Skip to main content

Forums / Games / Halo Infinite

SBMM - Why it is both good and bad

OP HaloBungie343I

  1. 1
  2. ...
  3. 2
  4. 3
  5. ...
  6. 6
SBMM, or Skill Based Matchmaking, is an algorithm used by developers to put you with players that are similar is skill to you. Games as far back as Halo 2, CoD Ghosts, and other similar games (primarily in the FPS genre) al have had some form of SBMM.

SBMM is not necessarily a bad thing. But the implementation of it in modern games has been very bad.

The point of this explanation is, while Halo has had SBMM in every Halo game to date with the exception of Halo: Combat Evolved, Halo: Combat Evolved PC, and Halo: Custom Edition, it has not been nearly as big as today nor has it affected matchmaking as it does today. This is how we have the three options available to us when matchmaking in Halo 5: Guardians and Halo: MCC. We have "Search for the fastest match," Find a Balanced Match," and "Extended Search." These alter the SBMM in a way to give us a match we want. One that gets us in quickly with any players of any skill close by. One gives us a match close but with similar skill. And one gives us a match with very similar skill anywhere the game can find one. Even if it isn't at the closest servers. But should remain in your region at the least.

Here is my thoughts on Halo: Infinite and SBMM:
Halo: Infinite needs to continue to use the same SBMM algorithms it has been using since Halo 2 and do NOT change to the modern system used by companies like Ubisoft, Activision, EA, and Epic Games.

I will not argue and ask 343 to not use SBMM as that would be pointless. 343 will use SBMM whether anyone here likes it or not. I just hope it will be the same as all previous Halos.

Anyone else have something to add? Do you disagree?
SBMM in ranked, no SBMM in social. Done (obv not that simple but its what they should start with and then good the strength of the SBMM from there)
Loose sbmm parameters in social hurts player retention. Just my experience, but whenever I get easy games I feel like I'm wasting time and whenever I get games where I have 99% chance of losing I end up just getting frustrated. Both scenarios usually lead to quitting the game and doing something else more worthwhile.
qlimm wrote:
Loose sbmm parameters in social hurts player retention. Just my experience, but whenever I get easy games I feel like I'm wasting time and whenever I get games where I have 99% chance of losing I end up just getting frustrated. Both scenarios usually lead to quitting the game and doing something else more worthwhile.
According to Josh Menke SBMM increases player retention. I think the big wave of games switching over to SBMM is due to that notion.
qlimm wrote:
Loose sbmm parameters in social hurts player retention. Just my experience, but whenever I get easy games I feel like I'm wasting time and whenever I get games where I have 99% chance of losing I end up just getting frustrated. Both scenarios usually lead to quitting the game and doing something else more worthwhile.
According to Josh Menke SBMM increases player retention. I think the big wave of games switching over to SBMM is due to that notion.
Ya, then we look at H5 and Cold War where no game is ever pure fun cause it's all a sweat fest of people with the same skill as you. Remove it or reduce it in social. If we want sweat, we will play ranked
qlimm wrote:
Loose sbmm parameters in social hurts player retention. Just my experience, but whenever I get easy games I feel like I'm wasting time and whenever I get games where I have 99% chance of losing I end up just getting frustrated. Both scenarios usually lead to quitting the game and doing something else more worthwhile.
According to Josh Menke SBMM increases player retention. I think the big wave of games switching over to SBMM is due to that notion.
Yes, that is what I was saying. I would define "loose" SBMM as win/loss probabilities leave the 35-65% range. The data doesn't lie, and I think that is why I've played H5 far more than MCC.

Case in point: I was just playing the new community forge playlist in MCC and I ran into an extremely unbalanced matchup where the players were farming my team in an objective gametype. I left and got banned. This is why SBMM is needed.
What's the difference between Halo's SBMM and the other devs you mentioned? I thought H5 even had a different system than previous Halo's.
Check out HiddenXperia’s YouTube breakdown of it. He shows some data and gives reasons for and against it as well as breaking down the pro’s and con’s.

matching primarily on SBMM has its trade-offs for sure but above all the game experience is what really matters. It can’t match you with someone with similar skill while sacrificing ping refresh.

I think 343i also needs to get rid of the horrible ranking system used in H5. From my understanding in arena, it bases your rank on games won and not on your specific skill level.

Arena should be based on KDA and rank accordingly by SBMM while social shouldn’t reflect any KDA directly...it is social and not meant to be a sweaty completion because if your KDA falls in that, it may effect your ranking in Arena.

