Skip to main content

Forums / Community / Matchmaking Feedback & Discussion

Matchmaking Update 3/17/2020

OP ZaedynFel

  1. 1
  2. ...
  3. 35
  4. 36
  5. 37
  6. 38
  7. ...
  8. 39
B4 I unlocked the gunfighter mag, I dreamed that it would be with me at all times😒. Is it possible that a patch can make my dream come true? I know, it's like im 5.lol🦸🏿‍♂️🍬🚂🚀🎈
Thanks for the update you guys are awesome
AR: The Ar can be buffed a little bit making it a bit useful and not useless.

Battle Rifle: increase range

magnum: all fine

smg: they can increase the range just a little

Dmr: all fine
This post has been edited by a moderator. Please do not post spam.
*Original post. Click at your own discretion.
Spoiler:
Show
When we going to make the game fun and playable for servers other than west US
This post has been edited by a moderator. Because we are not currently equipped to provide moderation in other languages, we must ask that all messages be posted in English. Feel free to use an online language translation service, and then create a new thread with the translated text. Sorry for the inconvenience.
cuando la gente deje de ser tóxica
This post has been edited by a moderator. Please refrain from making posts that do not contribute to the topic at hand.
*Original post. Click at your own discretion.
Spoiler:
Show
SNOWOLF409 wrote:
This post has been edited by a moderator. Because we are not currently equipped to provide moderation in other languages, we must ask that all messages be posted in English. Feel free to use an online language translation service, and then create a new thread with the translated text. Sorry for the inconvenience.cuando la gente deje de ser tóxica
Zaedynfel, Will individual performance be considered in the future? I would be curious and speculate how much more accurate matchmaking might get with individual statistics being considered. Like one individuals shooting accuracy, wins on particular maps, individual performance with specific weapons on maps, % of kills from things like grenades, etc

I would speculate that this data would ultimately make for better matchmaking statistical predictions.

There seems to be an overall sense from the community below maybe Onyx and Champion levels of an imbalance of matchmaking. I imagine the goal is to get matches as close to 50% in fairness of wins/loss as possible. With a statistical change of +1/-1 being off from that 50% being the ideal closest fairness imbalance.

Predictability of matchmaking for an individual match would be daunting understandably but I would speculate more data of the individual would start to get you closer to the 50% matchmaking goal.
Recourser wrote:
Zaedynfel, Will individual performance be considered in the future? I would be curious and speculate how much more accurate matchmaking might get with individual statistics being considered. Like one individuals shooting accuracy, wins on particular maps, individual performance with specific weapons on maps, % of kills from things like grenades, etc

I would speculate that this data would ultimately make for better matchmaking statistical predictions.

There seems to be an overall sense from the community below maybe Onyx and Champion levels of an imbalance of matchmaking. I imagine the goal is to get matches as close to 50% in fairness of wins/loss as possible. With a statistical change of +1/-1 being off from that 50% being the ideal closest fairness imbalance.

Predictability of matchmaking for an individual match would be daunting understandably but I would speculate more data of the individual would start to get you closer to the 50% matchmaking goal.
Individual performance is considered. It’s just kills per minute (tracked for each game type) vs the strength of opponents though. They’ve tested all those other individual stats and none worked as well as KPM only. The imbalance in matchmaking is due mostly to population, not the ranking system.

EDIT: Also, Zaedynfel doesn't work for 343 anymore.
IGRS wrote:
Recourser wrote:
Zaedynfel, Will individual performance be considered in the future? I would be curious and speculate how much more accurate matchmaking might get with individual statistics being considered. Like one individuals shooting accuracy, wins on particular maps, individual performance with specific weapons on maps, % of kills from things like grenades, etc

I would speculate that this data would ultimately make for better matchmaking statistical predictions.

There seems to be an overall sense from the community below maybe Onyx and Champion levels of an imbalance of matchmaking. I imagine the goal is to get matches as close to 50% in fairness of wins/loss as possible. With a statistical change of +1/-1 being off from that 50% being the ideal closest fairness imbalance.

Predictability of matchmaking for an individual match would be daunting understandably but I would speculate more data of the individual would start to get you closer to the 50% matchmaking goal.
Individual performance is considered. It’s just kills per minute (tracked for each game type) vs the strength of opponents though. They’ve tested all those other individual stats and none worked as well as KPM only. The imbalance in matchmaking is due mostly to population, not the ranking system.

