Skip to main content

Forums / Community / Matchmaking Feedback & Discussion

[Locked] Matchmaking Feedback Update – March 13

OP ZaedynFel

  1. 1
  2. ...
  3. 6
  4. 7
  5. 8
  6. ...
  7. 9
Challenges
A few of you have asked about having some set of daily goals to hit in the form of “quests”. Halo 4 had this in the form of rotating Challenges. We also like this feature and would love to improve on it even further, though we won’t be seeing it in Halo 5.

I really like this idea. Adds an element to playing the game daily without it getting stagnant.
This post has been edited by a moderator. Please do not call out individuals. This includes forum members, moderators, administrators, and non-forum members.*Original post. Click at your own discretion.
Spoiler:
Show
RzR J3ST3R wrote:
Would love to hear your thoughts, thanks again for all the attention you have given this thread.
Hey, thanks for that example!

So the data is not biased towards good players. In fact, the by far more popular majority case most likely swamps how CTF is played at the high end. Most players probably do what you saw, just run around slaying and now and then get lucky and score.

One story we could use to explain the counter-intuitive trend that we see in the data is that, like in your match's example, a lot of players run around and slay, on both teams. If that's the case, then, on average, the teams that slay well are more likely to get lucky and score, thus winning. So it comes back to slaying again, and that slaying creating more opportunities to get lucky and score. That appears to be how Halo CTF has been played by the "average" player for many years based on data from severla past Halos.

Notice I said "average", clearly NOT how it's "meant" to be played nor how it is played at the high end. But, nonetheless, how the large majority of millions of players have been experiencing it, and therefore we have to consider how things actually happen vs. how a much smaller group of folks thinks they "should" happen. If we ignore this, we run the risk of designing systems that alienate the majority with no benefit. It would be a bit unfair to expect solo players without mics to play the way an organized team does. But we could at least make them aware that there's a better way.

As you get towards the high end, it probably flips and changes into coordination being key, but that is probably washed out in the millions of games where it really is just random luck favoring the slaying team since they're more likely to be alive and able to grab a flag when they can.

Ideally, we'd have a system that supported that game play, but also showed that there's a "better" way. A system that didn't neglect the mainstream folks and called out the players that really do cause wins because of this, but also show players that there's a level above that type of play, which is how you personally are playing.

My current working idea is something like this:
  • Personal Rank that moves whenever you do not play in a full party. This is going to be biased towards solo performance, even at the expense of the outliers. This is necessary to reward the play experience of by far most of our players. It's also allows players who don't have mics or can't get into parties to still feel like they "shine" BUT the personal Rank would, in theory, not be featured as prominently in the UI. Just enough to find it.
  • Team Rank that is attached to specific full parties that only moves when you play with that specific party. This one would be purely win- and skill-based. Only goes up on a win, down on a loss, proportional to match difficulty. Teams could be named, and would be prominently featured in the UI, ideally with in-game leaderboards. Maybe even be the only place we allow Champions.
The goal is to support and how our mainstream solo folks are actually playing the game, while podiuming the aspiration to play as the organized team.

This stuff is always in flux though, I'm just sharing where my current head is.
Great post again.
I'm a little confused about how mmr and rank fit together, I remember the mmr/ranked post saying matchmaking doesn't really take into account rank and mmr fluctuates and notice again here a reference to communicating why something similar about how golds went up against diamonds cause they 'just played like diamonds' which is great. I would've thought though that certain mmr brackets are tied to ranks (it sounds that way with some of the language here, is that just to simplify?) and that a mmr system that ends up putting golds vs diamonds is highly likely to be over sensitive to random variation in performance.

