Forums / Community / Matchmaking Feedback & Discussion

[Locked] Matchmaking Feedback Update – June 11

OP ZaedynFel

  1. 1
  2. ...
  3. 2
  4. 3
  5. 4
  6. ...
  7. 50
To me , in ranked playlists winning is everything . Winning is skill; it just doesn’t make sense to use a system that awards you only 1 for a win and 15 for a loss . This is simple , win is a win , and winning represents skill no matter how you slice it.
ZaedynFel wrote:
It won't be getting HCS settings no. As the most popular Ranked playlist and the #2 playlist overall, it's more popular than any HCS-settings list has ever been by an order of magnitude.

But we will try and move the Slayer Forge maps over. This takes a bit of time because they need to be setup and play-tested with the current Slayer settings / loadouts / common weapons on map, etc.
I'm excited to hear this. I was hoping I'd be able to play these maps within a playlist that didn't have the HCS settings. I hope they get added to the Core Play playlist too.
**kindly answered by eLantern below
To me , in ranked playlists winning is everything . Winning is skill; it just doesn’t make sense to use a system that awards you only 1 for a win and 15 for a loss . This is simple , win is a win , and winning represents skill no matter how you slice it.
What about a person who teams up with 3 better players and then largely sits back hiding so that they simply don't die? Their team wins, but does that person show an increase in skill? That's an extreme example but the point is that winning a match doesn't mean you personally got better from that match; it just shows that your team was better. But the +1 and -15 thing is for rank, not skill; there is a subtle difference.
ZaedynFel wrote:
ZaedynFel wrote:
All Onyx Players Start in DiamondWith TrueSkill2 being stricter and with the changing population dynamics of the Ranked playlists, we are going to shift where we let players start for the next season.
The highest Rank players can start at will be Diamond 3. Onyx and Champ players will then work their way up from there.

In addition, the minimum Rank for being a Champ will just be the minimum Onyx Rank of 1500.

This allows us to align with our definition of Champ players being the top 200 Onyx players, while also requiring players to put in some games to prove they deserve the ranks we expect them to get.
Let me preface this by saying I am not opposed to revised/improved ranks or Trueskill 2. But something w/ MMR seems off (and I'm not talking about CSR inflation) and I hope it gets the attention it deserves, if needed.

You posted a couple MMR plots that were interesting- MMR not responding how I would anticipate if functioning correctly at the high end. The clearest case is laid out in the following link:

FFA vs Doubles MMR//

MMR trends down over a large sample size, despite a very high win rate in one of the game's most competitive playlists. There were other examples posted, but this is the most complete data set. In my opinion, this isn't how MMR should be functioning- there is not a realistic window for the player to improve and rank up. I'm obviously not looking at the raw data, but it seems like widespread problem. Perhaps it is something already being addressed. If not, it is worth noting that the player in the linked case would not rank up in the new season despite winning >80% of their games.

I've seen enough push back on the forums, Twitter, etc. telling players they were previously "boosted" or they aren't as good as they thought- in this case (and others) the evidence strongly suggests otherwise. I suspect that there is already some recognition of this, as you qualify the top players list in HCS w/ those having played more than 50 games - perhaps acknowledging that in addition to CSR inflation, there are a the number of fresh accounts, etc. currently in champ that may not have had the negative MMR influences other accounts have. *** Actually, re: the last comment, the HCS list provided does not remotely mesh with the leaderboards on Waypoint.
I answered this is the old thread.

The Ranks are different as expected because players that sit on their current inflated ranks won't go down.

HCS is not one of the lists affected by the issue.
I'm not sure what questions you are answering here. I would still be interested to understand why MMR would drop over time despite a 80% win rate. The FFA rank is already mostly converged. The doubles rank may *never* converge because of the win rate and the MMR trend. Resetting ranks and capping it at Diamond 3 alone doesn't fix that downward MMR trend.
Josh addressed Juror's rank oddity here and here.

