> 2533274906308237;83:
> > 2533274839818445;52:
> > > 2533274906308237;39:
> > > Just following up on an previous post you made about accounting for individual performance in the post-game MMR update:
> > >
> > > "To your last point incorporating individual performance while still honoring the winner: This is something we are working on this very moment for MMR. It’s going to take awhile because it’s a major improvement involving very intricate math to get right. This will improve MMR estimates greatly, although we won’t be applying it directly to CSR in Halo 5. It will indirectly affect CSR though, because CSR is based on enemy difficulty, and enemy difficulty will be more precise that it is now."My question: are you saying that individual performance will not affect post-game CSR updates, but it may cause you to get put into a much harder or much easier game afterwards (because your MMR will have changed significantly based on performance), and thus will give you an opportunity to gain or lose a lot of CSR in that next game (and subsequent games)?__ So basically the CSR gain/loss algorithm with remain ELO-like, based only on Win/Loss, but the matchmaking will change.
> > >
> > > I’ve been a very vocal opponent of including any individual performance stats into player ranks, but I think I really like your idea here, precisely because CSR is not directly affected. Maybe I’m finally understanding it… This would mean that CSR remains totally pure and focused on winning (not stat-hoarding), but matchmaking would be more efficient at knowing when to put low ranked people into harder games, which would then rank them up faster if they can win those hard games.
> >
> > Yes, you are correct.
>
> Been thinking about this, and one issue came to me.
>
> Let’s say player “SpartanJohn” (most generic GT i could think of) loses a few games in a row, cause he’s basically just playing for kill stats in an objective gametype, and not helping win. But his stat line looks good as a result. So now he ends up with an MMR that exceeds his CSR and his actual skill level, and he gets put into some tough games with higher MMR people. The MM system thinks it’s a fair match because it doesn’t know his MMR is inflated. Now his new teammates kinda get screwed… they may lose the game, lose MMR and CSR, all because SpartanJohn has an unrealistically high MMR.
>
> This is probably not a huge effect… if his new teammates lose a game because of SpartanJohn, they’ll likely make up that CSR quickly, if they’re as good as they think. In the end it probably comes out as a wash for them. But it might cause some frustration for them in the meantime.
Ye, there will probably be outliers here and there.
But keep in mind that it’s still constrained by winning. If you’re on the losing team, the system requires that the sum of the performances on the your team be less than the winning team’s. That means you really have to outperform your teammates to benefit in that way. This won’t happen super often if the matchmaking is good.
Do you know why any kind of party restriction system wasn’t implemented when the game came out?
The game is kicking me out of every single match I get in Halo CE or Halo 2. My NAT type is Open, I don’t have any other problem with any other game. WHAT IS GOING ON?!
Whenever I get kicked out a message saying “JOINING FAILED Your attempt to join this player has failed. Please try again.” I’m not even playing with anyone else. Please I’m trying to do a stream on twitch playing all the Halo games on MCC. But this keeps happening.
> 2533274796457055;7:
> > 2533274808246214;6:
> > One form of heavy aim? So other forms will still exist?
>
> This is going to sound snarky but I don’t mean it to.
>
> Read the thread before posting next time:
>
>
>
>
> > 2533274839818445;4:
> > > 2695457991278531;3:
> > > Niiiiiiice!!
> > > When you say “one form of heavy aim”, are you implying there are multiple causes of heavy aim? If so, will those fixes be out soon after this patch?
> > > thanks again for your transparency!!
> >
> > I’m just being careful about saying, “we’ve totally fixed heavy aim”
> >
> > We’ve fixed something that definitely causes aim to feel heavy.
> >
> > We’ll wait and see if there are others.
> >
> > Also, there are bound to be some things that can affect aiming that are out of our hands (ISP, etc.).
im so sick of all this damn lag
> 2533274826313307;87:
> The game is kicking me out of every single match I get in Halo CE or Halo 2. My NAT type is Open, I don’t have any other problem with any other game. WHAT IS GOING ON?!
Issues like this are best posted in the MCC support forums, this topic is for discussing feedback related to Halo 5’s matchmaking
Why when I searched focused do I always match Mexicans?
Knowing how inconsistent hcs mathcmaking is, could you possibly force all players to be in game chat while playing the game. ( Go back to the old days of games like mw2 where you couldnt matchmake unless you had a mic attached). It’s hard enough playing this playlist solo.
