Picking the top ‘n’ players from the pool is always going to be inaccurate.
Especially when ‘n’ is relatively small.
It didn’t work… and I’m not surprised they dropped it like a hot potato.
Yet it has been shown to be faster and more accurate than other systems.
So they were still messes. And even more broken.
No. Everything should be 1/1
In the first 3 years of Halo 5 when reaching champion you were not facing low diamonds. High diamonds would slip in but everyone you faced was onyx. This worked. Only until the population started going down that it didn’t work as well
We don’t know what data and how weighted.
It’s important to separate the influence of form vs the assumption of rank when you start a new playlist. The latter is very influential but of pretty much no significance.
A lot of people are getting the two issues mixed up.
That’s just not true. Quitter’s lose heaps of rank.
And yes, it’s frustrating that the system sometimes uses such losses to bring your CSR and MMR back together. It would be better if they waited for a loss in a valid game.
Again. We don’t know the weightings for sure.
There is definitely a weighting for kills per minute - but the behaviour of that is to rank you up faster, not necessarily higher.
As you close in on your rank the win becomes increasingly important.
They do need to something about this. A limit on ranks you can squad up with. Or a game limit on new accounts before they can play ranked.
Not the fault of the ranking system.
There were some. But they seem to have stopped?
There is undoubtedly things they can do to improve. Particularly around the CSR journey and what they show to the player.
Bro the elo system is dumb as hell, if they match of pure rank players would naturally be in line with their rank. Elo is just a way to make the grind easier during solo q. And this is coming from a player eho usually solo qs in ranked. the current elo system is -Yoink!-.
1 Like
https://www.reddit.com/r/halo/comments/sjcg8u/its_official_matchmaking_is_broken_social_games/?utm_source=share&utm_medium=ios_app&utm_name=iossmf
It’s been pretty clear there is significant weight.
It is true, I’ve seen it pan out this way myself.
Winning is important, but there is a high level of disparity in the ELO distribution with the current system.
One of the few ranking system that will allow you to queue right off of the bat with a new account. Some cheaters, not all, generally use a new burner account to play with.
Hopefully. Lol
The matchmaking in ranked is messed up. It’s impossible for anyone outside of 343 to know why it’s messed up. Since we can’t know why it’s messed up, 343 isn’t going to care about our suggestions for fixes.
That being said… It has been shown that performance in bot bootcamp and social playlists is affecting matchmaking in ranked… And that should be changed.
Is it though?
-
They smurfed a bad social MMR.
-
When you start a new playlist your closest existing MMR is used as a starting point. Normally this is a fair assumption that helps find your rank faster. But not when you smurf a bad MMR in a social playlist and it is the only one the system has on hand to seed the new playlist.
-
So in placement they started with really low ranked opponents. But because the system is volatile early the quality of the opponents increased quickly (as they should). The ranked MMR would have been rising fast.
-
At the end of placement you are given a starting CSR that is lower than your MMR This gives the system a bit of time to breathe. But it also means that your opponents are going to be higher than your CSR (giving you a chance to earn the MMR).
-
In the last couple of ranked games they struggled / lost. But their CSR wouldn’t have fallen because their MMR was still higher.
-
They then went back to lose some social games.
-
On returning to ranked they played some lower than expected ranked opponents. This could have just been chance (it was only two games and their skill curves are still wide / volatile) - or it could have been that their MMR had fallen much closer to their CSR. Remember that they were beginning to struggle before they went back to social.
I’m not saying there isn’t an influence. We know there is - courtesy of the shared data / offset model that TrueSkill2 uses. But this example of two games is not particularly compelling - especially when their MMR was probably falling quickly back towards their CSR post placement.
As long as you keep in mind that we are talking about two different ways ranking is being affected.
-
Is the seeding MMR when you start a new playlist. If you have only played social/bots this will be your starting MMR in ranked. It’s a significant influence at the start - but placement and a starting CSR < MMR tend to fix the problem. And importantly this only happens the first time you go into ranked on a new account.
-
Is the shared MMR data. We know that all the playlists share some MMR data and then have offsets (eg. for the x-axis and width of the curve). So we know they can influence each other. The evidence we have for it is very anecdotal. We don’t know what factors or weightings are involved. But I imagine the influence is not as drastic as some are claiming.
It doesn’t matter, letting other playlists influence ranked is completely unnecessary.
If it really has very little influence except in your first few ranked games, then it shouldn’t matter if that is removed.
It will learn where you should be quickly just like with a new account that doesn’t play any other playlists. Isn’t that why there are 10 placement games required?
If it has a lot of ongoing influence, then that should definitely be removed.
Agree.
But if the influence is minor… and their are solid performance gains for the system (eg. speed or amount of data that needs to be saved), then you may have a compelling reason to keep it in.
I haven’t read the discussion paper for a while - but they did definitely touch upon the shared data and how it reduced the memory footprint of the system.
you’re spot on mate. im still not quite sure what the issue is with the current system. if anyone has a good option, they’re not sharing it & just expecting 343 to invent one lol
heheh now THAT is way too optimistic for the halo community 