In particular he mentioned this...
Quote:
We're looking into a few rare cases like this, and have a working solution in prototype that maintains the accuracy we already have on normal cases, but also correctly captures the skills of players in this situation. It's not a widespread issue, and also doesn't affect most playlists / modes fortunately. 99.9% of cases where people have questions about their MMRs, etc., have been fine, but like any system, there are a few exceptions. In this case, we'll have a fix out for this in the near future.
eLantern wrote:
ZaedynFel wrote:
ZaedynFel wrote:
All Onyx Players Start in DiamondWith TrueSkill2 being stricter and with the changing population dynamics of the Ranked playlists, we are going to shift where we let players start for the next season.
The highest Rank players can start at will be Diamond 3. Onyx and Champ players will then work their way up from there.

In addition, the minimum Rank for being a Champ will just be the minimum Onyx Rank of 1500.

This allows us to align with our definition of Champ players being the top 200 Onyx players, while also requiring players to put in some games to prove they deserve the ranks we expect them to get.
Let me preface this by saying I am not opposed to revised/improved ranks or Trueskill 2. But something w/ MMR seems off (and I'm not talking about CSR inflation) and I hope it gets the attention it deserves, if needed.

You posted a couple MMR plots that were interesting- MMR not responding how I would anticipate if functioning correctly at the high end. The clearest case is laid out in the following link:

FFA vs Doubles MMR//

MMR trends down over a large sample size, despite a very high win rate in one of the game's most competitive playlists. There were other examples posted, but this is the most complete data set. In my opinion, this isn't how MMR should be functioning- there is not a realistic window for the player to improve and rank up. I'm obviously not looking at the raw data, but it seems like widespread problem. Perhaps it is something already being addressed. If not, it is worth noting that the player in the linked case would not rank up in the new season despite winning >80% of their games.

I've seen enough push back on the forums, Twitter, etc. telling players they were previously "boosted" or they aren't as good as they thought- in this case (and others) the evidence strongly suggests otherwise. I suspect that there is already some recognition of this, as you qualify the top players list in HCS w/ those having played more than 50 games - perhaps acknowledging that in addition to CSR inflation, there are a the number of fresh accounts, etc. currently in champ that may not have had the negative MMR influences other accounts have. *** Actually, re: the last comment, the HCS list provided does not remotely mesh with the leaderboards on Waypoint.
Josh addressed Juror's rank oddity here and here.

In particular he mentioned this...
Quote:
We're looking into a few rare cases like this, and have a working solution in prototype that maintains the accuracy we already have on normal cases, but also correctly captures the skills of players in this situation. It's not a widespread issue, and also doesn't affect most playlists / modes fortunately. 99.9% of cases where people have questions about their MMRs, etc., have been fine, but like any system, there are a few exceptions. In this case, we'll have a fix out for this in the near future.
Big thumbs up for the link- I had moved on from that thread and would have never have seen it. Not convinced it's not more widespread but glad it's at least on the radar.
Team Arena is crazy now, just had 4 of the most one sided games ever, I'm not the best player but seems to be a massive skill gap on the go here?

Just did slayer and got done 50-13, best guy on our team getting 6 kills.
This isn't me moaning because I suck at the game but more because these changes are making it not enjoyable to play.
Is there a reason behind this?

Also, if I play social, does that just mean I could end up playing against people a lot better than me, or worse? So non competitive games.
KotorBay85 wrote:
Team Arena is crazy now, just had 4 of the most one sided games ever, I'm not the best player but seems to be a massive skill gap on the go here?

Just did slayer and got done 50-13, best guy on our team getting 6 kills.
This isn't me moaning because I suck at the game but more because these changes are making it not enjoyable to play.
Is there a reason behind this?

Also, if I play social, does that just mean I could end up playing against people a lot better than me, or worse? So non competitive games.
In Team Arena you have won 56% of your matches against opponents who on average have the same skill as you (Plat).

In Slayer you have a 59% win % against Plat opponents as well.

But if you're playing early in the morning PDT, your mileage may vary on the gap between the best and worst player in the match.
@KotorBay85
Population troughs often lead to matches at the fringe of what's considered "acceptably" balanced. A skill cap that helps to allow the matchmaker to find matches for highly skilled players can tilt the odds even past the so-called "acceptable" limit (it kind of tricks the system). Once in game, players may or may not perform according to their history (people have bad/good games); plus, any quitting will throw the outcome odds out the proverbial window.