I’d like to know why MMR remains hidden. The community gets shown these shiny badges, attached to a numeric, visible, Competitive Skill Rank. This CSR goes up and down with W/L and is shown after every match. Now we have them listed on Leaderboards. It’s all we see, they get reset every season, and we’re told to get our placements and work our way up in rank. It’s only logical that players believe these ranks to be the basis for matchmaking. But when they try to apply their CSR to their matches it most often appears unbalanced and creates a crap ton of needless confusion, frustration and anger. There are multiple threads on here about it every day. Then they get told, “Oh that’s not your real Rank”. Well what is my real rank? “Oh we can’t tell you.” Utterly ridiculous.
EDIT: I reposted here after my OP was locked and I was directed to this thread by a Moderator. I guess it’s my fault for pointing out how many times this issue gets asked about in different threads.
> 2533274968722829;93:
> I’d like to know why MMR remains hidden.
Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
> 2533274839818445;84:
> Quick note. The stopgap is not in yet. The stopgap will be a month or so. The ideal solution is months and months out since it’s part of a full skill system overhaul and needs time to contain all the awesomeness.
>
>
>
>
> > 2533274815533909;79:
> > Well this is also good to know, but also pretty crappie at the same time… Just so I’m clear, if you have the search parameter set to focused, you’ll always play against full teams and never non full parties?
>
> Not quite. If you search outside of a full party and search Focused, you won’t match full parties until you have waited the max wait time, whatever we set that to. Right now it’s 40 seconds. Might increase it for this.
>
> If you are in a full party, it will search for a full party regardless of focused until you hit the max wait time, and then it will try for whatever it can get.
So the stopgap isn’t in yet? you mean the what you said here? When does that come into affect then? or am I just not understanding things here right lol …sorry if not lol
> This stopgap is like the classic solution. We will try to match full parties only against other full parties. It’s a bit inefficient in that there are plenty of full parties that are terrible and don’t need to match against other parties, and those parties will now wait a little longer and get slightly worse matches. But it should be helpful to non-full parties while we finish the ideal solution
Interesting regarding the wait times. Sounds like a pretty decent set-up. If you do increase wait times, any idea what you might bump it up too? rough guess.
> 2533274813317074;94:
> > 2533274968722829;93:
> > I’d like to know why MMR remains hidden.
>
> Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
The problem with a matchmaking system that has a hidden MMR to match players based on it’s predictions of who will win is completely flawed in the first place. If it is matching teams and has a projected winner then it isn’t doing its job of correctly matching teams. Because it should be matching people so closely that I can’t predict the winner.
> 2533274813317074;94:
> > 2533274968722829;93:
> > I’d like to know why MMR remains hidden.
>
> Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
Thanks for the response. I respect that you are trying to be helpful. I pretty much already understand the difference between CSR and MMR, but many players do not. The post you referenced only reinforces the confusion. If MMR is meant to be their “best guess at what your skill can be” then why are we awarded a visible, Competitive****Skill Rank, which is really just a “trophy”, and not a representation of where our actual, competitive skill is ranked? According to his answer, the new, numbered Leaderboards on Waypoint don’t really accurately represent those players’ estimated skill. They only show trophies earned by winning games. They have a hidden list of players numbered by skill. We both know that’s MMR. But 343’s emphasis on earning CSR, calling it a rank, re-setting it each season, and posting it after every match creates an illusion that it is actually an attempt at estimating skill. As far as the second link you referenced, players regularly post screen shots of CSR discrepancies in ranked matches wider than those Josh said H5 does not allow. I’m not going to research all of them, but I’ve commented on some of them in the past and that confusion contributes to the saltiness of the community and the quitting problem the game has. When players are shown two Onyx and two Golds, getting stomped by a Champ, an Onyx and two Plats they don’t know or care that some hidden algorithm thought their competitive skills were evenly matched.
> 2533274847840732;96:
> > 2533274813317074;94:
> > > 2533274968722829;93:
> > > I’d like to know why MMR remains hidden.
> >
> > Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
>
> The problem with a matchmaking system that has a hidden MMR to match players based on it’s predictions of who will win is completely flawed in the first place. If it is matching teams and has a projected winner then it isn’t doing its job of correctly matching teams. Because it should be matching people so closely that I can’t predict the winner.
