SBMM isn't used properly in games anymore

It used to be SBMM would be used to find players within a certain radius of your skill level, and then it would attempt to balance the teams out for a match that felt as fair as possible, as often as possible.

Now in Halo Infinite, and from what Bungie said in their recent TWAB the next SBMM update for Destiny 2 will do, is that the game does it’s very best to keep your win/loss rate as close to 50% as possible, or at least within the range of 40%-60%.

The result is that instead of playing consistent matches that feel fairly balanced, with the occasional steamroll either for or against you. The algorithm now tries to decide ahead of time if you, and everyone else who gets matchmade should be winning, or losing matches, and purposely stacks teams so you’ll either roll or get rolled, with balanced matches being few and far between.

I don’t know who decided this kind of system was a good idea, but I hate to see that it’s bleeding into more and more franchises. Your wins don’t matter anymore, and your losses don’t matter either. Matchmaking decides whether you will win or lose before the match even starts.

9 Likes

Isn’t this what it does now?

Within the limitations of available population that is.

Isn’t that the ultimate aim of any ranking / match-making service?

The downside of having people with higher win rates is that other’s have to have a correspondingly low one.

I just read through the TWAB. Can you point to me which bit describes this?

SBMM doesn’t really need to manipulate the matches specifically - it should all balance out naturally. If someone starts winning more than 50% all you have to do is “rank” them up and let the tough opposition do it’s work.

Your wins and losses decide if you rank up, stay the same, or rank down.

They always matter.

100% this is the issue. It’s all so artificial and illegitimate feeling. It’s like they designed the game to not be fun on purpose

5 Likes

Sbmm’s sole purpose is to hold the hands of non video game players/guys that really love to see multiplayer games burn and fall. There literally isn’t a single argument to defend it.

The bad kids ruined Halo.

They’re not kids, just old men

I can see how challenges would be “fun to run”.

When I read topics like this, it just shows how poorly games are being designed these days that It’s part of the reason I stopped playing Infinite. The game is designed with barely any incentive, to hold players back sorta speak. For what purpose?

I don’t get the direction gaming in general is heading other than business like 343 thinking only about their wallet. Correct me if I’m wrong, but wasn’t there a point back in the day that developers designed games to have an enjoyable experience, because this is the complete opposite.

1 Like

I am a mid level Platinum player p4 and no better. I have been playing against diamond players.
Although it’s only 2.5 levels higher it’s not 2.5 wins better the skill gap between the two ranks is massive and also playing against onyx who party up with a mid or low gold player to get dragged into lower lobbies so they can get loads of kills.

I am trying to find the fun/balance in MM
Is it because hardly anyone but us idiots are still playing ?

1 Like

But why should we be ‘ranking’ social matches in the first place? Why call it social if it’s actually following a ranking system? If it were truly social, winning/losing wouldnt matter and we could all just be randomly thrown into matches. Honestly its good to get unfair matches every now and then because you can see what the higher level players are doing and try to learn from it

You are using skills to create two even teams.

Ranked is matched around the player. Social around the team.

Most players just want to have a close match. Winning or losing isn’t the point. It’s about feeling you had a chance.

The evidence for players logging off for the night in H5 was pretty conclusive. Players will put up with losing streaks, poor performance, and even lag, as long as they felt the game was up for grabs.

This is great for the good players. They just spend game after game feasting on the dregs that random serves up to them.

An Onyx player in random set ups will be against four players ranked lower than them in 8/10 games. At an average level of G6/P1.

They only face two players ranked higher than them in 1.4% of games. So much for the “some good / some bad” line of thinking. It’s definitely not a case of “now and then” for everyone.

And if two Onyx level players buddy up… carnage!

Except that it’s not every now and then for players on either side of the bell curve.

And yes, I agree it’s good to chance the arm against better players. But I doubt a Silver player would learn much up against a couple of Onyx level players. We have a wide range of skills in our custom’s group - and if we play straight out slayer the Silver/Gold’s die. A LOT. And they often have no idea how it happened. And they certainly don’t get any chance to improve by “fighting back”. So I’m not sure how that helps them become better players.

I thought the guiding principles of any learning curve were that gradual exposure to higher skill sets was the best improvement path.

Because the goal of industry SBMM has changed:

The goal now isn’t to let the player show their skill and climb to their heights, that SBMM goal is so 2010.

The goal now is to keep forcing competitive matches every game and stagnantate the player progress to keep the player playing with an illusion that they are grinding to get better.

This is the main reason why I’ve stopped playing Ranked modes in almost every modern game.

1 Like

That’s not what it does now, not at all. Where previously, in Halo 3 for example, the SBMM system took your visible rank, and specifically matched you with players within 5 ranks of it. In a ranked playlists you would never see a General (50) if you weren’t at least a Brigadier (45), and you would never see a Brigadier if you weren’t at least a Colonel (40). And the algorithm purposely tried to match you with people as close to your level as possible. So as a level 30 player for instance, it would first look for level 30 players, then extend to 29-31, then 28-32 and so on.