The speed at which the "acceptable" balance range (team-to-team) expanse itself is quicker when searching Social playlists and the range of individual skills (i.e. best to worse player on the team) it'll consider constructing a team with is greater too.

Edit: Josh was quicker with his response.
Just had another game where I was on the winning side of CTF, guy on my side got 14 kills, their top guy had 4, one guy quit mid way but game was done in 3 minutes.

Surely something is wrong here?
To me , in ranked playlists winning is everything . Winning is skill; it just doesn’t make sense to use a system that awards you only 1 for a win and 15 for a loss . This is simple , win is a win , and winning represents skill no matter how you slice it.
Winning when already expected to win means the system already knows your skill and it hasn't changed.

Also, it's not a "reward" system so saying "awards you" is misleading. It's a skill system. If you want to go up, you have to do more than just win (e.g. Onyx beating Plat doesn't mean anything), you have to win in a situation when you weren't expected to win already, thus meaning your skill estimate was underestimated, in which case it would need to go up.
It seems like there is still an issue were a game starts 3v4 because someone got kicked from the server. Sometimes the 4th person that was supposed to be on that team doesn't even show up in the post-game carnage report, making it seem like the game loaded only 7 players into the match. Can this be fixed?
Syxnrgy wrote:
Who said it was rotational?
It's literally in the first post:
ZaedynFel wrote:
This list will start as a rotational and whether we keep it permanently will depend on how well it does.
Personally I think Ranked should strictly be competitive only options. Look at any other game (CSGO, RS6, LoL, SC2, Overwatchetc) found in eSports that has RANKED matchmaking... it is the same deal; it is because the settings are the most balanced for competitive play, whether players like it or not.

This should be the new matchmaking system from Day 1 for Halo: Infinite and what remains of Halo 5: Guardians:
Ranked
Team Arena (HCS 4v4 Settings)
Doubles (2v2 Settings)
Free-For-All (HCS Settings w/Kill Cap)
Rotational: Snipers etc.

Social
Big Team Battle (8v8 Settings)
Team Social (4v4 Settings - Slayer, Capture the Flag, Strongholds, KOTH, Assault, Dominion, Extraction, Oddball, Ricochet etc.)
Infection (Infection settings)
Rotational: Grifball etc.

We strictly DO NOT have a sustainable population anymore in the Halo Franchise due to people having different likes and dislikes which can be catered for with the different types of games on the market today etc. It is not the matter of making X great again, it simply is fact and we should come to terms with it... we just need to figure out how to optimize matchmaking search times, region connections etc. Directing more players into groups will accomplish this...
Syxnrgy wrote:
Syxnrgy wrote:
Who said it was rotational?
It's literally in the first post:
ZaedynFel wrote:
This list will start as a rotational and whether we keep it permanently will depend on how well it does.
Personally I think Ranked should strictly be competitive only options. Look at any other game (CSGO, RS6, LoL, SC2, Overwatchetc) found in eSports that has RANKED matchmaking... it is the same deal; it is because the settings are the most balanced for competitive play, whether players like it or not.

This should be the new matchmaking system from Day 1 for Halo: Infinite and what remains of Halo 5: Guardians:
Ranked
Team Arena (HCS 4v4 Settings)
Doubles (2v2 Settings)
Free-For-All (HCS Settings w/Kill Cap)
Rotational: Snipers etc.

Social
Big Team Battle (8v8 Settings)
Team Social (4v4 Settings - Slayer, Capture the Flag, Strongholds, KOTH, Assault, Dominion, Extraction, Oddball, Ricochet etc.)
Infection (Infection settings)
Rotational: Grifball etc.

