Forums / Community / Matchmaking Feedback & Discussion

[Locked] MATCHMAKING FEEDBACK UPDATE – November 5

OP ZaedynFel

  1. 1
  2. ...
  3. 11
  4. 12
  5. 13
  6. 14
  7. 15
  8. ...
  9. 27
But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
What are you searching? I select HCE, H2, H2A, 4v4, for slayer, snipers, and action sack and my matchmaking has been fairly quick.
Please use the dedicated feedback threads or support forums for any queries on MCC, thanks
I'm digging the Match Composer! It's actually cleaned up a lot of how hectic and discombobulated the social game types were! I'm extremely disappointed to see yet another update come and go without the inclusion of a voting system. I think map voting would go the extra distance in putting the map and game type cycling into the hands of the players. I hope to see it included asap in an update! And as always Reach and Firefight will be mentioned by me at every opportunity until I have them both!
LethalQ wrote:
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
What are you searching? I select HCE, H2, H2A, 4v4, for slayer, snipers, and action sack and my matchmaking has been fairly quick.
Mostly Halo 2 classic and I get bored from waiting so long that I watch YouTube and I finally find a game I even checked I have open Nat so that is not a problem I even called my ISP they said my WiFi is fine so it's gotta be the mcc servers I can't think of anything else of why it takes so long
LethalQ wrote:
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
What are you searching? I select HCE, H2, H2A, 4v4, for slayer, snipers, and action sack and my matchmaking has been fairly quick.
Mostly Halo 2 classic and I get bored from waiting so long that I watch YouTube and I finally find a game I even checked I have open Nat so that is not a problem I even called my ISP they said my WiFi is fine so it's gotta be the mcc servers I can't think of anything else of why it takes so long
I wish I can send you a video to show you how long it takes me
LethalQ wrote:
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
What are you searching? I select HCE, H2, H2A, 4v4, for slayer, snipers, and action sack and my matchmaking has been fairly quick.
Mostly Halo 2 classic and I get bored from waiting so long that I watch YouTube and I finally find a game I even checked I have open Nat so that is not a problem I even called my ISP they said my WiFi is fine so it's gotta be the mcc servers I can't think of anything else of why it takes so long
It more likely a low population thing. If you are in ranked it has to find others close to your rank. For social it has to find others looking for H2. If you are searching soical try and add another title or two and maybe game types for faster times. MCC doesn't have the storgest population especially players trying to play H2 and HCE.
LethalQ wrote:
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
What are you searching? I select HCE, H2, H2A, 4v4, for slayer, snipers, and action sack and my matchmaking has been fairly quick.
Mostly Halo 2 classic and I get bored from waiting so long that I watch YouTube and I finally find a game I even checked I have open Nat so that is not a problem I even called my ISP they said my WiFi is fine so it's gotta be the mcc servers I can't think of anything else of why it takes so long
East coast time, I've found the game is unplayable before 130 in the afternoon
LethalQ wrote:
LethalQ wrote:
I don't know if anyone else is experiencing this issue or doing anything about it but I waited 15 mins just to find a game and when I did it restarted then I waited 10 more minutes and I shut off my xbox this is definetly a matchmaking problem and this needs to be fixed ASAP I mean cmon it takes me not even 30 seconds just to find a game in halo reach and the mcc has been out for 4 years and its still a mess so 343 if you can hear me please please please fix the matchmaking its not fun waiting this long just to find one game and fyi its not a population issue this is a matchmaking bug that needs to be dealt with before anything else
MCC is hardly a mess anymore. You may want to go back and check it out . Especially matched composer. It is awesome.
And I tried match composer last night and it takes forever just to find a game and when I do find one it's slow as balls just to get the game started why can't anyone listen to me or should I just give up
<p>But why does it take me 15 mins just find one game when reach takes 30 seconds explain that cuz it ain't my internet</p>
What are you searching? I select HCE, H2, H2A, 4v4, for slayer, snipers, and action sack and my matchmaking has been fairly quick.
Mostly Halo 2 classic and I get bored from waiting so long that I watch YouTube and I finally find a game I even checked I have open Nat so that is not a problem I even called my ISP they said my WiFi is fine so it's gotta be the mcc servers I can't think of anything else of why it takes so long
East coast time, I've found the game is unplayable before 130 in the afternoon
Guys this is for Halo 5, not MCC.

