Forums / Community / Matchmaking Feedback & Discussion

Why is MMR such a secret???!

OP GEEbaybee90

  1. 1
  2. ...
  3. ...
  4. 2
With the new “tru2skill” ranking system it seems that it places a more significant emphasis on “MMR.” More people are curious and wondering what their MMR is including myself. If MMR plays such an integral part of game, why is it kept a secret from players as to what their MMR is?

mod edit: moved to Matchmaking Feedback & Discussion
This post has been edited by a moderator. Please do not repost content a moderator has removed, repost a topic that has been locked, or post about forum moderation decisions. If you have a question or concern about a forum moderation decision, please private message the applicable moderator.
*Original post. Click at your own discretion.
Spoiler:
Show
MMR is not used as the public rank because unlike CSR, MMR can change wildly from game to game. Example: let's pretend you're normally a low Diamond player in Slayer, and that's what your MMR is. You have to really good games, unnaturally good, where you essentially got lucky. Your MMR shoots up into Onyx. You're super happy. But then your next games, you play like normal, so your MMR goes back to low Diamond. Then, you have a few really bad games, where you were just playing like crap for some reason. Your MMR falls to low Plat. You feel terrible, because you know you play better than that, but you just had a few bad games.

MMR can change pretty wildly. Last season, someone's MMR changed 1200 points from one game. Seeing those changes reflected publicly would be quite jarring. Hence, CSR is used as public rank, because it can only change in increments of 1 or 15. It's much more stable. And now, since CSR only changes by 1 point if your MMR and CSR don't match, once the values converge, CSR becomes a much better measure of skill than ever before. So once you reach a point where you only gain 1 CSR for a win, you then know essentially what your MMR is.

Future Halos may have ways of showing players their hidden skill. But Halo 5 will continue to show CSR publicly, while using MMR on the backend to determine matches.
The impression I get from Dr Menke who created the system is that it's not possible to add stuff to the postgame matches in H5, but they're considering adding things in future games.
LUKEPOWA wrote:
The impression I get from Dr Menke who created the system is that it's not possible to add stuff to the postgame matches in H5, but they're considering adding things in future games.
This, and I also get the impression from Josh that if people knew it they might be able to boost easier and what not.
CSR doesn’t make any sense either because I’ll play a match against low diamonds and my rank will go up 1 point but then I’ll play a game with a team that has a couple of onxys level players on it, lose, and my rank goes down 15 points, obviously I’m not going to beat a team of onxys. This new tru2 skill makes it an endless grind. Wish things were like halo 4.
If you link your games in this thread, you can possibly get your questions answered.
Do not post about forum moderation.

Spoiler:
Show
MRVEEK wrote:
I agree the csr doesn’t make sense. And Halo 4’s matchmaking was superior. It allowed a better measure of accuracy of where one stands amongst the broad community, unlike halo 5 where most matches are filtered to put you against similar skilled players
What? Looser skill matching which was in H4 doesn't give you an accurate measure of where you stand in the community. That's like getting a champion rank for beating up bronzes all the time. H5's system is more accurate because it can figure out what your actual rank is better than previous systems using their new formula. Matching similarly skilled people is better because you're not gonna have champions stomping on bronze players all the time. Obviously, population issues can come into play and you'll sometimes get uneven matches, but the system will accommodate for that behind the scenes.

Have you read the first posts in this thread and this thread to understand CSR and MMR better?
LUKEPOWA wrote:
MRVEEK wrote:
I agree the csr doesn’t make sense. And Halo 4’s matchmaking was superior. It allowed a better measure of accuracy of where one stands amongst the broad community, unlike halo 5 where most matches are filtered to put you against similar skilled players
What? Looser skill matching which was in H4 doesn't give you an accurate measure of where you stand in the community. That's like getting a champion rank for beating up bronzes all the time. H5's system is more accurate because it can figure out what your actual rank is better than previous systems using their new formula. Matching similarly skilled people is better because you're not gonna have champions stomping on bronze players all the time. Obviously, population issues can come into play and you'll sometimes get uneven matches, but the system will accommodate for that behind the scenes.

Have you read the first posts in this thread and this thread to understand CSR and MMR better?
Oddly I agree with MRVEEK that H4 was a better system, but I say so for the opposite reasons - I felt like I always got better balanced matches in 4, not the free-for-all of no skill restrictions that he seemed to think it was. Had that really been the way H4 worked I'm sure the quitting would have been epidemic and the population would have dropped to nothing within a few months of launch... oh, wait. Well, that's no different than Halo 5, so there's that.