We strictly DO NOT have a sustainable population anymore in the Halo Franchise due to people having different likes and dislikes which can be catered for with the different types of games on the market today etc. It is not the matter of making X great again, it simply is fact and we should come to terms with it... we just need to figure out how to optimize matchmaking search times, region connections etc. Directing more players into groups will accomplish this...
I'd be fine with ranked snipers being a rotational ranked playlist, I just wish they would tell us exactly when it would be brought back. Also, the team social playlist you propose might be better with only 3 or 4 objective gametypes. Any more than that might be a bit too much imo.
ZaedynFel wrote:
The goal is to stop splitting our Ranked audience between two core competitive 4v4 experiences, and in addition make it clear where players transitioning over to Ranked can find the core experience. On the social side, we would like to provide a core 4v4 experience as well, as an obvious option to transition to after Quick Play.
Sounds like a great idea. The boost in population should really help with match quality.
ZaedynFel wrote:
To me , in ranked playlists winning is everything . Winning is skill; it just doesn’t make sense to use a system that awards you only 1 for a win and 15 for a loss . This is simple , win is a win , and winning represents skill no matter how you slice it.
Winning when already expected to win means the system already knows your skill and it hasn't changed.

Also, it's not a "reward" system so saying "awards you" is misleading. It's a skill system. If you want to go up, you have to do more than just win (e.g. Onyx beating Plat doesn't mean anything), you have to win in a situation when you weren't expected to win already, thus meaning your skill estimate was underestimated, in which case it would need to go up.
My one issue with this is that it seems that the system has a binary approach. Where if you are "projected" to win, even marginally, you have a very small CSR gain you can make. Basically, I have nothing to "gain" by winning but can lose significant CSR if a teammate decides to quite out or just play terrible. Their poor performance weighs me down. I do think the new system is better, I'm just iffy on this one part.

Is there any bias towards more heavily weighing recent performance in MMR?
Can you put something up so we can at least see what to expect when ranking after games? And why start at Diamond 3? I thought yall didnt want us to grind up?
Kashnasty wrote:
Can you put something up so we can at least see what to expect when ranking after games? And why start at Diamond 3? I thought yall didnt want us to grind up?
They are making a soft ceiling at d3 so that the champion ranks will still be a little bit of a grind since there is no real playlist progression anymore. It's to stop people from ranking champion 1 with 10 games played even if their mmr is really high.
This post has been edited by a moderator. Please refrain from making non-constructive posts.
*Original post. Click at your own discretion.
Spoiler:
Show
Syxnrgy wrote:
Syxnrgy wrote:
Who said it was rotational?
It's literally in the first post:
ZaedynFel wrote:
This list will start as a rotational and whether we keep it permanently will depend on how well it does.
Personally I think Ranked should strictly be competitive only options. Look at any other game (CSGO, RS6, LoL, SC2, Overwatchetc) found in eSports that has RANKED matchmaking... it is the same deal; it is because the settings are the most balanced for competitive play, whether players like it or not.

This should be the new matchmaking system from Day 1 for Halo: Infinite and what remains of Halo 5: Guardians:
Ranked
Team Arena (HCS 4v4 Settings)
Doubles (2v2 Settings)
Free-For-All (HCS Settings w/Kill Cap)
Rotational: Snipers etc.

Social
Big Team Battle (8v8 Settings)
Team Social (4v4 Settings - Slayer, Capture the Flag, Strongholds, KOTH, Assault, Dominion, Extraction, Oddball, Ricochet etc.)
Infection (Infection settings)
Rotational: Grifball etc.

We strictly DO NOT have a sustainable population anymore in the Halo Franchise due to people having different likes and dislikes which can be catered for with the different types of games on the market today etc. It is not the matter of making X great again, it simply is fact and we should come to terms with it... we just need to figure out how to optimize matchmaking search times, region connections etc. Directing more players into groups will accomplish this...
90% of my friends list plays 95% to 99% Slayer exclusively, and absolutely hate the HCS settings, and also dislike objective modes. If 343 was to follow your suggestion and completely remove Slayer-only, i know they would move to other games. Removing slayer isnt going to happen. Neither is updating it to HCS settings. There is no way that 343 would purposefully alienate its largest playerbase.
  1. 1
  2. ...
  3. 2
  4. 3
  5. 4
  6. ...
  7. 50