EDIT: Also, Zaedynfel doesn't work for 343 anymore.
They tested all individual stats? There are so many variables to consider and comparable items to track that could help improve things. But I can see the population size being an issue so I understand that.

as for Zaedynfel I didn’t know. He started the thread so I assumed but thanks.

Unfortunately for me I just started playing Halo 5 about 4 months ago so I haven’t experienced this balance in matchmaking ( most likely due to what you mentioned the population size) but I am hoping for a Much fairer situation when Infinite comes out so I’m trying to get ahead of the curve.

thank you for your insights.
I think the ranking and the matching of players is about right.
Sam O34 wrote:
I think the ranking and the matching of players is about right.
The issue is there are many players whose win rate is far below 50% who are decent players and have been playing for some time. That in my opinion is an issue. Anything less than 49% win rate (over time) should not happen, certainly not in an ideal scenario. And winning above 51% isn’t necessary ideal either because you should be getting challenged more.
Recourser wrote:
IGRS wrote:
Recourser wrote:
Zaedynfel, Will individual performance be considered in the future? I would be curious and speculate how much more accurate matchmaking might get with individual statistics being considered. Like one individuals shooting accuracy, wins on particular maps, individual performance with specific weapons on maps, % of kills from things like grenades, etc

I would speculate that this data would ultimately make for better matchmaking statistical predictions.

There seems to be an overall sense from the community below maybe Onyx and Champion levels of an imbalance of matchmaking. I imagine the goal is to get matches as close to 50% in fairness of wins/loss as possible. With a statistical change of +1/-1 being off from that 50% being the ideal closest fairness imbalance.

Predictability of matchmaking for an individual match would be daunting understandably but I would speculate more data of the individual would start to get you closer to the 50% matchmaking goal.
Individual performance is considered. It’s just kills per minute (tracked for each game type) vs the strength of opponents though. They’ve tested all those other individual stats and none worked as well as KPM only. The imbalance in matchmaking is due mostly to population, not the ranking system.

EDIT: Also, Zaedynfel doesn't work for 343 anymore.
They tested all individual stats? There are so many variables to consider and comparable items to track that could help improve things. But I can see the population size being an issue so I understand that.

as for Zaedynfel I didn’t know. He started the thread so I assumed but thanks.

Unfortunately for me I just started playing Halo 5 about 4 months ago so I haven’t experienced this balance in matchmaking ( most likely due to what you mentioned the population size) but I am hoping for a Much fairer situation when Infinite comes out so I’m trying to get ahead of the curve.

thank you for your insights.
Yep. They’ve experimented with many individual stats — from what I’ve seen they led to less accurate predictions compared to using solely KPM.

I think we’ll all be satisfied with how the system works with Infinite and the (hopefully, likely) enormous population from the new game.
IGRS wrote:
Recourser wrote:
IGRS wrote:
Recourser wrote:
Zaedynfel, Will individual performance be considered in the future? I would be curious and speculate how much more accurate matchmaking might get with individual statistics being considered. Like one individuals shooting accuracy, wins on particular maps, individual performance with specific weapons on maps, % of kills from things like grenades, etc

I would speculate that this data would ultimately make for better matchmaking statistical predictions.

There seems to be an overall sense from the community below maybe Onyx and Champion levels of an imbalance of matchmaking. I imagine the goal is to get matches as close to 50% in fairness of wins/loss as possible. With a statistical change of +1/-1 being off from that 50% being the ideal closest fairness imbalance.

Predictability of matchmaking for an individual match would be daunting understandably but I would speculate more data of the individual would start to get you closer to the 50% matchmaking goal.
Individual performance is considered. It’s just kills per minute (tracked for each game type) vs the strength of opponents though. They’ve tested all those other individual stats and none worked as well as KPM only. The imbalance in matchmaking is due mostly to population, not the ranking system.

EDIT: Also, Zaedynfel doesn't work for 343 anymore.
They tested all individual stats? There are so many variables to consider and comparable items to track that could help improve things. But I can see the population size being an issue so I understand that.

as for Zaedynfel I didn’t know. He started the thread so I assumed but thanks.

