Forums / Community / Matchmaking Feedback & Discussion

[Locked] Matchmaking Feedback Update – September 25

OP ZaedynFel

  1. 1
  2. ...
  3. 2
  4. 3
  5. ...
  6. 7
Menke Meddling

We have some new updates trickling in from Xbox Live that let us improve matchmaking.
The first is live now, and allows us to expand our searches for up to 10 minutes, instead of only 2.5. We currently time out our searches at 5 minutes though, so in practice, there’s still a 5-minute max for now. The main difference is we can expand search rules during the entire 5 minutes, instead of only half of it. For example, before this change, we could only expand the allowable skill gap for 2.5 minutes, and then after that rely on a set hard threshold that could no longer expand. Now, we can potentially continue expanding the allowable skill gap over the full 5 minutes, resulting in tighter matches.
This also means we can extend the amount of time we try and pair full fireteams with only other full fireteams. We tested this over the weekend at the full 5-minute mark. We may back off a minute on this to give a small window of time during which full fireteams can match with non-full fireteams so people wait a little less than now.
The second feature we are waiting on isn’t live yet, but it will make a larger difference once it is. This feature is a bit more complicated to explain. The short answer is, we can make matchmaking and team balancing much tighter when there are fireteams involved than we can today. Note: this is NOT the feature I’ve discussed the past that allows us to measure the skills of fireteams better, that one isn’t coming soon. This is purely a matchmaking improvement and not part of the skill system.
To understand the second feature requires a little more understanding of how the matchmaker currently works on skill. Today, the matchmaker looks at the average MMR of the fireteam you are in --- or just your solo MMR if you are alone, and only allows you to match within a set gap from that value. This means, if you are, e.g., a Silver player in HCS, you generally only see Bronze, Silver, and Gold players in your matches.
This works great if everyone is queuing alone. It’s OK if there are a few solo Gold players in a mostly Silver match because they will get evenly balanced across the teams by the team balancer. However, if there’s a full Fireteam of 4 Gold players, the matchmaker currently doesn’t consider the fact that they can’t get split up. Instead, it can potentially match 4 Golds vs. 4 Silvers. This is because the team balancer will rightly refuse to split up a fireteam.
Let’s say a gold player is a “2” and a Silver player is a “1”. The average gap in the case of a team of Golds vs. Silvers is still within the max tolerance since we have a 2.0 team and 1.0 team. But, total gap is now 4.0 (4x2 = 8 vs. 4x1 = 4), which is not a fair match. We currently have no way to both balance for the solo queue case (where 1.0 is just fine) and the full party case (where 1.0 is always unfair because it’s really 4.0), without making one or the other suffer either longer than needed wait times, or unfair matches. We currently err on the side of unfair matches because most players do queue solo, and use party restrictions to mitigate it. But, those restrictions do NOT mitigate it at all when a full party faces another full party. They actually make it worse.
With the new feature we are expecting soon™ the matchmaker will simultaneously restrict the maximum player to player gap and the team to team gap. This will mean that when a full Fireteam plays, it will always find opponents that are fair.
In fact, when we look in the data and find Fireteams beating non-Fireteams, in the vast majority of cases, the MMR system knew the match was unfair ahead of time, but we couldn’t prevent it with the current matchmaker. It wasn’t because Fireteams are inherently better than solo players, it was because we couldn’t match them right.
This means that most of the unfairness that comes from matching vs. full Fireteams goes away with the update, and means there isn’t much of a reason to keep the current party restrictions. They are causing unnecessarily long wait times in most cases, so when the new feature comes online, we will probably test removing them. It should result in matches that are actually more fair---even with solos matching vs. parties---and faster matching times.
We may also allow 12-player Fireteams in Warzone with the change, but repeat the warning that wait times will be very long for those Fireteams. The only group of players that will have a total MMR high enough to match a 12-player Fireteam will usually be another 12-player Fireteam, of which there won’t be many. So queue at your own peril. If you can’t match this way, you will want to back off on the party size until you can match, like today.
If we get this feature before next week, I will give an update in a tweet, and more explanation in next week’s post.
Cool! Glad you or someone at XBL figured out how to tweak the MM parameters to consider both fireteam avg. MMR and also individual MMRs. A few questions/comments:
  1. Will this raise any issues when your fireteam has diverse MMRs? Like if a gold teams up with a low onyx? Seems like that may cause issues with the requirement that player-to-player gaps be small.
  2. I was initially concerned that adding this additional restriction would lead to even longer wait times than we see today... since I've already seen increased wait times with recent MM tweaks (for the most part, the quality of ranked matches I'm getting has been reasonable). But you think that's mostly due to the party restrictions? So if you can remove those party restrictions, this new person-to-person MMR gap limit will be manageable? It would be awesome if this simultaneously improved MM quality and queue times. So you're saying maybe the person-to-person gap will be limited to 1.0 ranks, but the fireteam-fireteam TOTAL gap (rather than average gap) may also be limited to 1.0?
  3. Are you thinking party sizes of 7-11 will also be allowed in WZ / WZA, with this new restriction? I'm assuming so.
Will the 12 man warzone be a separate playlist with 12 man fireteams only?
Will the 12 man warzone be a separate playlist with 12 man fireteams only? and will it be possible to search with anything between 6 and 12 man fireteams?
The OP stated that it would be the same playlist, just with the 12-man restriction removed. I asked the same question about 7-11 man fireteams.