I used to get the feeling in game (when I played a lot and was improving) that rank was fairly stable but matching too sensitive, perhaps making rank stable as an artifact (i'd rofl stomp then get rofl stomped then roflstomp then get roflstomped so sure it's 50% but it's not a match and it's not climbing ranks either. - The odd occassaion of beating a team ranked like 8 lvls higher feels like it isn't rewarded in Elo/Msr adequately either, a nice big bump up even if I'm only going to lose it again would've made that experience less tiresome).
I would've thought that both mmr and rank were heavily weighted by 'average performance' and that 'an off game' or lucky 'winning streak' would only effect it minimally and as such a certain amount of deviance from average is given as leeway in both mmr and rank - e.g let's say almost everyone (95%) in platinum 3 is mmr 1200 +\- 150. To me something like that would make sense as a ranking and mmr system but it sounds like h5 uses something different?
I mean surely after a few games, a single game is little more than random noise and the average is what matters, seems unhelpful that mmr can fluctuate so much faster than rank can.
The odd gold vs diamond blowout is ok but if it's so often it needs a new UI to explain to ppl that aren't enjoying it what just happened...seems like some improvement could be made to make the two systems more aligned instead, or does that lead to unacceptable wait times?
Really interested to hear more.
Thanks
MMR doesn't always more around a lot. Though currently we let it bounce a bit more during seasonal placement matches in case players really did change a lot (most do not). It really is heavily weighted on average performance, and recent stuff doesn't heavily sway it, like you said.

But the point is we want to reserve the right to do whatever we want with MMR. Maybe one day we'll base it on game mode instead of playlist. Maybe we'll roll all kinds of behaviors into it. Maybe it'll move in strange, counterintuitive ways that aren't "fair" from a reward system point of view. For example, maybe a high-ranked team beats a lower ranked team and the MMR system decides the outcome was too close, and wants to LOWER the MMR of the winning team. It may be mathematically the correct thing to do, but it seems unfair to the team that won, and we'd never do that in CSR. It would be like a having a competitive sports game where the underdog almost wins, so we actually give them the win even when they lose. But MMR needs to be free to do that.

MMR is meant to be expected performance. CSR is meant to be actual recent performance. Currently seasonal. So we want you to re-justify the MMR we think we know about you each season. This way, if MMR decides you aren't as good, you can prove it wrong with CSR.

To be fair, in the main case, the two numbers should and often do synch up and go in tandem, in which case it may seem redundant. But I still like the separation for the cases where they aren't. Also, I like the separation so I can make CSR more straightforward in the design and UX.
TGI LeCheS wrote:
A) back in Halo2, I remember finishing a matchmaking game and going back to my PC on bungie.net/stats... There was a bar being filled with color blue showing how far you were from next rank... If I'm not wrong Halo 5 beta also had something like this... Why was this dismissed in the final version of the game? Is it possible to see it back???

You can already see this on your profile here on waypoint the same way you could on bungie.net/stats

You currently haven't been placed yet this season, but once you do, you will see it on that bar next to each playlists's CSR.

Change it to last season (Jan Feb 2017) and you'll see that you were Gold 5 in Doubles, almost half-way to Gold 6.
ZaedynFel wrote:
Great post again.
I'm a little confused about how mmr and rank fit together, I remember the mmr/ranked post saying matchmaking doesn't really take into account rank and mmr fluctuates and notice again here a reference to communicating why something similar about how golds went up against diamonds cause they 'just played like diamonds' which is great. I would've thought though that certain mmr brackets are tied to ranks (it sounds that way with some of the language here, is that just to simplify?) and that a mmr system that ends up putting golds vs diamonds is highly likely to be over sensitive to random variation in performance.

I used to get the feeling in game (when I played a lot and was improving) that rank was fairly stable but matching too sensitive, perhaps making rank stable as an artifact (i'd rofl stomp then get rofl stomped then roflstomp then get roflstomped so sure it's 50% but it's not a match and it's not climbing ranks either. - The odd occassaion of beating a team ranked like 8 lvls higher feels like it isn't rewarded in Elo/Msr adequately either, a nice big bump up even if I'm only going to lose it again would've made that experience less tiresome).
I would've thought that both mmr and rank were heavily weighted by 'average performance' and that 'an off game' or lucky 'winning streak' would only effect it minimally and as such a certain amount of deviance from average is given as leeway in both mmr and rank - e.g let's say almost everyone (95%) in platinum 3 is mmr 1200 +\- 150. To me something like that would make sense as a ranking and mmr system but it sounds like h5 uses something different?
I mean surely after a few games, a single game is little more than random noise and the average is what matters, seems unhelpful that mmr can fluctuate so much faster than rank can.
The odd gold vs diamond blowout is ok but if it's so often it needs a new UI to explain to ppl that aren't enjoying it what just happened...seems like some improvement could be made to make the two systems more aligned instead, or does that lead to unacceptable wait times?
Really interested to hear more.
Thanks
MMR doesn't always more around a lot. Though currently we let it bounce a bit more during seasonal placement matches in case players really did change a lot (most do not). It really is heavily weighted on average performance, and recent stuff doesn't heavily sway it, like you said.