Unfortunately for me I just started playing Halo 5 about 4 months ago so I haven’t experienced this balance in matchmaking ( most likely due to what you mentioned the population size) but I am hoping for a Much fairer situation when Infinite comes out so I’m trying to get ahead of the curve.

thank you for your insights.
Yep. They’ve experimented with many individual stats — from what I’ve seen they led to less accurate predictions compared to using solely KPM.

I think we’ll all be satisfied with how the system works with Infinite and the (hopefully, likely) enormous population from the new game.
To clarify, it's both KPM & DPM that are used, but KPM is significantly weighted over DPM.

And as you explained they did experiment with vast amounts of individual stats, pretty much all as far as I'm aware, including every medal. But, perhaps worth mentioning is that a couple objective-based metrics did show some minor predictive results that merited them for inclusion consideration. In the end, they opted not to include them because of concerns that they'd end up being exploitable by players and their impact toward match result predictions were so minor that they were deemed essentially negligible. If I remember correctly it was "Flags Pulled" and something else which I'm forgetting at the moment but it related to one of the other modes... hmmmm... Bomb, Oddball, Strongholds.... ah... I can't recall. I'd have to go digging through all the communications I had with Josh to find it.

The other thing Josh mentioned was that all the testing they did demonstrated that overly complicating the formula with a multitude of statistical metrics initially thought to be important introduced copious amounts of data redundancy regarding player skill that then led to less accurate predictive results. I know I was one such individual that found that hard to believe. But, in the end I had to acknowledge that the facts are the facts. So it's either accept the truth or become someone who simply can't accept the reality of the matter. I've seen a crap ton of the latter constantly try telling Josh what the facts really are without any credibility other than what they imagine they possess.

P.S. I miss Josh working on Halo.
eLantern wrote:
IGRS wrote:
Recourser wrote:
IGRS wrote:
Recourser wrote:
Zaedynfel, Will individual performance be considered in the future? I would be curious and speculate how much more accurate matchmaking might get with individual statistics being considered. Like one individuals shooting accuracy, wins on particular maps, individual performance with specific weapons on maps, % of kills from things like grenades, etc

I would speculate that this data would ultimately make for better matchmaking statistical predictions.

There seems to be an overall sense from the community below maybe Onyx and Champion levels of an imbalance of matchmaking. I imagine the goal is to get matches as close to 50% in fairness of wins/loss as possible. With a statistical change of +1/-1 being off from that 50% being the ideal closest fairness imbalance.

Predictability of matchmaking for an individual match would be daunting understandably but I would speculate more data of the individual would start to get you closer to the 50% matchmaking goal.
Individual performance is considered. It’s just kills per minute (tracked for each game type) vs the strength of opponents though. They’ve tested all those other individual stats and none worked as well as KPM only. The imbalance in matchmaking is due mostly to population, not the ranking system.

EDIT: Also, Zaedynfel doesn't work for 343 anymore.
They tested all individual stats? There are so many variables to consider and comparable items to track that could help improve things. But I can see the population size being an issue so I understand that.

as for Zaedynfel I didn’t know. He started the thread so I assumed but thanks.

Unfortunately for me I just started playing Halo 5 about 4 months ago so I haven’t experienced this balance in matchmaking ( most likely due to what you mentioned the population size) but I am hoping for a Much fairer situation when Infinite comes out so I’m trying to get ahead of the curve.

thank you for your insights.
Yep. They’ve experimented with many individual stats — from what I’ve seen they led to less accurate predictions compared to using solely KPM.

I think we’ll all be satisfied with how the system works with Infinite and the (hopefully, likely) enormous population from the new game.
To clarify, it's both KPM & DPM that are used, but KPM is significantly weighted over DPM.

And as you explained they did experiment with vast amounts of individual stats, pretty much all as far as I'm aware, including every medal. But, perhaps worth mentioning is that a couple objective-based metrics did show some minor predictive results that merited them for inclusion consideration. In the end, they opted not to include them because of concerns that they'd end up being exploitable by players and their impact toward match result predictions were so minor that they were deemed essentially negligible. If I remember correctly it was "Flags Pulled" and something else which I'm forgetting at the moment but it related to one of the other modes... hmmmm... Bomb, Oddball, Strongholds.... ah... I can't recall. I'd have to go digging through all the communications I had with Josh to find it.