I get what you’re saying, but stckrboy never said the system predicted a winner. I think he meant what you said: that the “system expects the outcome of your games to be”… “so close that I can’t predict a winner.”
My problem with the hidden system is that we are told it works a lot better than the visible system we are constantly teased and awarded with. The old “don’t pay any attention to the man behind the curtain” is annoying at best.
> 2533274968722829;98:
> > 2533274847840732;96:
> > > 2533274813317074;94:
> > > > 2533274968722829;93:
> > > > I’d like to know why MMR remains hidden.
> > >
> > > Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
> >
> > The problem with a matchmaking system that has a hidden MMR to match players based on it’s predictions of who will win is completely flawed in the first place. If it is matching teams and has a projected winner then it isn’t
>
> My problem with the hidden system is that we are told it works a lot better than the visible system we are constantly teased and awarded with. The old “don’t pay any attention to the man behind the curtain” is annoying at best.
Especially when we’re told that the MMR and CSR system is pretty much the same as the system used for Halo 3. But if I’m understanding correctly, the 1-50 ranking in Halo 3 was MMR (could definitely be wrong here) while Captain, General, etc represented CSR. If that’s the case, then we could see our MMR rather than keeping it hidden. I do wish we could see our MMR, or at least something similar to what has been mentioned in other updates about maybe showing an average MMR of each team.
> 2533274968722829;97:
> > 2533274813317074;94:
> > > 2533274968722829;93:
> > > I’d like to know why MMR remains hidden.
> >
> > Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
>
> As far as the second link you referenced, players regularly post screen shots of CSR discrepancies in ranked matches wider than those Josh said H5 does not allow. I’m not going to research all of them, but I’ve commented on some of them in the past and that confusion contributes to the saltiness of the community and the quitting problem the game has. When players are shown two Onyx and two Golds, getting stomped by a Champ, an Onyx and two Plats they don’t know or care that some hidden algorithm thought their competitive skills were evenly matched.
I can attest to this too. One that sticks out for me personally (please note that this has been a season or so ago and my own matches seem to have since gotten much more balanced) was a game of FFA I played in which I was a Platinum 4 and matched against 2 Onyx players (one was 1800 and something and the other was 1969), 2 Champion players (88 and 24), and an unranked player. While this was before the FFA update where ranks meant next to nothing, it was still a much wider gap than has been discussed (I’ve seen in previous updates that you can expect to match players within one full rank tier of your own rank). To be fair, I didn’t lose any CSR since the players who beat me were of a MUCH higher rank and I tied with the Champion 88 and beat the Onyx 1969). That may not be the best example since like I said, it was pre-update and ranks really didn’t tell you anything about a player’s skill, but that is just one example that sticks out. I also have a screenshot of a doubles game in which I was a Gold 2 or 3, my teammate was a Champion and we played against a team of two Onyx players (not sure of the specific ranks when it comes to where they fit within the Onyx and Champion tiers).
I understand that CSR and MMR are sometimes different, but if there’s that big of a gap, then it would appear that something is wrong. I personally would like to see the most accurate representation of my skill possible. I know that MMR is said to be complicated to show since it’s not just a single number, but even if it was just shown as a stat on Waypoint rather than in-game and was depicted as something like “this is an estimate of what rank we believe you to be” or “this is an estimate of what we expect your rank to get to.” I just find it interesting and would like to know what the system expects me to be capable of, even if it’s just a rough estimate.
To all those who are complaining that CSR means nothing because MMR is the only important stat - you guys need to go back and read ZaedynFel’s old posts more carefully. As you play more games in a season (usually 30-60 i think Josh said), your CSR will adjust to match your MMR’s expected skill value. So by the end of a season, CSR and MMR should be roughly in agreement for anyone who played enough games in that playlist.
> 2533274847840732;96:
> > 2533274813317074;94:
> > > 2533274968722829;93:
> > >
>
> The problem with a matchmaking system that has a hidden MMR to match players based on it’s predictions of who will win is completely flawed in the first place. If it is matching teams and has a projected winner then it isn’t doing its job of correctly matching teams. Because it should be matching people so closely that I can’t predict the winner.
Dude, you’re way over-simplifying matchmaking. Two teams will very rarely have the EXACT same team average MMR. Usually there will be a slight difference favoring one team. And based on the way CSR (or any Elo-like system) works, that slight difference in the teams’ average CSRs allows the computer to calculate the probability each team will win. So maybe the computer knows beforehand there’s a 70% chance team A will win. It doesn’t know for sure, there’s still a chance either team could win.
if you required two teams to have identical average MMRs, we’d all be waiting days to find a single game in matchmaking.