The current matchmaking system instead looks at your win/loss rate, and if it is above 60%, it tries to match you into games you are supposed to lose, comparing your skill and the skill of your teammates with those of your enemies. And if you’re below 40%, it tries to do the opposite. Instead of trying to find similarly skilled players to play with, so that each match feels winnable for as many players in the match as possible. It does the opposite, it builds a team of players who’s “turn” it is to lose, then finds a team of higher skilled players who’s “turn it is to win” and makes them play against eachother.

Sure, this new version of SBMM helps to ensure everyone wins about half of their games. But instead of it being because they are playing at their skill level, where going on a win streak is a sign of improvement, as you overcome other players that used to give you trouble, it’s because they are matched with weaker players roughly half the time, and stronger players the other half of the time. If you go on a win streak in Infinite, it’s not because you got any better, it’s because the game put you in matches you were supposed to win several times in a row.

It’s a really awful system. Especially since it creates a poor environment for improvement in the first place. Spending half your games playing against significantly weaker opponents reinforces bad gameplay habits, and spending the other half getting steamrolled prevents learning opportunities. Even matches, where you know winning consistently is an accomplishment, and where small improvements will make a noticeable impact on a match is where players learn, and grow the best.

Also, if you didn’t already know this get ready. But your win/loss rate is purposely fixed between 40 and 60% regardless of skill level because of the system, and is NOT taken into account when determining your actual player rank. Instead it looks at things like how often you kill or die to other players, and compares that to their rank.

The ranked playlist’s system of ranking is actually a sham. It will always push you towards your global “rank”. When it determines how much exp you gain or lose when you win or lose a game, it scales it based on how close you are to that rank. Which is why you rubber and a lot. When you are below the rank the game already considers you, you get almost no penalty for losing, and a large gain for winning. And when you’re above it you’ll get the opposite effect.

I, for instance, gain very small amounts of XP if I win at Platinum V. But will lose a significant amount of XP if I lose, so no matter how hard I push if I get matchmade into a win streak, that first loss will sent me right back to Play V.

Lucid on the other hand (just taking an extreme example), would still be gaining tons of XP per match at Plat V, but would lose almost nothing if he lost a match. Because the game would be pushing him to Onyx. Infinite would still match me with Lucid if it was my turn to lose a match and his turn to win. And neither of us would get anything out of the match, it would just be an easy game for his team, and an impossible game for mine.

I’ve never been too exceptional at Halo, at all. But I always felt like most of my matches were fair, and like I could contribute to a win if I just tried a little harder, played a little better, or learned a little more about the game. Modern SBMM has completely removed that part of the game. Now I walk in expecting about half of my games to feel unwinnable, and the other half to be pub stomps.

Also, to address the TWAB, it’s specifically the line where they said they intend for the new SBMM system to keep players between a 40-60% win rate. That directly implies the system is going to start pushing you into matches you’re supposed to win or lose to bring that average up or down. Which is how modern games, including Infinite, have started doing it. It’s not a good system, not in the slightest.

1 Like

Which is what I thought we were still doing, within the limits of the population, adjusting for squads, and stretching things just a little bit to prevent undue rank locking.

I honestly haven’t seen anything about this.

I’m on a lunch break at work and just frantically googled Halo / TrueMatch - and couldn’t find any reference to W/L ratios. Just the usual suspects of matching on skill and latency.

TrueSkill2 does have weighting for form. Is this what you are referring to? If you have won a few matches - especially at your expected skill rank - the system should give you tougher opponents to play against. A chance to rank up.

This is fundamentally different to creating specific win teams and losing teams. This would involve big imbalances in average MMRs between the teams.

When I plotted over 100 games of team MMR I couldn’t find any significant changes in opponent MMR.

It would be lovely if you had a reference that I could read about TrueMatch using W/L as a metric.

I would have thought that W/L ratio in ranking is kind of redundant.

If the system is working your W/L should be 50%.

It looks at the result, who you beat, and your personal performance (KPM and DPM). There could be other metrics but none that have been mentioned (either by 343 or Microsoft).

This is only true if you play as expected.

If you perform (consistently) above your station you will rank up.

It’s up to you to actually improve.

You’ll get small amounts if the win was token (ie. vs the same or lower rank).

If you haven’t “earned” that CSR your MMR will be lower so it is “paid back” on your next loss.

If you want to keep it and progress to Platinum 6 you need to start winning against Platinum 6 opponents (establish at 50% win rate at that level) and/or start getting 1.5 KPM against those opponents.

And at this stage of the season it’s a bit more compounded. We’re all going to be a bit rank locked (narrow, stable MMRs).

I’ve read through the TWAB and that’s not the mind set I got.

It’s more that the system creates a 40-60% win rate. That’s what you want the system to do. Ideally it would be a smaller range - but that tends to be unrealistic (you can’t predict how random players function as a team) and slow.