The other thing Josh mentioned was that all the testing they did demonstrated that overly complicating the formula with a multitude of statistical metrics initially thought to be important introduced copious amounts of data redundancy regarding player skill that then led to less accurate predictive results. I know I was one such individual that found that hard to believe. But, in the end I had to acknowledge that the facts are the facts. So it's either accept the truth or become someone who simply can't accept the reality of the matter. I've seen a crap ton of the latter constantly try telling Josh what the facts really are without any credibility other than what they imagine they possess.

P.S. I miss Josh working on Halo.
At the end of the day, in my opinion, all that matters is if they were able to reach the optimal goals of 50% win loss ratio with a +1/-1 variability. If they did (minus our current low population matchmaking issue) then great, if not then there should be another approach considered.

I’m a believer in that there are no statistical probability predictions that can’t be achieved. Obviously with humans there are a lot of variables, unpredictables, but we do possess long term patterns.

My belief isn’t based on opinion but proof of concept in scientific results. Based on what you’ve described I believe they have done a great job of working through a lot of the data. And ultimately if they were able to achieve that 50% win rate with +1/-1 variability then they achieved what they should have.

I’m not saying I’m right over them, or that I don’t accept the facts they came to, I just know that if coming to a statistical prediction projection that didn’t achieve this goal then there was still possibility/predictability’s to be worked out.

For now, I have faith that they have and am hopeful that Infinite is at this optimal fair matchmaking. Because this will ultimately help hold players motivation to play. If they see, statistically winning half their matches, they will feel satisfied and will want to keep on playing.

There will always be those who still think things aren’t fair even with these facts because they unrealistically want to win some 75% or some very high % of their matches. This kind of win rate should only be possible at the Champion/Pro level. Because at that level theoretically those players could reach levels that the system isn’t quite tracking. Which would explain Sentinels winning as often as they do, as long as players like Frosty continue to improve themselves regardless of the system they can continue to statistically beat the competition.

Ones personal feelings about the statistics doesn’t change the facts of them, no matter how much one may want them to be in their favor. The only reasonable way one should expect to win more than they lose is if they mark ably improve their performance over the population within their own current rank. Which is why this new training and AI bot system they have developed could potentially be one of the best things they have put together if done right.

If ranking up was the equivalent of playing campaign and working your way against a new level that ultimately is the same but still has its unpredictability. Then achieving the next rank will be and feel more like beating a new level on campaign which is so perfect and amazing. Obviously you would still have to rank up by playing the humans who are far more unpredictable but this will drive you so much closer to achieving this that you will be able to taste the next rank.

Anyway, Thank you for your insights and knowledge in this matter.
This post has been edited by a moderator. Because we are not currently equipped to provide moderation in other languages, we must ask that all messages be posted in English. Feel free to use an online language translation service, and then create a new thread with the translated text. Sorry for the inconvenience.
hasta la hora me sigue emparejando con campeones cuando apenas soy platino u oro :(
Recourser wrote:
eLantern wrote:
Quote:
...
To clarify, it's both KPM & DPM that are used, but KPM is significantly weighted over DPM.

And as you explained they did experiment with vast amounts of individual stats, pretty much all as far as I'm aware, including every medal. But, perhaps worth mentioning is that a couple objective-based metrics did show some minor predictive results that merited them for inclusion consideration. In the end, they opted not to include them because of concerns that they'd end up being exploitable by players and their impact toward match result predictions were so minor that they were deemed essentially negligible. If I remember correctly it was "Flags Pulled" and something else which I'm forgetting at the moment but it related to one of the other modes... hmmmm... Bomb, Oddball, Strongholds.... ah... I can't recall. I'd have to go digging through all the communications I had with Josh to find it.

The other thing Josh mentioned was that all the testing they did demonstrated that overly complicating the formula with a multitude of statistical metrics initially thought to be important introduced copious amounts of data redundancy regarding player skill that then led to less accurate predictive results. I know I was one such individual that found that hard to believe. But, in the end I had to acknowledge that the facts are the facts. So it's either accept the truth or become someone who simply can't accept the reality of the matter. I've seen a crap ton of the latter constantly try telling Josh what the facts really are without any credibility other than what they imagine they possess.

