve mentioned a few times that I’d like to call out the MMR of each team (not player) so you can see how close each match actually is. It’s not something we can do in the short term, but just so you know we are aware there is a good solution to this.
It almost always feels bad when you lose, but it’s often random performance differences in a given match that causes that, and not an actual discrepancy in skill. Sometimes players perform much better or worse than usual or expected. That leads to almost all of the “unfair feeling” cases we see.
Now, that all said, there are some fairly rare exceptions where our higher-skilled, though not top players can get into uneven matches. This happens because our top players would never get to play at all if we didn’t let them match against worse players. But, again, this is rare because the number of high-skilled players is much much larger than the number of top players. The high-skilled mostly just match other high-skilled players. It just “feels” memorably bad when those large gap matches do happen because the difference in skill is so obvious.