> 2533274906308237;100:
> To all those who are complaining that CSR means nothing because MMR is the only important stat - you guys need to go back and read ZaedynFel’s old posts more carefully. As you play more games in a season (usually 30-60 i think Josh said), your CSR will adjust to match your MMR’s expected skill value. So by the end of a season, CSR and MMR should be roughly in agreement for anyone who played enough games in that playlist.
>
>
>
>
> > 2533274847840732;96:
> > > 2533274813317074;94:
> > > > 2533274968722829;93:
> > > >
> >
> > The problem with a matchmaking system that has a hidden MMR to match players based on it’s predictions of who will win is completely flawed in the first place. If it is matching teams and has a projected winner then it isn’t doing its job of correctly matching teams. Because it should be matching people so closely that I can’t predict the winner.
>
> Dude, you’re way over-simplifying matchmaking. Two teams will very rarely have the EXACT same team average MMR. Usually there will be a slight difference favoring one team. And based on the way CSR (or any Elo-like system) works, that slight difference in the teams’ average CSRs allows the computer to calculate the probability each team will win. So maybe the computer knows beforehand there’s a 70% chance team A will win. It doesn’t know for sure, there’s still a chance either team could win.
>
> if you required two teams to have identical average MMRs, we’d all be waiting days to find a single game in matchmaking.
I agree with you completely, however my issue (if it can really be called that) is just with wishing we could see our MMR since that’s what matchmaking is based on. Especially for those times earlier in the season before CSR and MMR become closer to being in agreement. That way for the matches that appear to be uneven, you could look and see that based on the MMR, the match is a lot closer than the visible ranks might suggest.
> 2533274906308237;100:
> To all those who are complaining that CSR means nothing because MMR is the only important stat - you guys need to go back and read ZaedynFel’s old posts more carefully. As you play more games in a season (usually 30-60 i think Josh said), your CSR will adjust to match your MMR’s expected skill value. So by the end of a season, CSR and MMR should be roughly in agreement for anyone who played enough games in that playlist.
>
> > 2533274847840732;96:
> > > 2533274813317074;94:
> > > > 2533274968722829;93:
> > > >
> >
> > The problem with a matchmaking system that has a hidden MMR to match players based on it’s predictions of who will win is completely flawed in the first place. If it is matching teams and has a projected winner then it isn’t doing its job of correctly matching teams. Because it should be matching people so closely that I can’t predict the winner.
>
> Dude, you’re way over-simplifying matchmaking. Two teams will very rarely have the EXACT same team average MMR. Usually there will be a slight difference favoring one team. And based on the way CSR (or any Elo-like system) works, that slight difference in the teams’ average CSRs allows the computer to calculate the probability each team will win. So maybe the computer knows beforehand there’s a 70% chance team A will win. It doesn’t know for sure, there’s still a chance either team could win.
>
> if you required two teams to have identical average MMRs, we’d all be waiting days to find a single game in matchmaking.
Then why bother resetting the “ranks” every month or two for active players, especially those below 1800 CSR?? It just throws out all the hard work the players and the system has done to try to get MMR and CSR to parallel, which is basically the issue here. Then we have the same circus every season, followed by a litany of threads complaining about placements, quitting, smurfs, etc, and it never has a chance to resolve itself. Honestly, explaining the system that way doesn’t jive with seasonal resets. Is it an incentive for every player to play 30-60 games in any (or all) the ranked lists they hope to get a rank in that matches their skill? That means if they want decent matches in three ranked lists they need to start by playing 150 games every season? Then they need to be matched against other players who have also played that many games to start to get balanced matches with a fair representation of rank that matches skill. Good luck wth that. For everyone else there is a feeling of getting screwed, so they leave. Then you end up with a small population of monthly grinders who play against each other over and over and over. If it’s such a good system that needs data to reconcile the two values then stop breaking it every month/season. Looking at the current leaderboard for Slayer, I know there are multiple reasons why a player can go 75-1(wtf MM?) with a 1.9 KD and only be ranked 49th, but those explanations don’t make me anxious to queue up Slayer any time soon.