But the point is we want to reserve the right to do whatever we want with MMR. Maybe one day we'll base it on game mode instead of playlist. Maybe we'll roll all kinds of behaviors into it. Maybe it'll move in strange, counterintuitive ways that aren't "fair" from a reward system point of view. For example, maybe a high-ranked team beats a lower ranked team and the MMR system decides the outcome was too close, and wants to LOWER the MMR of the winning team. It may be mathematically the correct thing to do, but it seems unfair to the team that won, and we'd never do that in CSR. It would be like a having a competitive sports game where the underdog almost wins, so we actually give them the win even when they lose. But MMR needs to be free to do that.

MMR is meant to be expected performance. CSR is meant to be actual recent performance. Currently seasonal. So we want you to re-justify the MMR we think we know about you each season. This way, if MMR decides you aren't as good, you can prove it wrong with CSR.

I'm sure I read somewhere there have two cases where the winner of a chess tournament has lost rank points/Elo because the person they beat was ranked much lower than them....I do kinda think that that is fair and mathematically correct. Thanks for the explanation though, I think that clears things up for me.
I'm sure I read somewhere there have two cases where the winner of a chess tournament has lost rank points/Elo because the person they beat was ranked much lower than them....I do kinda think that that is fair and mathematically correct. Thanks for the explanation though, I think that clears things up for me.
That's interesting. I hadn't heard about people losing points in chess for winning, but I have heard of them losing points for a draw. If a significantly stronger opponent can only draw, they may not be as good as we thought, etc.
ZaedynFel wrote:
I'm sure I read somewhere there have two cases where the winner of a chess tournament has lost rank points/Elo because the person they beat was ranked much lower than them....I do kinda think that that is fair and mathematically correct. Thanks for the explanation though, I think that clears things up for me.
That's interesting. I hadn't heard about people losing points in chess for winning, but I have heard of them losing points for a draw. If a significantly stronger opponent can only draw, they may not be as good as we thought, etc.
I wonder if, in that case, the winning player was far and way much better than everybody else who entered the tournament. Like, for (an extreme) example, the number 1 player in the world entered a tournament filled with players not even ranked in the top 1000 in the world. In a case like that, I'd find it understandable (but still odd) if said top player lost some points for winning the tournament since the system wouldn't have been impressed with the victory since it was never in doubt, or at least it shouldn't have been.
ZaedynFel wrote:
I'm sure I read somewhere there have two cases where the winner of a chess tournament has lost rank points/Elo because the person they beat was ranked much lower than them....I do kinda think that that is fair and mathematically correct. Thanks for the explanation though, I think that clears things up for me.
That's interesting. I hadn't heard about people losing points in chess for winning, but I have heard of them losing points for a draw. If a significantly stronger opponent can only draw, they may not be as good as we thought, etc.
I wonder if, in that case, the winning player was far and way much better than everybody else who entered the tournament. Like, for (an extreme) example, the number 1 player in the world entered a tournament filled with players not even ranked in the top 1000 in the world. In a case like that, I'd find it understandable (but still odd) if said top player lost some points for winning the tournament since the system wouldn't have been impressed with the victory since it was never in doubt, or at least it shouldn't have been.
Yeah well. A blind squirrel finds a nut every so often.
ZaedynFel wrote:
RzR J3ST3R wrote:
Would love to hear your thoughts, thanks again for all the attention you have given this thread.
Hey, thanks for that example!

we have to consider how things actually happen vs. how a much smaller group of folks thinks they "should" happen. If we ignore this, we run the risk of designing systems that alienate the majority with no benefit. It would be a bit unfair to expect solo players without mics to play the way an organized team does. But we could at least make them aware that there's a better way.

As you get towards the high end, it probably flips and changes into coordination being key, but that is probably washed out in the millions of games where it really is just random luck favoring the slaying team since they're more likely to be alive and able to grab a flag when they can.