Move to the MCC section for the feedback threads please and thanks.
Long search times and teams ruin BTB for me. BTB SF seems to have finally killed BTB.

I can’t play against BTB teams. I try to not quit out in general but BTB teams is where I draw the line. It’s such a horrid experience.

New Maps is probably too little too late but it’s worth a try. I posted about it forever ago but I was upset they removed a bunch of maps in the past from BTB.
ZaedynFel wrote:
eLantern wrote:
RzR J3ST3R wrote:
This Roy/Flamesword clinic on stream is why /min rates should not be used in Halo MMR. Slow and steady.
We recently did a TS2 batch run which significantly reduced the influence of kpm on rank, hence the reason tons of doubles players have moved up from mid-Diamond to Onyx and Champ.

The batch run also resulted in a massive increase in Doubles accuracy, pushing it over 70% compared to the under 60% it was at before.

Also, no, we won't change how we estimate skill in matchmaking since that's what players experience there, regardless of how things may play differently at event. Though the overall Ranks line up pretty well with who the top Pros are, at least within a tolerable delta in MMR. So even there, it's not far enough off for us to worry.

So high MMR players belong at events, and if they want to show who really deserves the top, they should go.
I am appreciative that the team is continually pulling levers to improve performance.

I did bold a portion above because there have been multiple instances now where you have made similar replies that suggests a bit of a disconnect between the lab bench and the end user. These plays are not unique to LAN events. They don't play differently. They are currently rewarded much differently and, in my opinion, we shouldn't have to go to events to have playing for the win rewarded.

I acknowledge that TS2 is a more accurate predictor than anything that has pre-dated it. The side-effect, intended or not, is that is has shifted what EV+ plays are in individual games to selfish plays. One could argue that it all averages out globally over time- and I'd counter that is likely only because your average user does not have a clue how TS2 works. "Power users" still mostly don't understand it, judging by the feedback threads. And maybe an educated Gold 4 player is probably less likely to change their play style to reach Platinum 1.

I do think it possible that if a) the team is able to find a solution to deconvolute the +/-15-1 problem it has been tweaking (a symptom of not relying on W/L) or b) a future UI shows CSR numbers beyond just onyx- players will eventually get wise and we all relive the quality-of-life problems of early Reach Arena.

I haven't shifted much from TS1 to TS2. But playing with an educated bunch, there is a noticeable shift in that quality-of-life.

There are big two problems I see:
1) Educated players realize that "playing for the win/team" is not always in their best interest. For example, rather than your more memorable 49-49 Halo standoff, players realize it is better to kill/die quickly. The current implementation is reliant upon the player being ignorant to how TS2 works for it.
2) TS2 solidifies a meta, and playing (and winning) outside of that that meta is punished. If you were to strip it all down, remove all of the +/-15 massaging, etc., and actually report the game's true success metrics vs a standard kill counter, the feedback would be abysmal. And it probably should be, because the predictor is re-defining success.

I offer no perfect alternatives. In my opinion, these rate-based metrics are fantastic for initial placement/bracketing matches. But beyond that, it should be W/L-heavy. Otherwise, you are reliant upon your ranking system remaining a big black box and having an uneducated population. Grind it out wins, shock upsets, sneaky plays, big moments that all make Halo exciting, ultimately all mean much less now, whether your average user realizes it yet or not. That's unfortunate.
ZaedynFel wrote:
eLantern wrote:
RzR J3ST3R wrote:
This Roy/Flamesword clinic on stream is why /min rates should not be used in Halo MMR. Slow and steady.
I am appreciative that the team is continually pulling levers to improve performance.

1). I did bold a portion above because there have been multiple instances now where you have made similar replies that suggests a bit of a disconnect between the lab bench and the end user. These plays are not unique to LAN events. They don't play differently. They are currently rewarded much differently and, in my opinion, we shouldn't have to go to events to have playing for the win rewarded.

2). I acknowledge that TS2 is a more accurate predictor than anything that has pre-dated it. The side-effect, intended or not, is that is has shifted what EV+ plays are in individual games to selfish plays. One could argue that it all averages out globally over time- and I'd counter that is likely only because your average user does not have a clue how TS2 works. "Power users" still mostly don't understand it, judging by the feedback threads. And maybe an educated Gold 4 player is probably less likely to change their play style to reach Platinum 1.