Above all don’t sacrifice internet connections where if you are matched based on skill with someone halfway around the world, it could potentially create lag and other related problems in return making players jump around on your screen randomly which in return would lower your chance of landing shots and thus lower your KDA. Just check out the video LOL
man please dont give 343 any ideas, i swear bocw is so annoying with it's sbmm, i think sbmm is only good for beginners, so they don't get bodied by good players on their first match, at least till level 10, after level 10 no sbmm
it bases your rank on games won and not on your specific skill level.
"Skill level" is a vague subjective way to talk about skill that contains an implicit value judgement about what you think matters most. A skill rating tries to reduce a multi-dimensional entity into a single number that somehow does... something? Rather than letting personal values guide you directly, you should ask what that "something" is. It should be a quantifiable result and not some vague subjective goal.

In case of TrueSkill that Halo uses, that "something" is very simple: predicting the outcome of a match (the final ordering of teams), given a set of players with some skill ratings. This is a good objective goal, because regardless of what any single person thinks is the most important part of skill, everyone can agree that the team that beat the other team is the better team. As soon as you start considering things like individual scores within a team, you run into more subjective value judgements. The more specific you get, the easier it becomes to disagree with the basic assumptions of your skill rating system on a subjective basis. We can all agree that the team that won was the better team, but we won't all agree that the player who got X kills and Y deaths did better than their teammate who got Z kills and W deaths, for some values of X, Y, Z, and W.

It's easy to be all idealistic about skill rating systems. It's extremely difficult to come up with one that does something remotely useful. Even if you do, chances are that whatever you come up with, people are going to hate it as much as the other guy's system on similar subjective and idealistic grounds.
tsassi wrote:
it bases your rank on games won and not on your specific skill level.
"Skill level" is a vague subjective way to talk about skill that contains an implicit value judgement about what you think matters most. A skill rating tries to reduce a multi-dimensional entity into a single number that somehow does... something? Rather than letting personal values guide you directly, you should ask what that "something" is. It should be a quantifiable result and not some vague subjective goal.

In case of TrueSkill that Halo uses, that "something" is very simple: predicting the outcome of a match (the final ordering of teams), given a set of players with some skill ratings. This is a good objective goal, because regardless of what any single person thinks is the most important part of skill, everyone can agree that the team that beat the other team is the better team. As soon as you start considering things like individual scores within a team, you run into more subjective value judgements. The more specific you get, the easier it becomes to disagree with the basic assumptions of your skill rating system on a subjective basis. We can all agree that the team that won was the better team, but we won't all agree that the player who got X kills and Y deaths did better than their teammate who got Z kills and W deaths, for some values of X, Y, Z, and W.

It's easy to be all idealistic about skill rating systems. It's extremely difficult to come up with one that does something remotely useful. Even if you do, chances are that whatever you come up with, people are going to hate it as much as the other guy's system on similar subjective and idealistic grounds.
The easiest contradiction to that is seeing some of the ranks this system has turned out. Two onyx, one champion and a silver against 3 platinums and a diamond. The "average" of the team may be equal (according to Win/Loss ranking) but the sum of their actual skills is most certainly not. Furthering the point of seeing matches where 1 players goes 20-5 carrying their entire team to victory (often times being a onyx matched with a bunch of silvers or golds against diamond/platinum range team) which is in fact contradictory to your statement that "the better team won" because if we give each rank a number, add them up and divide to find the average of the team, then technically in my last example the "better team" actually lost but rewarded less skilled players for having someone carry them. Whereas ranking each individual on their individual stats/performance.

Where if a true KDA system was used (because the goal of most of the arenas is for the team to get to 'X' amount of kills) thus Kills-Deaths-Assists would be the ranking of what a players skill level would/should be (not whether they lucked out and won a bunch of matches), you would have players with a very similar skill level which when in a Slayer styled game, everyone would agree a true KDA is in fact the judgemental factor of that playlist. You can absolutely make the statement that a player with a positive KDA did better than someone with a negative KDA. Lets not forget that KDA factors in assists and not just whether you went positive in the game. The use of the system is subjective because to be honest; the subject is math and numbers don't lie.

Granted there's factors like having an off-game that can play into it but having an off-game is on that specific player whereas getting paired with garbage teammates who have a sub 1 kda can cause you to lose consecutive matches and thus drop your rank or even paired with someone on their off-game for that matter; even if the particular player had a positive KDA all of their losing matches. Most refer to Halo 2 or 3's (not sure which) numeric 1-50 ranking system as the 'preferred' ranking as opposed to H5's system.

All the above is referring to Slayer styled modes.