TS2 is, imo, a mixed bag. It seems to work better (though low population makes it hard to know for sure, as you say), but I sure would like to know just how much weight they're giving to some of the new parameters that are involved. All in all, though, I have to agree with the OP that while there may be good reasons to have MMR seperate from CSR, it certainly is a cause for suspicion amongst players when the ranking system is not the transparent mechanism used to make matches. Why not just give everybody a champion CSR since it doesn't really mean anything anyway?
All I’m saying is, regardless of MMR being hidden because of wild fluctuations from match to match, It would still be nice to know, as player, what my MMR is. I’m sure it wouldn’t be an issue to even show a players “average MMR” from match to match. If MMR is a much better indicator of “skill”, doesn’t make sense to keep it secret from players. I consider myself a causal gamer, but it’s discouraging to play when, even when I place highest on a loosing team my CSR goes down 15 points. I agree with another poster who mentioned, the system shouldn’t be fixed but rather variable. MMR is basically like taking a test and never getting the results back.
All I’m saying is, regardless of MMR being hidden because of wild fluctuations from match to match, It would still be nice to know, as player, what my MMR is. I’m sure it wouldn’t be an issue to even show a players “average MMR” from match to match. If MMR is a much better indicator of “skill”, doesn’t make sense to keep it secret from players. I consider myself a causal gamer, but it’s discouraging to play when, even when I place highest on a loosing team my CSR goes down 15 points. I agree with another poster who mentioned, the system shouldn’t be fixed but rather variable. MMR is basically like taking a test and never getting the results back.
I think I have an expansion of your test analogy that explains why they don't show MMR.

First off imagine MMR as a test that never changes or very rarely does. Now after you take that test (game) you receive back the final result (CSR). However you don't get back the test and can't see the individual questions that really explain thoroughly when happened (MMR). There are several reason why keeping the answers to the questions a secret are good. First off since the test doesn't change often if you gave people solutions to the questions they could use those to pinpoint specific questions and get the answers always right (basically play the system and always gain MMR). Right now since you just received the general results there is no surefire way to discover what you did to exactly increase or decrease your MMR.

As Chimera has said MMR can fluctuate wildly. So while you are taking 15 tests (games) in a day you may do really bad on one and get lucky and do well on another. Since these aren't the average of what you nornally do it's not fair to base your grade (rank) off those few data points. Now if you start doing better in every game then your CSR will rise with your MMR. Basically it reduces confusion as people would get even angrier if they suddenly lost 100 points due to a few bad games.

Side point: it seem you haven't read the posts linked so I'll explain what going up and down 1 or 15 means

If you WIN: going up 1 means your pregame CSR is higher than post game MMR. Going up 15 means your pregame CSR is less than your post game MMR.

If you LOSE: going down 1 means your pregame CSR was lower than your post game MMR. Going down 15 means your pregame CSR was higher than your post game MMR.
Oddly I agree with MRVEEK that H4 was a better system, but I say so for the opposite reasons - I felt like I always got better balanced matches in 4, not the free-for-all of no skill restrictions that he seemed to think it was.
I don't really believe that you always got better matches. I played H4 and I had plenty of games that were mismatches and blowouts which is similar to any social system like in Reach or H5 for example.
LUKEPOWA wrote:
Oddly I agree with MRVEEK that H4 was a better system, but I say so for the opposite reasons - I felt like I always got better balanced matches in 4, not the free-for-all of no skill restrictions that he seemed to think it was.
I don't really believe that you always got better matches. I played H4 and I had plenty of games that were mismatches and blowouts which is similar to any social system like in Reach or H5 for example.
I would say "over all" or "on average" the matchmaking seemed to be better. I readily admit that's purely anecdotal evidence, but that game seemed so much less hostile than this one, and matchmaking quality was a big part of that. Or so I remember it. On the other hand, these perceptions could be down to nothing more than a lack of visible ranks in that game. In Halo 4 I might get kicked around by some guy in a match, but when that match was over then I put it out of my mind. In Halo 5 I look at his rank in the post-game lobby and if it's three tiers higher than mine then I become infuriated at matchmaking and Halo 5 is forever etched in my brain as the garbage game that thinks Silver/Diamond matches are totally okay.
I would say "over all" or "on average" the matchmaking seemed to be better. I readily admit that's purely anecdotal evidence, but that game seemed so much less hostile than this one, and matchmaking quality was a big part of that. Or so I remember it. On the other hand, these perceptions could be down to nothing more than a lack of visible ranks in that game. In Halo 4 I might get kicked around by some guy in a match, but when that match was over then I put it out of my mind. In Halo 5 I look at his rank in the post-game lobby and if it's three tiers higher than mine then I become infuriated at matchmaking and Halo 5 is forever etched in my brain as the garbage game that thinks Silver/Diamond matches are totally okay.
The hostility could be related to the competitiveness of ranked plus how the game was designed with not getting a lot of breathing room in-between fights due to smaller maps and abilities allowing you to move around faster. I think not having visible ranks does play a role in social, but for the most part, you're going to get more games with people of average to below average skill so you don't have to try as hard to win fights and you also don't have that pressure to perform well which can make it feel like a better experience even in uneven games.