Also, ZaedynFel I'm curious whether there has been any progress made on a feature that would allow players to more easily re-join a match they'd lagged out of.
RzR J3ST3R wrote:
Cool! Glad you or someone at XBL figured out how to tweak the MM parameters to consider both fireteam avg. MMR and also individual MMRs. A few questions/comments:
  1. Will this raise any issues when your fireteam has diverse MMRs? Like if a gold teams up with a low onyx? Seems like that may cause issues with the requirement that player-to-player gaps be small.
  2. I was initially concerned that adding this additional restriction would lead to even longer wait times than we see today... since I've already seen increased wait times with recent MM tweaks (for the most part, the quality of ranked matches I'm getting has been reasonable). But you think that's mostly due to the party restrictions? So if you can remove those party restrictions, this new person-to-person MMR gap limit will be manageable? It would be awesome if this simultaneously improved MM quality and queue times. So you're saying maybe the person-to-person gap will be limited to 1.0 ranks, but the fireteam-fireteam TOTAL gap (rather than average gap) may also be limited to 1.0?
  3. Are you thinking party sizes of 7-11 will also be allowed in WZ / WZA, with this new restriction? I'm assuming so.
  1. It will still use the average for a Fireteam's MMR. Gold + Low Onyx is going to look like high Plat. But it will still get you tighter matches because the matchmaker will know that the Gold and Onyx player can't be separated, and require opponents with high enough MMR to balance.
  2. Yes, like that. Not exactly 1.0, but that basic idea. The actual value is tuned per playlist a bit because different game modes and team sizes result in different odds between teams. Some modes can have a larger gap, but result in the same fairness.
  3. Yes, the change should adapt fine to any team size. A high-end Warzone player is going to have a rating clipped to 2.0 standard deviations (SD)s. 12 of those guys are going to have a total around 24 SDs. The max gap between WZ teams will be around 5 SDs at worst. This means opponents will have to sum up to at least 19 SDs, meaning they will have to average around 1.58 SDs per player. So the average player-player gap enforced by the NEW rule would be around 0.42, which is actually MORE restrictive than than the current direct player-player rule. That restrictive of a gap is why 12-player fireteams will have a hard time finding matches, but they can at least try. 12-player fireteams will have access to around only 6% of the population compared to the 50% a solo player gets. A 7-player Fireteam is going to bring in "14 SDs" So we would need at least 14 SDs worth of skill for the opponents. The worst player allowed right now in a WZ match with top players is a "0" So if we had 7 2s and 5 0s on one team, that would be 14 SDs. The opponents would then have to sum up to at least (14 - 5) = 9 SDs. 9/12 = 0.75 SDs each, for a gap of 1.25, which is still more restrictive than the 2.0 WZ gap, but easier to find than 1.58 SDs. So a 7-player fireteam can matchmake with 23% of the population, which is a lot more than the 12-player.
You also had a question about re-joining matches, but I have no more info on that at this time.
ZaedynFel wrote:
(4x2.0 = 8 vs. 1.0x2 = 4)
Just a quick math correction for you.

That 2 within the 1x2 should be a 4 (number of players) which is how you would get your outcome of 4 in the equation of SD times Number of Players equals Four (_x_=4).
eLantern wrote:
ZaedynFel wrote:
(4x2.0 = 8 vs. 1.0x2 = 4)
Just a quick math correction for you.

That 2 within the 1x2 should be a 4 (number of players) which is how you would get your outcome of 4 in the equation of _x_=4.
Oh, nice catch. Fixed.
I can't believe it was something so simple as the system thinking it could split parties when it couldn't (it probably wasn't that simple at all to find this error) Well done! Looking forwards to jumping in.

Soon(tm) Wow, I guess once you work at Blizzard it kinda sticks. Can I ask, did you also work on Heroes of the Storm? SCII matchmaker is brilliant, always gives me fun games.

I need to thank you as well for that move changing the ping parameters for AU a few weeks ago, I've been reading your threads but totally missed that part somehow. What a genius move, 170ms to the West Coast is so much better than 250 to the East or Europe, thank you. Definitely the best compromise. Still want that default low ping lock for Halo 6 though when the population is back.