3). I do think it possible that if a) the team is able to find a solution to deconvolute the +/-15-1 problem it has been tweaking (a symptom of not relying on W/L) or b) a future UI shows CSR numbers beyond just onyx- players will eventually get wise and we all relive the quality-of-life problems of early Reach Arena.

4). I haven't shifted much from TS1 to TS2. But playing with an educated bunch, there is a noticeable shift in that quality-of-life.

5). There are big two problems I see:
1) Educated players realize that "playing for the win/team" is not always in their best interest. For example, rather than your more memorable 49-49 Halo standoff, players realize it is better to kill/die quickly. The current implementation is reliant upon the player being ignorant to how TS2 works for it.
2) TS2 solidifies a meta, and playing (and winning) outside of that that meta is punished. If you were to strip it all down, remove all of the +/-15 massaging, etc., and actually report the game's true success metrics vs a standard kill counter, the feedback would be abysmal. And it probably should be, because the predictor is re-defining success.

6). I offer no perfect alternatives. In my opinion, these rate-based metrics are fantastic for initial placement/bracketing matches. But beyond that, it should be W/L-heavy. Otherwise, you are reliant upon your ranking system remaining a big black box and having an uneducated population. Grind it out wins, shock upsets, sneaky plays, big moments that all make Halo exciting, ultimately all mean much less now, whether your average user realizes it yet or not. That's unfortunate.
1). Competitive tournament play is extremely different compared to the average MM gameplay. Extremely high level MM play might be similar but considering T2.0 is meant to measure skill for the entire population 343 can't base the skill ranking just on those high level players. I also want to point out that the winners of the H3 event, Gabriel and Fantasy, consider their gameplay very aggressive and fast paced. So while you cherry picked out one game from Flamesword and Roy of how they came back by playing slow to allow the other two to make mistakes the winners of the events literally said that they won the final series 3-0 because they played how they normally do instead of playing scared/slower.

2). In general the basic population is going to be/meant to be uneducated about the ranking system. Almost all video game ranking systems are even less transparent black boxes compared to H5. Additionally the metric used, Kills per minute, doesn't actually support selfish plays nearly as much as you think. You have said that you search solo to that you can get more kills and screw over your teammate. However, you aren't really screwing over your teammate since the faster/more you get kills the quicker the game ends causing both of your KPMs to go up. Just using a kills/score metric like Reach and 4 did was disastrous and caused selfish plays to be better than winning. KPM doesn't work like that though. Additionally, KPM was chosen because it predicted wins the best out of any metric, even win %. Why would you use a worse metric to predict wins?

3). The +/- 1/15 isnt used anymore. There was a CSR update a few weeks back that now has CSR update anywhere between -30 and 30 based on two metrics. The expected values of KPM, DPM, and whether you won or not compared to how you actually did, and your MMR vs the opponents average MMR. So the first brings your CSR closer to your MMR while the second brings your CSR closer to your enemy's MMR. This change rewards winning more than before and has allowed several doubles players who were stuck in diamond to move to onyx and champ where they actually belonged. However, in some playlists with bad population, such as FFA, that second part of CSR change could cause players to "lose" CSR unfairly which is why 343 put out a fix for specific playlists. Josh has also stated multiple times that he wished he could show more data in the UI post game report of H5 but there aren't enough devs to work on it since they are all assigned to future projects.

4). It would be interesting to see if the players you know who changed their playstyle to be more aggressive have been winning more often when compared to before.