The final statement has never wrung truer however the pursuit of something better is always what is preferred. I think H5 naming their system 'True Skill' is also a bit deceiving given how the system operates as well.
Where if a true KDA system was used (because the goal of most of the arenas is for the team to get to 'X' amount of kills) thus Kills-Deaths-Assists would be the ranking of what a players skill level would/should be (not whether they lucked out and won a bunch of matches), you would have players with a very similar skill level which when in a Slayer styled game, everyone would agree a true KDA is in fact the judgemental factor of that playlist. You can absolutely make the statement that a player with a positive KDA did better than someone with a negative KDA. Lets not forget that KDA factors in assists and not just whether you went positive in the game. The use of the system is subjective because to be honest; the subject is math and numbers don't lie.
There are an infinite amount of different ways to combine three numbers (K, D, and A) and get a single number out of them. The specific way you choose again contains an implicit value judgment, and you get different orderings of (K,D,A) triples depending on which way you choose. You're putting more of your personal values into the system the more specific you try to be. You're not doing anything inherently more productive.
tsassi wrote:
Where if a true KDA system was used (because the goal of most of the arenas is for the team to get to 'X' amount of kills) thus Kills-Deaths-Assists would be the ranking of what a players skill level would/should be (not whether they lucked out and won a bunch of matches), you would have players with a very similar skill level which when in a Slayer styled game, everyone would agree a true KDA is in fact the judgemental factor of that playlist. You can absolutely make the statement that a player with a positive KDA did better than someone with a negative KDA. Lets not forget that KDA factors in assists and not just whether you went positive in the game. The use of the system is subjective because to be honest; the subject is math and numbers don't lie.
There are an infinite amount of different ways to combine three numbers (K, D, and A) and get a single number out of them. The specific way you choose again contains an implicit value judgment, and you get different orderings of (K,D,A) triples depending on which way you choose. You're putting more of your personal values into the system the more specific you try to be. You're not doing anything inherently more productive.
By getting more specific with the values then inherently you'll get more specific matches in skill and isn't that the idea to begin with?Having evenly skilled players matched together and then giving the 'W' to the better team? Halo is still a 'Team Shooter' is it not? The values can be given through common sense use of what is good and bad in this specific case. I think everyone will agree with the below:

Kills>Assists>Deaths would be the primary layout since the idea of winning is to get the most kills...assists are a better additive to the equation whereas dying is the exact opposite of what is necessary to get the win. Inherently then distributing accordingly someone with a higher KDA (or in this case KAD) are better skilled players when the skill of the game is Slaying.
tsassi wrote:
Where if a true KDA system was used (because the goal of most of the arenas is for the team to get to 'X' amount of kills) thus Kills-Deaths-Assists would be the ranking of what a players skill level would/should be (not whether they lucked out and won a bunch of matches), you would have players with a very similar skill level which when in a Slayer styled game, everyone would agree a true KDA is in fact the judgemental factor of that playlist. You can absolutely make the statement that a player with a positive KDA did better than someone with a negative KDA. Lets not forget that KDA factors in assists and not just whether you went positive in the game. The use of the system is subjective because to be honest; the subject is math and numbers don't lie.
There are an infinite amount of different ways to combine three numbers (K, D, and A) and get a single number out of them. The specific way you choose again contains an implicit value judgment, and you get different orderings of (K,D,A) triples depending on which way you choose. You're putting more of your personal values into the system the more specific you try to be. You're not doing anything inherently more productive.
By getting more specific with the values then inherently you'll get more specific matches in skill and isn't that the idea to begin with? Having evenly skilled players matched together and then giving the 'W' to the better team? Halo is still a 'Team Shooter' is it not?
What do you mean by "more specific matches in skill"? What is your measure of success? What is the quantity that tells you how well your skill rating system is performing?
tsassi wrote:
tsassi wrote:
Where if a true KDA system was used (because the goal of most of the arenas is for the team to get to 'X' amount of kills) thus Kills-Deaths-Assists would be the ranking of what a players skill level would/should be (not whether they lucked out and won a bunch of matches), you would have players with a very similar skill level which when in a Slayer styled game, everyone would agree a true KDA is in fact the judgemental factor of that playlist. You can absolutely make the statement that a player with a positive KDA did better than someone with a negative KDA. Lets not forget that KDA factors in assists and not just whether you went positive in the game. The use of the system is subjective because to be honest; the subject is math and numbers don't lie.
There are an infinite amount of different ways to combine three numbers (K, D, and A) and get a single number out of them. The specific way you choose again contains an implicit value judgment, and you get different orderings of (K,D,A) triples depending on which way you choose. You're putting more of your personal values into the system the more specific you try to be. You're not doing anything inherently more productive.
By getting more specific with the values then inherently you'll get more specific matches in skill and isn't that the idea to begin with? Having evenly skilled players matched together and then giving the 'W' to the better team? Halo is still a 'Team Shooter' is it not?
What do you mean by "more specific matches in skill"? What is your measure of success? What is the quantity that tells you how well your skill rating system is performing?
Similar players matched by KDA will result in 'groups' of teams of similar skill being determined. Thus matching those players against other players with similar KDA typically would make an even playing field and thus the group of 4 (in this case Slayer) that works as the better team would win the match. The losing team would only drop rank if they performed less than what they have been averaging while those with a KDA, regardless on W/L, who stayed the same or went higher would increase their ranks since according to their average prior to the game, was better despite winning or losing or teammates performance.