The mismatched ranked games in H5 are very likely due to population. The way I look at it is; the matchmaker couldn't find anyone within your team's skill range so it's either this or you don't find a game at all. It's an unfortunate compromise, but we can't do the social system because then there wouldn't be any point to trying to get a similarly skilled game as much as possible in order to get an accurate rank. On the plus side, there are a couple things behind the scenes which can help minimize the impact of those games to your rank.
Most developers keep their matchmaking system relatively secret so people don't "play to the formula." I'm actually quite surprised how transparent 343 has been in discussing their system.
I just wish I could see my MMR it stresses me out more than arena does because I can't see it. I usually find out by looking at all the other player's service records and their ranks in arena.
I just wish I could see my MMR it stresses me out more than arena does because I can't see it. I usually find out by looking at all the other player's service records and their ranks in arena.
Being able to see MMR would likely stress people out more since it's more unstable than CSR.
There's a 20+ page paper written about it, it's not exactly a secret Lol.

There's even an open API for stats that you can test it on (and an entire season worth of stats to use as a baseline, from when they placed you exactly at your MMR), so nothing is stopping anyone from reverse engineering the code.

But most people fall into the laymen category and aren't going to understand all that anyway, which is why they dumb it down to a more long-term average of CSR. That's why your MMR is hidden, because there's a lot that goes into it and nobody is going to take the time to understand it and instead, just complain about things they don't understand.

What really matters to the user at the end of the day is whether the rank is accurate, and CSR provides a more general picture of that to give feedback based on.

LUKEPOWA wrote:
The impression I get from Dr Menke who created the system is that it's not possible to add stuff to the postgame matches in H5, but they're considering adding things in future games.
This, and I also get the impression from Josh that if people knew it they might be able to boost easier and what not.
He seemed to note those concerns as more of a general "initial hesitation" at the beginning, or just "taking it slow" to see how it works in a blind setting (ie. as a control, and to make sure there's no serious bug that got overlooked).

But in fact, one of the main goals with Trueskill 2 was to make it resistant to manipulation by design (eg. everything being based per-minute, everything as a relative comparison to other ranks, and CSR acting as a clamp) so that they wouldn't have to keep re-working it and dealing with manipulation tactics on a case-by-case basis.

So I really don't think the boost-ability aspect is a serious concern on a more long-term basis, more of an initial footnote for early testing. The true aim in regards to boosting is actually to remove the need for secrecy in the first place (eg. as Luke noted with their desire for more information in upcoming titles).

In fact, you can see this in his posts -- he has given more & more details as time has went on, and by this point I'm sure anyone clever enough could figure out a basic framework for generating estimate MMR graphs like he has posted recently -- although there wouldn't be much point because, as the system continues to prove its resistance to manipulation, it becomes more & more likely that those graphs will just become the standard anyway.
LUKEPOWA wrote:
Oddly I agree with MRVEEK that H4 was a better system, but I say so for the opposite reasons - I felt like I always got better balanced matches in 4, not the free-for-all of no skill restrictions that he seemed to think it was.
I don't really believe that you always got better matches. I played H4 and I had plenty of games that were mismatches and blowouts which is similar to any social system like in Reach or H5 for example.
I would say "over all" or "on average" the matchmaking seemed to be better. I readily admit that's purely anecdotal evidence, but that game seemed so much less hostile than this one, and matchmaking quality was a big part of that. Or so I remember it. On the other hand, these perceptions could be down to nothing more than a lack of visible ranks in that game. In Halo 4 I might get kicked around by some guy in a match, but when that match was over then I put it out of my mind. In Halo 5 I look at his rank in the post-game lobby and if it's three tiers higher than mine then I become infuriated at matchmaking and Halo 5 is forever etched in my brain as the garbage game that thinks Silver/Diamond matches are totally okay.
I played a decent amount of Halo 4 because I enjoyed the game, but there's no way I'd consider the matchmaking quality better in Halo 4 than Halo 5.

Halo 4 dealt with an improperly implemented JiP system for all match types that was being relied on to back-fill matches that constantly seemed to start with an uneven number of players on each team because they prioritized search time over ensuring team's were adequately balanced and filled out.

Without dedicated servers P2P connections granted host advantages and it also led to a ton of host migrations during matches; especially, when combined with the lousy JiP system that was implemented into Halo 4.

Halo 4 allowed splitscreen players access to online matchmaking and the added strain that created often negatively affected the stability of the match too.

The relied upon JiP system combined with P2P network performance strains (many related to splitscreening players) and host migrations would lead to such annoying in-game pauses (aka black screens) right in the middle of a match's action.

Also, in Halo 4 you'd occasionally run into some players making use of a classic cheat tactic via lag switching.

Lastly, Halo 4 did eventually introduce CSRs, but you had to visit the Halo 4 Waypoint app/site to visualize them, so I understand what you're trying to get at regarding visual in-game ranks.

Now, despite my ragging on Halo 4's JiP system I would like to make it clear that I was happy during that period of time that 343i finally adopted JiP into the franchise, but I was highly critical of how they went about it. Even today, with a vastly better JiP system implemented into Halo 5, there's room for future improvement.

As for in-game visible ranks, I think the way Halo 5 handles them is pretty damn good. People know what they're getting into when they search a ranked playlist, but you can only gain the information post-match instead of pre-match like older titles use to provide you. Sure you might get upset by witnessing what ranks were in a match, but with improvements toward rank integrity that hopefully becomes less of an issue. Increased population would be another method to help improve upon the issue you mention.
  1. 1
  2. ...
  3. ...
  4. 2