5a). Again 343 put in a change that has caused winning to matter more than before. Additionally part of the MMR calculation for each game determines whether or not you should win. If you go against that calculation and lose when you are not supposed to because you rushed out in a 49-49 to end the game maybe ten seconds before it normally would have then your MMR will be negatively affected.
5b). T2.0 hasn't redefined the meta for H5. The meta for H5 at a high level has always been to play fast/aggressive to try to split spawn the enemy and collapse on them. T2.0 just put the fact that that meta was the most predictable way to win in the spotlight. Playing outside of that meta will cause you to statistically lose more games that you would have had you played H5 "correctly"

6). The issue with W/L heavy ranking is that it is much easier to exploit and doesn't give nearly as accurate of a skill ranking. We come down to the issue of whether the ranking system should report actual skill or just a rough measure of that. We had that before the CSR changes were put into place. Players were +/- 600 CSR pretty easily since they could be carried by better players to win more games. It's a hard decision and one that 343 can't take lightly since the current ranking system feels worse, especially back when CSR only changed by 1 or 15. Josh has mentioned that he would want a team rank to be included in Infinite that was only W/L based in addition to the individual rank which would be like we have today.
FYI, we have made a slight change to CSR updating that you probably won't notice unless you tend to play against teams whose average skill is much lower than yours.

Before, the update was CSR vs. avg(YourMMR, OpposingTeamMMR)

Now it is CSR vs. OpposingTeamMMR + (YourMMR - YourTeamMMR))

This results in the system correcting match difficulties so if a high player (1900) plays against 1500s with other 1500s, it will raise the opponents perceived skill to match the actual difficulty for the player. This results in high-skilled players having their CSRs converge on MMRs instead of on the midpoint between opponents and their own MMR. So top players won't be underestimated anymore.

This is similar to the oldest method, but without the inflation issues.
ZaedynFel wrote:
FYI, we have made a slight change to CSR updating that you probably won't notice unless you tend to play against teams whose average skill is much lower than yours.

Before, the update was CSR vs. avg(YourMMR, OpposingTeamMMR)

Now it is CSR vs. OpposingTeamMMR + (YourMMR - YourTeamMMR))

This results in the system correcting match difficulties so if a high player (1900) plays against 1500s with other 1500s, it will raise the opponents perceived skill to match the actual difficulty for the player. This results in high-skilled players having their CSRs converge on MMRs instead of on the midpoint between opponents and their own MMR. So top players won't be underestimated anymore.

This is similar to the oldest method, but without the inflation issues.
So, if I go back to this example:

Team 1 consists of:
  • Player A: CSR = 1750 | MMR = 1800
  • Player B: CSR = 1300 | MMR = 1250
  • Player C: CSR = 1250 | MMR = 1200
  • Player D: CSR = 1100 | MMR = 1050
Team 1 Average MMR = 1325

Team 2 consists of:
  • Player E: CSR = 1450 | MMR = 1400
  • Player F: CSR = 1300 | MMR = 1350
  • Player G: CSR = 1300 | MMR = 1250
  • Player H: CSR = 1150 | MMR = 1100
Team 2 Average MMR = 1275

Focusing on Player A, he/she is now faced with 1750 vs (1275+(1800-1325)) or 1750. 1750 vs 1750 means that whatever the match outcome is his/her CSR wouldn't shift drastically. Probably around +/- 1 CSR, but this is much more fair to the highest skilled player on the team in matches when they're paired with lower level players to balance a team.

Previously, it would have been 1750 vs ((1800+1275)/2) or 1537.5. 1750 vs 1537.5 means that a loss would likely result in a big -30 CSR impact while a win could only earn them a +1 CSR.

I'm glad that the concern I initially had regarding the previous method unfairly impacting the highest MMR/CSR individuals on a team has been addressed. I am wondering if this new method causes an inverse concern for the individuals with the lowest MMR/CSR on a team?

Focusing on Player D, he/she is now faced with 1100 vs (1275+(1050-1325)) or 1000. 1100 vs 1000 means that a loss would likely result in a big -30 CSR impact while a win could only earn them a +1 CSR, right? This seems somewhat okay because their Pre-Match CSR was already above their Pre-Match MMR, their team was expected to win, they had the best player in the game on their team, and this method lessens the possibility of being carried to inflated ranks by higher skilled teammates. I can certainly see reasons why this method is accepted as being much better overall.