> 2533274889429520;99:
> > 2533274968722829;98:
> > > 2533274847840732;96:
> > > > 2533274813317074;94:
> > > > > 2533274968722829;93:
> > > > > I’d like to know why MMR remains hidden.
> > > >
> > > > Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
> >
> > My problem with the hidden system is that we are told it works a lot better than the visible system we are constantly teased and awarded with. The old “don’t pay any attention to the man behind the curtain” is annoying at best.
>
> Especially when we’re told that the MMR and CSR system is pretty much the same as the system used for Halo 3. But if I’m understanding correctly, the 1-50 ranking in Halo 3 was MMR (could definitely be wrong here) while Captain, General, etc represented CSR. If that’s the case, then we could see our MMR rather than keeping it hidden. I do wish we could see our MMR, or at least something similar to what has been mentioned in other updates about maybe showing an average MMR of each team.
>
>
> > 2533274968722829;97:
> > > 2533274813317074;94:
> > > > 2533274968722829;93:
> > > > I’d like to know why MMR remains hidden.
> > >
> > > Josh touched on that in this post, but has also talked about it in previous updates - essentially MMR is the estimation of your skill, what the system expect the outcome of your games to be. CSR is the visible representation of what you have actually earned. It’s honestly no different from Halo 3 which ran on pretty much the same system just with different names
> >
> > As far as the second link you referenced, players regularly post screen shots of CSR discrepancies in ranked matches wider than those Josh said H5 does not allow. I’m not going to research all of them, but I’ve commented on some of them in the past and that confusion contributes to the saltiness of the community and the quitting problem the game has. When players are shown two Onyx and two Golds, getting stomped by a Champ, an Onyx and two Plats they don’t know or care that some hidden algorithm thought their competitive skills were evenly matched.
>
> I can attest to this too. One that sticks out for me personally (please note that this has been a season or so ago and my own matches seem to have since gotten much more balanced) was a game of FFA I played in which I was a Platinum 4 and matched against 2 Onyx players (one was 1800 and something and the other was 1969), 2 Champion players (88 and 24), and an unranked player. While this was before the FFA update where ranks meant next to nothing, it was still a much wider gap than has been discussed (I’ve seen in previous updates that you can expect to match players within one full rank tier of your own rank). To be fair, I didn’t lose any CSR since the players who beat me were of a MUCH higher rank and I tied with the Champion 88 and beat the Onyx 1969). That may not be the best example since like I said, it was pre-update and ranks really didn’t tell you anything about a player’s skill, but that is just one example that sticks out. I also have a screenshot of a doubles game in which I was a Gold 2 or 3, my teammate was a Champion and we played against a team of two Onyx players (not sure of the specific ranks when it comes to where they fit within the Onyx and Champion tiers).
>
> I understand that CSR and MMR are sometimes different, but if there’s that big of a gap, then it would appear that something is wrong. I personally would like to see the most accurate representation of my skill possible. I know that MMR is said to be complicated to show since it’s not just a single number, but even if it was just shown as a stat on Waypoint rather than in-game and was depicted as something like “this is an estimate of what rank we believe you to be” or “this is an estimate of what we expect your rank to get to.” I just find it interesting and would like to know what the system expects me to be capable of, even if it’s just a rough estimate.
I know Josh was only referring to certain ranked play lists when he mentioned limits on CSR gaps, but I just took a screen shot of a 50-3 Slayer / Team Skirmish game (only one player from each team quit too) that matched three Champs and a Plat 6 vs a Gold 1, and three unranked players who were SR 12, 2, and 2. I know it’s unranked, but seriously? Is there any effort at all to balance that list? Anyone wanna play social ? 
So to simplify all of your CSR / MMR questions, I think one thing that would resolve a lot of these concerns is to show the MMR of each TEAM at the end of each match. So you would see the usual bundle of CSRs on each player, but above / below the teams you would see, e.g., Diamond 1 vs. Diamond 3 and realize the MMR was actually fair.
This is common in other games. In those games, you also sometimes see large CSR discrepancies, but the MMR gap looks fine.
In 99.99% of all matches (there may be occasional glitches in server communications), the average MMR gap is definitely within the bounds of “fine”
Most matches are within 300 MMR ( 1.0 standard deviation) unless it’s a playlist with a looser constraint.
So I’d like to show that MMR on both sides if I could. I don’t know if it’s something I can change anytime soon, but it’s something I think would reduce the communication confusion.