On the subject of rank

So the blog post was about as expected, no new or surprising information to anyone with even a cursory understanding of how ranking systems work in shooters. Nevertheless there are two things that I think are worth addressing, those being the influence of socials on ranked mmr and the use of mmr rather than csr to put ranked matches together.

I get why you might want to use performance in socials to set a baseline for ranked placements, but doing so with no protections in place for the different nature of those games is a staggeringly bad idea. Naturally, it’s worth noting the case of the unfortunate player who played bot bootcamp and then got pushed straight into the onyx lobbies for his placements, got destroyed and still placed in d1. This is a great example of why there should be no mmr in the bot playlist, but also serves as a worst case scenario of letting socials have too much impact on the baseline.

People don’t necessarily play the same in socials as they do in ranked. They might play socials to sweat against players who dont care and pad their stats, or they might be the player who doesn’t care and is just there to do challenges or have something to do with their hands while paying attention to something else. there is zero guarantee that socials will provide an accurate baseline, so you need to be ready to immediately discard whatever impact socials have on ranked mmr the moment you see a difference in performance. Otherwise you make it blindingly easy for a top tier player to place into bronze and stomp their way to the top, or vice versa for enabling middling players to boost their rank.

And on that subject, though you didn’t acknowledge it in your post, it is very clear from playing the game and many people conducting tests, and abusing the system, that socials continue to impact your mmr long after you have finished your placements. When I was grinding fiesta killjoys for the weekly, I had to intentionally play like trash so that my enemies would spree (because good players don’t let you spree - side effect of too strong sbmm in socials). When I finished doing that and went back into ranked, the first few games were absurdly easy. Not only were the skill of the lobbies substantially lower than my typical games, the matches were skewed to give me better teammates than enemies to help me win. The only reason I can assume this happened is because my fiesta games lowered my global mmr to the point where the game just ignored my diamond csr and put me into lobbies full of plat/golds, and distributed the players to balance out the mmr, ignoring the very lopsided csr averages.

Socials simply should not do this. If I play socials to warm up, and intentionally play like an idiot because I just want to get into as many gunfights as possible before going into ranked, am I now boosting my rank gains by lowering my mmr? If I use bots to warm up, am I doing the inverse, punishing myself and my teammates by putting me in lobbies I can’t compete in? Even if the influence is not so strong off of 1-2 games, what if I play for a day with friends who are much better or worse than me? When I then go back into ranked, should that performance in socials (another case where strong sbmm in socials is a problem) be used to weight my ranked games?

Clearly not. Social and ranked mmr should be fully segregated, or at the very least should have no affect on each other once placements are completed for the first time. Subsequent placements can use previous ranked mmr as a baseline, which would be more accurate in any case.

The next issue I have is with the use of mmr to balance ranked games, rather than csr. In short, this system punishes players for good performance by increasing the difficulty of subsequent games separately from their progress through the ranks. It also decreases the accuracy of the ranks themselves.

I have no issue with mmr being used to adjust the rank gains and losses on the basis on performance relative to the estimated difficulty of the match, but it causes real problems when it becomes a tool to attempt to force perfect 50/50 games while ignoring the visible player rank. A player who performs very well in one game will immediately experience a series of difficult games, as the mmr apparently adjusts much faster up than down, but this is true whether or not they won the game where they performed well. So a player who works very hard trying to carry a losing team gets their mmr boosted, but they see no csr gain to acknowledge their hard work and skill.

Instead, they are rewarded with even harder games, which they are less likely to win, and which will cost them greater amounts of csr as the system judges them to be higher skill relative to the lobby than csr might suggest. This will then continue until they have enough games at subpar performance to lower their mmr to where they will be given games that arent weighted against them, and in which winning will now give them greater amounts of csr for beating the odds as determined by the invisible mmr system. All of this reads to players as a punishment for good performance and a reward for mediocrity, all for the sake of appeasing the system’s need to force 50/50 win/loss.

A better system would be to use csr to balance matchmaking, and have mmr exist solely to nudge csr as close to true skill as possible. Essentially, mmr should serve as a buffer to keep csr gain/loss from moving away from where a player “belongs” too rapidly once they have reached the level where their performance is the average. This then exclusively rewards players for improvement and good performance, and punishing them only for poor performance.

No one wants to play a game that actively prevents them from climbing by pushing them into games way outside their rank bracket. If you are better than your rank suggests, you should climb until this is no longer the case. If you are worse than your rank suggests, you should drop until this is no longer the case. End of story.

And as an aside, loosen up the sbmm on socials. The current parameters create a massive sweatfest that promotes complacency by never matching you against better players, while failing to reward players for becoming more skilled by increasing the difficulty of matches as you get better. This is great in a ranked experience where you want to play at your best against similarly skilled players, and you have a visible rank to display your competence and track your improvement. But it sucks when you want a casual experience.

Note that I’m not asking to stomp noobs in quickplay, I just think that a looser sbmm does a better job at giving everyone a decently rounded and relaxed experience where you will do somewhat better on average as you become better on average. It also disincentives smurfs from ruining low ranked games by giving everyone more of a chance to get the occasional great game where they feel like a god.