Anyways, the primary issue that still remains is the MMR lag, but I'm aware that that won't get addressed within Halo 5. By the way, has the team devised a method to remove the skill-lag for Post-Match CSR adjustments within a future title (aka Infinite)?
QX wrote:
ZaedynFel wrote:
eLantern wrote:
RzR J3ST3R wrote:
This Roy/Flamesword clinic on stream is why /min rates should not be used in Halo MMR. Slow and steady.
1). Competitive tournament play is extremely different compared to the average MM gameplay. Extremely high level MM play might be similar but considering T2.0 is meant to measure skill for the entire population 343 can't base the skill ranking just on those high level players. I also want to point out that the winners of the H3 event, Gabriel and Fantasy, consider their gameplay very aggressive and fast paced. So while you cherry picked out one game from Flamesword and Roy of how they came back by playing slow to allow the other two to make mistakes the winners of the events literally said that they won the final series 3-0 because they played how they normally do instead of playing scared/slower.

2). In general the basic population is going to be/meant to be uneducated about the ranking system. Almost all video game ranking systems are even less transparent black boxes compared to H5. Additionally the metric used, Kills per minute, doesn't actually support selfish plays nearly as much as you think. You have said that you search solo to that you can get more kills and screw over your teammate. However, you aren't really screwing over your teammate since the faster/more you get kills the quicker the game ends causing both of your KPMs to go up. Just using a kills/score metric like Reach and 4 did was disastrous and caused selfish plays to be better than winning. KPM doesn't work like that though. Additionally, KPM was chosen because it predicted wins the best out of any metric, even win %. Why would you use a worse metric to predict wins?

3). The +/- 1/15 isnt used anymore. There was a CSR update a few weeks back that now has CSR update anywhere between -30 and 30 based on two metrics. The expected values of KPM, DPM, and whether you won or not compared to how you actually did, and your MMR vs the opponents average MMR. So the first brings your CSR closer to your MMR while the second brings your CSR closer to your enemy's MMR. This change rewards winning more than before and has allowed several doubles players who were stuck in diamond to move to onyx and champ where they actually belonged. However, in some playlists with bad population, such as FFA, that second part of CSR change could cause players to "lose" CSR unfairly which is why 343 put out a fix for specific playlists. Josh has also stated multiple times that he wished he could show more data in the UI post game report of H5 but there aren't enough devs to work on it since they are all assigned to future projects.

4). It would be interesting to see if the players you know who changed their playstyle to be more aggressive have been winning more often when compared to before.

5a). Again 343 put in a change that has caused winning to matter more than before. Additionally part of the MMR calculation for each game determines whether or not you should win. If you go against that calculation and lose when you are not supposed to because you rushed out in a 49-49 to end the game maybe ten seconds before it normally would have then your MMR will be negatively affected.
5b). T2.0 hasn't redefined the meta for H5. The meta for H5 at a high level has always been to play fast/aggressive to try to split spawn the enemy and collapse on them. T2.0 just put the fact that that meta was the most predictable way to win in the spotlight. Playing outside of that meta will cause you to statistically lose more games that you would have had you played H5 "correctly"

6). The issue with W/L heavy ranking is that it is much easier to exploit and doesn't give nearly as accurate of a skill ranking. We come down to the issue of whether the ranking system should report actual skill or just a rough measure of that. We had that before the CSR changes were put into place. Players were +/- 600 CSR pretty easily since they could be carried by better players to win more games. It's a hard decision and one that 343 can't take lightly since the current ranking system feels worse, especially back when CSR only changed by 1 or 15. Josh has mentioned that he would want a team rank to be included in Infinite that was only W/L based in addition to the individual rank which would be like we have today.
The "cherry picked" game was the same that Bravo and Onset went out of their way to describe how intelligent of a strategy it was, not just for that game but in general. I could have also picked the next game where Flamesword deviated from that "disciplined" play, unnecessarily pushed- and gave away 6-7 rockets- and was panned by the same commentators.

Or I could have pointed out the posts from several months ago that we discussed, where I showed that many of my best games were "slower", that the variance in success increased a great deal for the rates in question. I wasn't cherry picking anything. Aggressive play is great. It. is. not. the. only. way.

I don't know how to distill that any more simply, but here goes: When we used to do large data set analysis we often did something called gating. Essentially you looked for groups that were distinguishable from each other. Great when removing gunk from a data set. Absolutely terrible when you have two groups that look very different but perform the same function. In this case- win the game. You can win more than one way. That's my point, has always been my point. The current ranking system enforces and rewards one type, effectively gating it, regardless of the game outcome.