Ideally, we'd have a system that supported that game play, but also showed that there's a "better" way. A system that didn't neglect the mainstream folks and called out the players that really do cause wins because of this, but also show players that there's a level above that type of play, which is how you personally are playing.

My current working idea is something like this:
  • Personal Rank that moves whenever you do not play in a full party. This is going to be biased towards solo performance, even at the expense of the outliers. This is necessary to reward the play experience of by far most of our players. It's also allows players who don't have mics or can't get into parties to still feel like they "shine" BUT the personal Rank would, in theory, not be featured as prominently in the UI. Just enough to find it.
  • Team Rank that is attached to specific full parties that only moves when you play with that specific party. This one would be purely win- and skill-based. Only goes up on a win, down on a loss, proportional to match difficulty. Teams could be named, and would be prominently featured in the UI, ideally with in-game leaderboards. Maybe even be the only place we allow Champions.
The goal is to support and how our mainstream solo folks are actually playing the game, while podiuming the aspiration to play as the organized team.
I want to make sure I correctly understand your future plan... because if I'm getting it, then I have some HUGE complaints and questions for you. If I'm misunderstanding, sorry about that. But it sounds like you're implying that this 'personal rank' and 'team rank' would replace the single CSR number we currently have, and your MMR/CSR adjustment will be affected by your K/D. (I say MMR/CSR because you've explained that CSR converges to MMR with more games played as you "prove" your MMR to the system - so CSR is fundamentally dependent on your MMR)

  1. I think you misunderstand me -- I primarily play solo queue, and if I do play with a fireteam, it's rarely the same specific To4. The game I linked you was solo queue. I fact, I think the two teammates who simultaneously hate-messaged me were probably playing as a party. So if anyone was going to be organized, it should have been them, not me. Although well-organized parties certainly have some advantages, it doesn't require a-priori "organization" to play in a smart manner and contest key map positions to support your teammates, that's called "knowing how to play Halo well," even in solo-queue, and should absolutely factor into your rank."
  2. I used to be stuck at lower ranks, and I could never understand why I was sometimes out-slaying my higher-ranked teammates in objective gametypes. I used to think... "man, why does my onyx teammate suck!?" But then I watched my games in theater from all POVs, watched some pro players on stream, and LEARNED how to play the game smarter, all while in solo-queue.
  3. Please tell me, according to your proposed system, could I lose MMR/CSR to my gold-ranked teammates for that game I linked..?? Despite the fact I outplayed them as evident in theater mode?? Note that the highest ranked person (plat 5) on the enemy team ALSO had the worst K/D on his team - it wasn't just me. Here's a totally separate game where the highest ranked players on BOTH teams also had the worst K/D. In that second example, I'm plat 4, and I go 22-13. My teammate is high onyx (CSR 1869, ranked #671 in the world in Team Arena playlist) and he goes 16-21. So I'm better at plat 4 than the onyx 1869 guy...?? Um, no. Again, this is all solo queue. I see this phenomenon frequently, both on the winning and losing teams. Clearly my Onyx 1869 teammate was doing something better/smarter than I am...
  4. If a higher ranked player stands to lose CSR/MMR in those situations because you decide that individual slaying stats are the prime contributor to wins, rather than combining smart map positioning with good gun skill and situational awareness to help your teammates (even in solo queue), it will be infuriating to me and to many. In fact, the CSR system will become a complete joke. Players will cease trying to win games and instead try to stat-buff themselves, which will lead to lower-quality games for everyone.
  5. Under the proposed system, especially if I always solo-queue, I have reduced incentive to try and win, and greater one to stat-buff. Worse, it will make the CSR experience meaningless for solo-queue players, since I basically can't get a real CSR ('team rank') unless I can find a reliable To4 to always play with.
  6. BTW, I'm very glad to note that the past week of this current season, after 16 months of this game, I FINALLY feel like the CSR ranks are close to accurate. Players who outrank me tend to school me, and vice-versa. Halo5arena.com seems to corroborate that - I win almost every game where my team avg. CSR is higher, and lose almost every game where we're lower. So, thank you very much for making the change reducing the number of games it takes to "prove" your CSR and converge close to your MMR. It really made a huge difference and I finally feel like my rank is a truer indicator of my skill. I still want to evaluate it the rest of the season, but it seems like a massive improvement. Plus, now I can actually try to improve... I can watch how higher ranked players beat me, and try to counter it. But now I'm afraid you are proposing undoing all your great work that has finally paid off into the current system...
Thanks again for your time.
This post has been edited by a moderator. Please do not call out individuals. This includes forum members, moderators, administrators, and non-forum members.*Original post. Click at your own discretion.
Spoiler:
Show
This system is so dumb, I literally get punished for being good and searching alone. Because the system puts terrible players on my team and considers it balanced. There is absolutely nothing balanced about Proving Grounds right now.
This system is so dumb, I literally get punished for being good and searching alone. Because the system puts terrible players on my team and considers it balanced. There is absolutely nothing balanced about Proving Grounds right now.
The better you play the worse the team you get.