The easiest quantity of data that is going to tell you your skilled ranking is working is going to obviously be player feedback. Depending on the feedback would be your measure of success. Especially at a time when feedback can received in seconds compared to weeks/months/years.
The easiest quantity of data that is going to tell you your skilled ranking is working is going to obviously be player feedback. Depending on the feedback would be your measure of success. Especially at a time when feedback can received in seconds compared to weeks/months/years.
Ah, I see. You're looking for a player rating system that optimizes for player feedback. In that case, all I can say is I admire your optimism.
Also to be fair, I think H5's current ranking system or pairing system could be fair in the Social Playlist. Putting different levels of skill players together where the additive sum is close to equal to the opposing team because the playlist is Social, designed for fun and a 'less sweaty' type of gaming. That system is accurate to a degree (like having an operation where you know the general area the incision needs to be made) whereas the system im describing is accurate to the point of surgical precision (where you need to make sure not to nick an artery) if that makes any sense...its more precise. I believe the system I'm proposing to be a better fit for Arena.
tsassi wrote:
The easiest quantity of data that is going to tell you your skilled ranking is working is going to obviously be player feedback. Depending on the feedback would be your measure of success. Especially at a time when feedback can received in seconds compared to weeks/months/years.
Ah, I see. You're looking for a player rating system that optimizes for player feedback. In that case, all I can say is I admire your optimism.
I think you're being a bit pessimistic in your response. it is no different than any of the changes that 343i themselves have relied upon especially given that they are "returning to the Halo roots" realizing that some of the changes they've made were not highly doted upon. The fact is the current system is good to a degree but Arena players are more responsive to a system I mentioned revolving around the KDA. Furthermore the notion of returning the KDA for Arena isn't something new...players have been asking for it (specifically 1-50) for more than just a few weeks/months/years. I mean even in the latest update there is acknowledgment of a specific team that is monitoring player feedback (I believe they were referred to as the Live Team).

"hey i was positive last game so im not the reason we lost". The logic is sound tight and makes sense in a mode where having more kills determines the winner (AKA a positive KDA). Where as the current H5 system gives the appearance of "I went -7 but I had a guy (or two) on my team that went +13 and we won so I rank up". That is totally backwards and a blind man can see that.
qlimm wrote:
Loose sbmm parameters in social hurts player retention. Just my experience, but whenever I get easy games I feel like I'm wasting time and whenever I get games where I have 99% chance of losing I end up just getting frustrated. Both scenarios usually lead to quitting the game and doing something else more worthwhile.
According to Josh Menke SBMM increases player retention. I think the big wave of games switching over to SBMM is due to that notion.
Ya, then we look at H5 and Cold War where no game is ever pure fun cause it's all a sweat fest of people with the same skill as you. Remove it or reduce it in social. If we want sweat, we will play ranked
This just doesn't make sense to me?

Surely a close match vs people of similar skill is where the fun is? Close matches for the 'win'.

I don't see how my opponent being 'sweaty' makes my enjoyment of the game any different?

In fact, I think there is something ironic in losing and calling your opponents 'sweats'. You are basically admitting that you care enough about the result to insult the opposition and/or make excuses because you lost. Therefore you are, by definition, a sweat as well.

There is nothing fun about a mis-match in skill - either for the team being stomped or the team doing said stomping. It just ends up being a complete waste of time.
Darwi wrote:
qlimm wrote:
Loose sbmm parameters in social hurts player retention. Just my experience, but whenever I get easy games I feel like I'm wasting time and whenever I get games where I have 99% chance of losing I end up just getting frustrated. Both scenarios usually lead to quitting the game and doing something else more worthwhile.
According to Josh Menke SBMM increases player retention. I think the big wave of games switching over to SBMM is due to that notion.
Ya, then we look at H5 and Cold War where no game is ever pure fun cause it's all a sweat fest of people with the same skill as you. Remove it or reduce it in social. If we want sweat, we will play ranked
This just doesn't make sense to me?

Surely a close match vs people of similar skill is where the fun is? Close matches for the 'win'.I don't see how my opponent being 'sweaty' makes my enjoyment of the game any different?

In fact, I think there is something ironic in losing and calling your opponents 'sweats'. You are basically admitting that you care enough about the result to insult the opposition and/or make excuses because you lost. Therefore you are, by definition, a sweat as well.

There is nothing fun about a mis-match in skill - either for the team being stomped or the team doing said stomping. It just ends up being a complete waste of time.
Dilly Dilly!
  1. 1
  2. ...
  3. 2
  4. 3
  5. ...
  6. 6