P.S. I miss Josh working on Halo.
At the end of the day, in my opinion, all that matters is if they were able to reach the optimal goals of 50% win loss ratio with a +1/-1 variability. If they did (minus our current low population matchmaking issue) then great, if not then there should be another approach considered.
To be clear, there are multiple systems and particular decisions at work within Halo 5.

There's the player skill assessor system, referred to as TrueSkill2, which uses advanced machine learning to predict match outcomes based on statistical understandings of individual player skill through time. The fantastic news in regards to TrueSkill2 is that its predictive accuracy is around (and likely above) 70% which is pretty amazing given the accuracy of most other predictive skill systems struggle to reach 60% (like the one being used in the MCC).

There's the visual competitive skill ranks (CSRs), which are earned within the "Ranked" arena environment and are designed to more or less reflect and follow a player's TrueSkill metric, but with some different rulesets aimed at curbing volatility and rewarding players strictly on W/L outcomes while preventing runaway rank inflation/deflation from one's TrueSkill. It's not really designed to be a lengthy experience grind to converge your visual rank to your actual skill unless you're a surefire Onyx level player striving to earn and maintain a Champ rank.

There's the matchmaking system which was recently (in early-mid 2020) updated to incorporate machine learning. This "smart" matchaker is referred to as TrueMatch. This system has dramatically sped up the time it takes to find matches by learning what search parameters relating to skill range & connection quality (by determining what data centers should be available) are ideal per region and time of day in order to keep searches from typically going beyond 5 minutes; however, it's also been designed to allow massively lopsided matches to be created in order to give all players -- regardless of skill difference -- an opportunity to play.

And there's the team picker/balancer system which aims to even out and balance the players & fireteams as best as possible that the matchmaker pulls into a lobby based on the match outcome predictions derived from the TrueSkill2 system. I believe this may also be the system that applies the skill value boost based on party size in order to compensate for the assumed bump in skill that tends to occur when people are playing in a group instead of solo and communicates this additional information to the matchmaker.

So the system in whole isn't strictly designed to ensure all matches possess 50:50 odds, but instead its aim is to shoot for those match odds whenever the best case scenario is present and to do so with some of the most advanced and accurate systems available in the industry. The problem is that the available player population at any given moment when a player choses to search often doesn't allow for that best case scenario to be their reality; therefore, you get a fair share of frustrating situations where match balance feels unfair and/or connections may be somewhat poor. These things certainly lead to players possessing win percentages that can go well beyond or under 50%.
Quote:
Anyway, Thank you for your insights and knowledge in this matter.
No problem.
eLantern wrote:
Recourser wrote:
eLantern wrote:
Quote:
...
To clarify, it's both KPM & DPM that are used, but KPM is significantly weighted over DPM.

And as you explained they did experiment with vast amounts of individual stats, pretty much all as far as I'm aware, including every medal. But, perhaps worth mentioning is that a couple objective-based metrics did show some minor predictive results that merited them for inclusion consideration. In the end, they opted not to include them because of concerns that they'd end up being exploitable by players and their impact toward match result predictions were so minor that they were deemed essentially negligible. If I remember correctly it was "Flags Pulled" and something else which I'm forgetting at the moment but it related to one of the other modes... hmmmm... Bomb, Oddball, Strongholds.... ah... I can't recall. I'd have to go digging through all the communications I had with Josh to find it.

The other thing Josh mentioned was that all the testing they did demonstrated that overly complicating the formula with a multitude of statistical metrics initially thought to be important introduced copious amounts of data redundancy regarding player skill that then led to less accurate predictive results. I know I was one such individual that found that hard to believe. But, in the end I had to acknowledge that the facts are the facts. So it's either accept the truth or become someone who simply can't accept the reality of the matter. I've seen a crap ton of the latter constantly try telling Josh what the facts really are without any credibility other than what they imagine they possess.

P.S. I miss Josh working on Halo.
At the end of the day, in my opinion, all that matters is if they were able to reach the optimal goals of 50% win loss ratio with a +1/-1 variability. If they did (minus our current low population matchmaking issue) then great, if not then there should be another approach considered.
To be clear, there are multiple systems and particular decisions at work within Halo 5.