You don't have to say anything about the already doomed from the start matches.
I literally had a match recently in ranked where the lowest ranked guy on the opposite team was higher than the highest rank on mine. How it expects a team of gold and mid platinum to fight against a team of diamond and onyx and not get pub stomped i'll never know.

A gold should never fight a diamond let alone multiple diamonds and an onyx. It defeats the purpose of ranks if were doing stuff like this.
I can't even enjoy playing Proving Grounds because of the randomness of who I match and get on my team.
This system is so dumb, I literally get punished for being good and searching alone. Because the system puts terrible players on my team and considers it balanced. There is absolutely nothing balanced about Proving Grounds right now.
I can't even enjoy playing Proving Grounds because of the randomness of who I match and get on my team.
You're currently a Plat 5 in Proving Ground. (Just a statement of fact in preparation for my next comments.)

In your last 10 games, the ranges of ranks were mostly high Golds to mid Diamonds. There were two Onyx players. Once on your team, once on the opposing team, and a couple mid Golds in one match, one on each team.

The ranks look pretty balanced overall.
This system is so dumb, I literally get punished for being good and searching alone. Because the system puts terrible players on my team and considers it balanced. There is absolutely nothing balanced about Proving Grounds right now.
I can't even enjoy playing Proving Grounds because of the randomness of who I match and get on my team.
You're currently a Plat 5 in Proving Ground. (Just a statement of fact in preparation for my next comments.)

In your last 10 games, the ranges of ranks were mostly high Golds to mid Diamonds. There were two Onyx players. Once on your team, once on the opposing team, and a couple mid Golds in one match, one on each team.

The ranks look pretty balanced overall.
The problem is the ranks aren't accurate. I've seen platinums that have the skill of Champions, and I've seen platinums that don't know how to maneuver their character around the map. You can't just look at the ranks and automatically assume it's balanced.
Josh, It would be just fantastic if you can do something about the Solo versus To3/To4. Preferably sooner rather than later. Going into a game knowing you're cannon fodder for the opposition isn't fun. It also doesn't encourage people to want to continue playing either. At this stage something is better than nothing assuming the something is a little more than just maintaining the current system.
Bucshot888 wrote:
ChallengesA few of you have asked about having some set of daily goals to hit in the form of “quests”. Halo 4 had this in the form of rotating Challenges. We also like this feature and would love to improve on it even further, though we won’t be seeing it in Halo 5.

I really like this idea. Adds an element to playing the game daily without it getting stagnant.
this should been and maybe could be in halo 5 with an update cause it would add more incentive to play the game and unlock stuff like halo 3 did,like halo reach or a combination of both?
ZaedynFel wrote:
RzR J3ST3R wrote:
Would love to hear your thoughts, thanks again for all the attention you have given this thread.
Hey, thanks for that example!

So the data is not biased towards good players. In fact, the by far more popular majority case most likely swamps how CTF is played at the high end. Most players probably do what you saw, just run around slaying and now and then get lucky and score.

One story we could use to explain the counter-intuitive trend that we see in the data is that, like in your match's example, a lot of players run around and slay, on both teams. If that's the case, then, on average, the teams that slay well are more likely to get lucky and score, thus winning. So it comes back to slaying again, and that slaying creating more opportunities to get lucky and score. That appears to be how Halo CTF has been played by the "average" player for many years based on data from severla past Halos.