So yes, I would pick a slightly less accurate "worse" global predictor if it encompassed more ways of being effective. If it didn't dictate play towards something that may not actually lead to greater team success, I would prefer it every time. If it did not promote, at it's core, every game be a slightly cooperative FFA where the optimal play would often be stealing power weapons from teammates, etc, I would pick that every time. If you gave me two choices, A) a system that asked "were you successful?", or B) a system that asked "did you make a play that based on the entire population you were likely to be successful?" I pick the result based system... tremendous value in System B, but ranks should be outcome based, IMO.

There must be alternatives middle ground approaches. Using kill rates exclusively for placement matches. Using kill rates to determine those exploits (oh, this guy is performing 2 levels below the rest of his team EVERY game they win). Skill caps in parties. Etc. etc.
QX wrote:
ZaedynFel wrote:
eLantern wrote:
RzR J3ST3R wrote:
This Roy/Flamesword clinic on stream is why /min rates should not be used in Halo MMR. Slow and steady.
The "cherry picked" game was the same that Bravo and Onset went out of their way to describe how intelligent of a strategy it was, not just for that game but in general. I could have also picked the next game where Flamesword deviated from that "disciplined" play, unnecessarily pushed- and gave away 6-7 rockets- and was panned by the same commentators.

Or I could have pointed out the posts from several months ago that we discussed, where I showed that many of my best games were "slower", that the variance in success increased a great deal for the rates in question. I wasn't cherry picking anything. Aggressive play is great. It. is. not. the. only. way.

I don't know how to distill that any more simply, but here goes: When we used to do large data set analysis we often did something called gating. Essentially you looked for groups that were distinguishable from each other. Great when removing gunk from a data set. Absolutely terrible when you have two groups that look very different but perform the same function. In this case- win the game. You can win more than one way. That's my point, has always been my point. The current ranking system enforces and rewards one type, effectively gating it, regardless of the game outcome.

So yes, I would pick a slightly less accurate "worse" global predictor if it encompassed more ways of being effective. If it didn't dictate play towards something that may not actually lead to greater team success, I would prefer it every time. If it did not promote, at it's core, every game be a slightly cooperative FFA where the optimal play would often be stealing power weapons from teammates, etc, I would pick that every time. If you gave me two choices, A) a system that asked "were you successful?", or B) a system that asked "did you make a play that based on the entire population you were likely to be successful?" I pick the result based system... tremendous value in System B, but ranks should be outcome based, IMO.

There must be alternatives middle ground approaches. Using kill rates exclusively for placement matches. Using kill rates to determine those exploits (oh, this guy is performing 2 levels below the rest of his team EVERY game they win). Skill caps in parties. Etc. etc.
Originally you mentioned the one game where Roy & Flame came back because of passive gameplay and how that should show 343 that their system is lacking. Ultimately Optic lost due to a passive playstyle against GMS demonstrated by GMS' 3-0 win in the final series. Also its H3 vs H5 which play significantly different even in doubles simply because of the evasion mechanics in H5.

As I stated when we had that conversation without MMR data it is hard to determine that the best games you had were in terms ones where you played slower. You may also play better in a slower game for a multitude of reasons. Aggressive play is not the only way but players who play more aggressive win more than slower players on average in H5. That's my issue with you saying that aggressive play shouldn't be rewarded as much as any other playstyle because they both end up with the same result of winning the game. If both aggressive playstyles and passive playstyles won games at the same rate then using KPM would be a bad measure of skill. But Josh has specifically stated that players who achieve a higher KPM (caused by an aggressive playstyle) win games at a high rate than passive players. I clearly haven't had the same kind of statistical/data analysis background/experience that you have. I just don't understand how both passive and aggressive playstyles can be rewarded the same if they don't come to the same conclusion at the same rates.

343 has already shown that they are trying to find a middle ground approach considering the removal of only 1/15 and the addition of the second part of CSR change mentioned in Josh's reply.
QX wrote:
QX wrote:
ZaedynFel wrote:
eLantern wrote:
RzR J3ST3R wrote:
This Roy/Flamesword clinic on stream is why /min rates should not be used in Halo MMR. Slow and steady.
https://www.reddit.com/r/halo/comments/4p52te/kills_and_deaths_per_minute_in_the_pro_league/

Would you rather be Lethul or Ninja?
I use this only for illustrative purposes. Both are excellent, top 0.01% tier- and I assume even these metrics support that. Only use this as a relative example.