There's the player skill assessor system, referred to as TrueSkill2, which uses advanced machine learning to predict match outcomes based on statistical understandings of individual player skill through time. The fantastic news in regards to TrueSkill2 is that its predictive accuracy is around (and likely above) 70% which is pretty amazing given the accuracy of most other predictive skill systems struggle to reach 60% (like the one being used in the MCC).

There's the visual competitive skill ranks (CSRs), which are earned within the "Ranked" arena environment and are designed to more or less reflect and follow a player's TrueSkill metric, but with some different rulesets aimed at curbing volatility and rewarding players strictly on W/L outcomes while preventing runaway rank inflation/deflation from one's TrueSkill. It's not really designed to be a lengthy experience grind to converge your visual rank to your actual skill unless you're a surefire Onyx level player striving to earn and maintain a Champ rank.

There's the matchmaking system which was recently (in early-mid 2020) updated to incorporate machine learning. This "smart" matchaker is referred to as TrueMatch. This system has dramatically sped up the time it takes to find matches by learning what search parameters relating to skill range & connection quality (by determining what data centers should be available) are ideal per region and time of day in order to keep searches from typically going beyond 5 minutes; however, it's also been designed to allow massively lopsided matches to be created in order to give all players -- regardless of skill difference -- an opportunity to play.

And there's the team picker/balancer system which aims to even out and balance the players & fireteams as best as possible that the matchmaker pulls into a lobby based on the match outcome predictions derived from the TrueSkill2 system. I believe this may also be the system that applies the skill value boost based on party size in order to compensate for the assumed bump in skill that tends to occur when people are playing in a group instead of solo and communicates this additional information to the matchmaker.

So the system in whole isn't strictly designed to ensure all matches possess 50:50 odds, but instead its aim is to shoot for those match odds whenever the best case scenario is present and to do so with some of the most advanced and accurate systems available in the industry. The problem is that the available player population at any given moment when a player choses to search often doesn't allow for that best case scenario to be their reality; therefore, you get a fair share of frustrating situations where match balance feels unfair and/or connections may be somewhat poor. These things certainly lead to players possessing win percentages that can go well beyond or under 50%.
Quote:
Anyway, Thank you for your insights and knowledge in this matter.
No problem.
This is a very thorough and helpful explanation. Which ultimately excites me for Halo Infinite where more fair matchmaking will be commonplace and keep one motivated.

I personally have found myself discouraged on more than one occasion and I have just started a few months ago. And I have seen this discouragement become more common amongst many players which led me to look into this whole matchmaking to begin with.

Wanting to be able to decide whether to drop Halo all together or not because of this unfair matchmaking but I wanted to look into all angles before making this decision. I also wanted to see how much 343i cares about their Halo players and listens to player feedback.

Even though you may not be 343i you have helped me see that they do care and listen to players feedback And specifically helped me not drop Halo but look forward to Infinite which will inevitably have the population size to allow for far more fair matchmaking. I’ve decided on viewing Halo 5 more casually going forward.

I’ve always been a competitive player and I haven’t played Halo since Halo2. Coming back I came in with a little past experience and expectations. Also, I’m big believer that if one is adaptive, open to criticism, and looks at their own gameplay and decision making and make improvements one can always get better and so over time inevitably rank up. But this has been an unnecessary struggle. I’ve increased my Sniping skills, my grenade skills, my movement, my decision making, my aim, and yet as soon as I get one game away from ranking up I get poorly matched and drop right back down.

Its a frustration that led me to have to make a decision to either keep playing or not. One can only stand so much frustration especially when making so many improvements with what feels like no rank/reward. A gamer has to feel like their improvements mean something but if all your improvements don’t show up in results over time then one has to ask if there is a point. Will continued improvement make a difference or will you just get dropped right back down regardless of how much you improve.

So thank you eLantern and IGRS for all your insights. I’ll now play with little expectations of rank change while working on my game improvements regardless of win results. Though I hope Halo 5 hasn’t lost too many players with its matchmaking issue. I don’t think a lot of people will understand what’s going on and then find this information to come to understand it.

Here’s to hoping Infinite brings in past players and new players coming to play and stay.
I do always feel like I never get csr unless I use a booster but I will still say i’m enjoying the multiplayer even though the matchmaking takes a while.
  1. 1
  2. ...
  3. 35
  4. 36
  5. 37
  6. 38
  7. ...
  8. 39