Notice I said "average", clearly NOT how it's "meant" to be played nor how it is played at the high end. But, nonetheless, how the large majority of millions of players have been experiencing it, and therefore we have to consider how things actually happen vs. how a much smaller group of folks thinks they "should" happen. If we ignore this, we run the risk of designing systems that alienate the majority with no benefit. It would be a bit unfair to expect solo players without mics to play the way an organized team does. But we could at least make them aware that there's a better way.

As you get towards the high end, it probably flips and changes into coordination being key, but that is probably washed out in the millions of games where it really is just random luck favoring the slaying team since they're more likely to be alive and able to grab a flag when they can.

Ideally, we'd have a system that supported that game play, but also showed that there's a "better" way. A system that didn't neglect the mainstream folks and called out the players that really do cause wins because of this, but also show players that there's a level above that type of play, which is how you personally are playing.

My current working idea is something like this:
  • Personal Rank that moves whenever you do not play in a full party. This is going to be biased towards solo performance, even at the expense of the outliers. This is necessary to reward the play experience of by far most of our players. It's also allows players who don't have mics or can't get into parties to still feel like they "shine" BUT the personal Rank would, in theory, not be featured as prominently in the UI. Just enough to find it.
  • Team Rank that is attached to specific full parties that only moves when you play with that specific party. This one would be purely win- and skill-based. Only goes up on a win, down on a loss, proportional to match difficulty. Teams could be named, and would be prominently featured in the UI, ideally with in-game leaderboards. Maybe even be the only place we allow Champions.
The goal is to support and how our mainstream solo folks are actually playing the game, while podiuming the aspiration to play as the organized team.

This stuff is always in flux though, I'm just sharing where my current head is.
Ok, I understand what your thinking/trying to do (I think... I've always I been more of a visual person lol) but if I am getting this, as I said in previous posts on this thread, I think this isn't a good idea.

Again, how do you figure out when someone does well? If I lose and I'm top spot on my team with a 16-3-9 and do 1800 damage, am I better then then the last place person on my team who goes 7-12-7 and does 2200 damage?? What would you be basing these personal stats on, just K/D or what? If stats start becoming bases on K/D, you'll have a whole crap load of people who only focus on THERE stats and don't care about winning anywhere near as much. I agree with a lot of what RzR J3ST3R wrote too. I think it'll make people who assist a lot or are doing the Objective in an Objective game types not feel as important if they don't get many kills because your now basing it on K/D only (if I'm understanding it right) To me, assists mean just as much as a kill.

Also, I think a lot of people rarely play with the same 3 people in say team Slayer for example. I probably play with 3 to 8 people on team Slayer. So how would that work in your team party rank system if I'm not playing with the same three people all the time??

I do agree that if you are going to do a system like this, solo players should not be able to reach the highest ranks of say champion (please change rank names to military names like it was Halo 3, I'm talking for H6 and beyond here, no point changing them in H5. Besides what sounds better I'm a Platinum 5, onyx 1700 or I'm a lieutenant commander or Brigadier General.... I think most would like the ladder, we are the UNSC after all..)

I still think the win/lose system is the best. I think most people's complaints when playing solo is that they are matched with or against peoples whos skill level is far worse or far superior to theirs. I feel like it's the matchmaking system that needs tweaking more than anything. So maybe you could use individual skill ranking to match people up here, but keep that individual skill ranking hidden maybe or something... I don't know. I just know if someone​ is Platinum 3, their team should be all Platinum Players and they should be playing against all Platinum Players. If this isn't possible it should be the closest Spartans to that rank.

One thing is for sure, if you're going to implement this type of system I would try and Halo 5 first, not Halo 6. See what people think.
The problem is the ranks aren't accurate. I've seen platinums that have the skill of Champions, and I've seen platinums that don't know how to maneuver their character around the map. You can't just look at the ranks and automatically assume it's balanced.
From looking at game histories and ranks in the proving grounds playlist, I suspect that the playlist is still too new for the ranks to be accurate. There are people who are high onyx in team arena, but only platinum in proving grounds. I suspect over the next several weeks the ranks will normalize in that playlist and get more accurate. It takes a while for the MMR system to adapt in a 100% new playlist where nobody has prior MMRs to help the system initialize.
  1. 1
  2. ...
  3. 6
  4. 7
  5. 8
  6. ...
  7. 9