Question, with all these improvements, will the game get to a point where matches are too tight? Sometimes a challenge or a game or two to let loose is nice ya know.
Josh, do you know why they keep raising the games required to complete in order to raise or lower the WZ win percentages a tenth of a percent (e.g. 1% - 1.1%)? It used to be 3 games in the beginning, then 4, more recently 5, maybe 6 (wasn't keeping track) and now it takes 7 games. It's not a big deal or anything, but I was just curious about the reasoning to keep raising it higher and higher over this long period of time.
LUKEPOWA wrote:
Josh, do you know why they keep raising the games required to complete in order to raise or lower the WZ win percentages a tenth of a percent (e.g. 1% - 1.1%)? It used to be 3 games in the beginning, then 4, more recently 5, maybe 6 (wasn't keeping track) and now it takes 7 games. It's not a big deal or anything, but I was just curious about the reasoning to keep raising it higher and higher over this long period of time.
isn't this solely dependent on how many games you played? like, just because of the math? If you'd only played 10 games then 1 win changes your win % a lot. But if you've played 1000, then it takes a lot more wins to change your win %. Am I missing something here?
Haven't thought of it like that, but that could be it. All I know is that it's been taking more games to move up. I'll try it on my other account and see if it takes less games.
RzR J3ST3R wrote:
LUKEPOWA wrote:
Josh, do you know why they keep raising the games required to complete in order to raise or lower the WZ win percentages a tenth of a percent (e.g. 1% - 1.1%)? It used to be 3 games in the beginning, then 4, more recently 5, maybe 6 (wasn't keeping track) and now it takes 7 games. It's not a big deal or anything, but I was just curious about the reasoning to keep raising it higher and higher over this long period of time.
isn't this solely dependent on how many games you played? like, just because of the math? If you'd only played 10 games then 1 win changes your win % a lot. But if you've played 1000, then it takes a lot more wins to change your win %. Am I missing something here?
Correctomundo! The Warzone win percentage is displayed as a whole percentage, so we don't see the changes at the <1% level.

So, LUKEPOWA, when you reached 1000 WZ games played, it would have taken >10 wins to see a 1% increase in your win rate. When you reached 2000 games played, it would have taken >20 wins to see a 1% increase. Now you've played over 2800 WZ games. If you want to see your win rate go up, you'll need to win >28 games. The more games you play, the more wins it will take to show a noticeable change in your win rate, simply because the win rate is displayed as a whole number, not 59.000001% or something like that. And obviously, if you lose some games, it will lower your win rate while still adding to your denominator, meaning every loss only increases the number of wins needed to see an increase in your WZ win rate.
Alright, good to know. Thanks for the info.
All in well bringing in another playlist.....but the game is so buggy and broken and dont have high hopes of this being rolled out without even more bugs
I mean will Warzone have working Arclights by then? will there be an population count?......will that full 12 man team be able to back out after a game without being stuck in the menu and then having to dashboard?

Probably not 👍
Thank god this is getting adressed
Is JIP ever getting fixed? Just had an H3 playlist Oddball game end right as I joined.

Link
Request

-HCS settings in Doubles. Please at least test this out. If theres a playlist this would work in its got to be Doubles.

-BTB maps. The current maps are so stale. Why are Basin, Estuary Unearthed gone?
Unearthed would be fine with maybe one more spartan laser. The only issue was one team controlling both wasps .
Estuary has a really awesome flow for CTF. Just remove the banshees and this map will be perfect. Banshees never really worked on this small BTB map.
These were great maps. We need something new or something to come back in BTB.

-Community maps playlist. Ghosts in the Shell is some of the most fun I've had because it was all new maps. Halo 5 has an incredible forge. Lets make use of it. Even if spawns aren't perfect who cares? Lets have some new stuff.

-Tunings update is so terrible there are not good enough words to describe how i feel about it. I don't know what I'll do if these changes are implemented. I know there is a place to express this point already but seriously I'll pass out flyers at my local community center if it helps.
Could we please have two more playlist?
Halo 3 throwback Doubles ( RANKED )
and
Halo 3 throwback Slayer ( RANKED )

This would bring so many people back. If you are not going to make us an anniversary at least give us ranked competitive games.
The mix between Halo 5 and Halo 3 on the throwback list is really fun, but not having it ranked will have a lot of people gone soon.

Thanks.

343 is heading in the right direction.
Could we please have two more playlist?
Halo 3 throwback Doubles ( RANKED )
and
Halo 3 throwback Slayer ( RANKED )

This would bring so many people back. If you are not going to make us an anniversary at least give us ranked competitive games.
The mix between Halo 5 and Halo 3 on the throwback list is really fun, but not having it ranked will have a lot of people gone soon.

Thanks.

343 is heading in the right direction.
To do something like this we would first have to verify the population is large enough to support adding another 2 lists.

Just making a playlist won't bring the population that it needs to be healthy, and if it's not healthy, then matchmaking is bad and the list loses people steadily.

So we'll be watching the new list. If it gets a massive amount of players, we could justify more lists out of it. If not, it means there's no way to make those lists.
Okay cool, thanks for the response. That does make sense. How has the population for H3 throwback looked so far?
  1. 1
  2. ...
  3. 2
  4. 3
  5. ...
  6. 7