Globally, I am certain aggressive playstyles outpace passive or anchoring playstyles. No argument. But there are going to really good "slow" players that are probably better than your aggressive players of that rank- and TS2 won't reflect that. Some cases obviously more extreme than others, and in most cases, changes will probably be small (within the delta, as Josh put it).

And that's fine. No system is going to be perfect. I just very much prefer the system where I don't have a voice in my head to fast play every engagement vs. playing each situation for the win, even if it's slow. I have played games in every Halo where I know the players on the other team, I know he has a better shot than me and I can't just fly into an encounter- but I know I can bait them, out-think them and sometimes steal a win. That's not rewarded much now. From a game quality perspective, I think it's a negative.

So for those reasons like that, I am comfortable trading a couple points of global accuracy. That is my opinion, and I'd guess would be a lot more players if they understood the system.
And that's fine. No system is going to be perfect. I just very much prefer the system where I don't have a voice in my head to fast play every engagement vs. playing each situation for the win, even if it's slow. I have played games in every Halo where I know the players on the other team, I know he has a better shot than me and I can't just fly into an encounter- but I know I can bait them, out-think them and sometimes steal a win. That's not rewarded much now. From a game quality perspective, I think it's a negative.
That's not true.
  1. The CSR system still only rewards based on wins. It will self-correct to adjust to one's assessed skill, but match outcomes (W/L) are still the only manner in which you can increase or decrease your rank.
  2. The TS2/MMR system is still primarily based on match outcomes despite the inclusion of individual KPM (and to a much lesser degree DPM) influences that can sway some adjustments in opposition of the positive or negative team adjustment. Team adjustments remain restricted by match outcomes (W/L). The purpose of individual influences like KPM is to further identify personal skill amongst team skill which directly leads to increased predictive accuracy.
  3. From a game quality perspective, more accurate skill-based matchmaking (SBMM) is the most critical component. And from an individual rank perspective, rank integrity is [certainly one of] the most critical component[s]. These two things are, more or less, being achieved now.
QX wrote:
QX wrote:
ZaedynFel wrote:
eLantern wrote:
RzR J3ST3R wrote:
This Roy/Flamesword clinic on stream is why /min rates should not be used in Halo MMR. Slow and steady.
https://www.reddit.com/r/halo/comments/4p52te/kills_and_deaths_per_minute_in_the_pro_league/

Would you rather be Lethul or Ninja?
I use this only for illustrative purposes. Both are excellent, top 0.01% tier- and I assume even these metrics support that. Only use this as a relative example.

Globally, I am certain aggressive playstyles outpace passive or anchoring playstyles. No argument. But there are going to really good "slow" players that are probably better than your aggressive players of that rank- and TS2 won't reflect that. Some cases obviously more extreme than others, and in most cases, changes will probably be small (within the delta, as Josh put it).

And that's fine. No system is going to be perfect. I just very much prefer the system where I don't have a voice in my head to fast play every engagement vs. playing each situation for the win, even if it's slow. I have played games in every Halo where I know the players on the other team, I know he has a better shot than me and I can't just fly into an encounter- but I know I can bait them, out-think them and sometimes steal a win. That's not rewarded much now. From a game quality perspective, I think it's a negative.

So for those reasons like that, I am comfortable trading a couple points of global accuracy. That is my opinion, and I'd guess would be a lot more players if they understood the system.
I guess my only worry would be how much of a loss to global accuracy we would take. Personally going back to the old system (I know you are not advocating for that just an example) which was more W/L would be worse for rank integrity. I'm sure most players would be fine with it since they could achieve higher ranks. Several of my friends were able to get champions in various playlists that they absolutely didn't deserve because CSR was able to drift far from MMR.

As a side note I think it is dangerous to think that playing every engagement fast will inherently increase rank. If you are a better passive player (myself included since my positioning, movement, and spawn knowledge much better than my comparative shot) then Im willing to bet playing passive will net you a better KPM than playing a Ninja playstyle would.
  1. 1
  2. ...
  3. 11
  4. 12
  5. 13
  6. 14
  7. 15
  8. ...
  9. 27