TLDR: Separate social and ranked mmr; one should never affect the other once the first placements are done, possibly not even then. And use csr to put together ranked matches, allowing the system to actually do its job as an indicator of skill, while mmr functions as a weight to pull csr closer to actual skill despite fluctuating win/loss.

2 Likes

You may be interested in this thread:
https://forums.halowaypoint.com/t/that-rank-system-blog-post-wasnt-very-good-or-helpful/508303

Some interesting discussion around the topic in there.

Hush. You are using way too much common sense for 343 to ever understand your feedback.

WTF are you talking about?! The SBMM in social is extremely loose, so much that the fast majority of matches are one-sided with huge skill gaps between and within teams. You are almost aways matched against (way) better players, unless you are in the top of the skill-list obviously. And by far the majority of matches have players with KDA’s of 10+, and reguraly even 20+. A strict SBMM would not have that.
It is extremely sweaty because it is so loose, since everything depends on those very skilled players, those determine the outcome, while the rest is either just cannon fodder (they have no fighting chance against the skilled players) or filling (the skilled players are ‘stealing’ almost all the kills). This creates also a lot of quitters because people just don’t have any fun with that.

I can promise you it isn’t.

The reason the balance is so bad is because the mmr moves very quickly, so if you do well you are quickly “rewarded” with higher skill/sweatier lobbies. And of course, to prevent players from easily tanking their mmr to abuse lower skill lobbies, the system is designed to be less sensitive to poor performance, preventing you from dropping back down if the mmr overcorrected upwards.

I disagree. A strictly balanced game would have some of that due to purely random fluctuations. Strict sbmm is not the same thing as strict balance. When the sbmm is operating on unstable parameters, such as the mmr system used in Infinite, the balance will get worse as the sbmm gets stricter due simply to the random variance in player performance from game to game. This only gets worse when nonrandom, i.e. behavioral variance, comes into play.

Consider, for instance, how players doing certain challenges perform compared to people just playing normally. What do you think happens to their mmr while their stats tank game after game trying to do stupid things like get kills with bad guns or let enemies spree in fiesta? And what happens when they complete the challenges and stop playing like a bot? They stomp their way back up to the sweats because no one even resembling their skill level is in those games to stop them.

There are a lot of challenges that are going to get you farmed even without trying to feed, because what it wants you to do is just not a smart way to play the game. And sometimes you are just tired or not paying attention, and you play without thinking about the game, which will also lower your mmr, and give the same result when you pick up the game and play normally.

In fact, it’s the exact opposite. The strict sbmm creates this problem.

For example, let’s say a mid plat player ends up in lobbies with players in the high silver-low gold range. If the plat is the only player who performs above a silver/gold level, they will stomp, ignoring their team and soloing the enemies because there is no one to stop them. The enemy team gets to experience being fodder and the teammates get to experience being the filler

But if there are a few other plats or mid-high golds mixed in with the lower skill players across both teams, the plat can’t have as much impact. Sure, they will probably still do well, but they generally won’t be able to completely wreck the enemy team, especially if they play solo and get teamshot. And the less skilled players in the lobby will still not do great, but at least they have teammates that can help them out, and aren’t relegated to just walking around aimlessly while one guy kills the entire enemy team.

Surely this would be a better experience for all involved? The less skilled players do a bit worse on average stats wise, but spend way less time getting stomped into the floor. And the more skilled players do a bit better on average, a natural reward for putting in the time and energy to get good, and get playlists that aren’t sweaty beyond belief.

Speaking of fun, consider the experience for people who want to group with their friends, who might have wildly different skill levels. You know, people who play the social playlists to have fun and socialize with their friends? Those games are never going to be balanced, no matter how well you precision engineer the mmr system.

With restrictive sbmm, such a group of friends will be given a very consistent experience. Consistently terrible. In every match, the less skilled teammates get to experience being complete fodder, while the higher skill players are given the choice to either sweat like mad or accept that every single match is going to a short, one-sided stompfest. Because nothing says have fun with your friends quite like being unable to play together without making a smurf account and feeling like a complete tool every time you get a kill because no one in the lobby is even close to a match for you.

If sbmm were relaxed, mixed skill teams would mostly be up against general population, which is to say gold-plat. So the worse teammates might not do great on average, but they could at least play the game with their friends without feeling like a complete anchor playing respawn simulator every game. And the higher skilled players get a relaxed environment to share their hobby with their friend, maybe give them some tips and help them improve, without needing to sweat their balls off for a snowball’s chance in hell of having a moderately close game.

Can we finally start penalizing players for quitting matches? I’m tired of losing ranks because someone quits 5 seconds in and i have to play 3v4 or even 2v4. Either penalize appropriately for it or don’t penalize at all, but stop penalizing players for not quitting the game when they have trash teammates

If they quit 5 seconds in, it is likely the game just refused to load them in
(players characters are already loaded in before the actual player is loaded, this is also frustrating in JIP matches in social, since you often start with a death).

The game just still have to many bugs, lag issues and crashes.

And there is already a ‘quit’ penalty in place, in ranked you get it after your 2nd ‘quit’ (it used to be after the 1st, but now its the 2nd because of the many bugs) and that gets worse every time you get a new ban. So they are already harsh on that, even with all those crashes.