Ranked is completely flawed anyway with all the gold accoubts flooding the Diamond matches. Losing games becausr if some dude going -12 and dropping massive amount of CSR is something 343 has to finally adress. But as we all know… absolutely nothing will change.
2 Likes
I don’t know that it is, I haven’t seen anything to suggest that it is. Maybe it used to be, but a lot has changed since 2011.
But it doesn’t matter, because the rank number isn’t used to put matches together. So no, it doesn’t change my perception.
They use words and phrases. Wow.
When they put a match together, they use the mean value of that curve, which is a number. Teams are matched according to the average of those values for the players involved.
It doesn’t matter that they use a curve to get the number, it doesn’t matter how they change the number over time, it’s just them using statistics to increase the accuracy of the mmr NUMBER.
I’m done trying to explain this to you.
Human psychology has not fundamentally changed over the last decade. It doesn’t matter that the system can find exactly where you belong, humans want to see progress and feel rewarded for their efforts.
But don’t take my word for it:
https://twitter.com/MaxHoberman/status/1502633023497555974?s=20&t=O2gyeja8NYtVyedpbOvxUA
Bottom line, attempting to balance too strongly and/or too quickly ruins the experience.
Game must be pulling my mmr from last season if im diamond 6 matching 1600-1700s. Still, i should be matching other diamond 6’s at the rank im at.
1 Like
When you go into placement each season it keeps your MMR as a seeding point. We think it’s just the width of the curve and it’s volatility that get’s refreshed -which gives the system a chance to re-rank you if needed.
And at the end of placement your CSR starts below your MMR to a maximum of Diamond 6 (it used to be Diamond 2).
Your MMR is obviously higher as per your matches - and it will only take a few wins for them to catch up.
You asked for citations to back “my opinion”. I gave them to clearly show that it’s mu minus 3 sigma and not the mean. And your response is that’s not good enough for “your opinion”.
Sigh.
And the funny thing is that you are complaining that you are being given team-mates with trash ranks - but most of that could easily be explained by their CSR being well below the mean of the MMR. Like three times sigma below.
You were the one who used their wording on that page to support your argument. Despite the fact that literally the first bit of the article to show off a huge picture of a player’s rank depicted as a huge glorius curve.
When they put the match together they use the whole width of the curve. That’s why placement matches are so volatile (and they hide your rank). Your sigma value is important.
And in match-making it would be very easy to divide the teams so that the average curve on both sides is the same. And at the end you can take both these averaged curves, subtract them, and use the result curve to calculate the odds between the two teams.
The maths paper by Moser talks about these sorts of calculations.
It does when the shape of the curve changes the behaviour of your rank.
A low flat curve and a tall narrow curve can both have the same mean. But the former curve will be volatile (in terms of match-making and it’s ability to move along the x-axis) while the latter curve will resist change.
I agree that a lot of what happens can be summarised by the mean. But you can’t predict what is going to happen without the sigma.
Agree. I just don’t think a skill rank is the place to feed this part of your psyche. The harsh reality is that skill is incredibly difficult to improve, whether we are talking about Halo, chess, or crotchet.
You are going to hit your ceiling. And because we are coming into this as FPS “veterans” it’s not going to take anywhere near as long as if we were taking up an entirely foreign hobby.
We desperately need a skill based XP rank. Weighted to wins, medals, and score.
Yep, and one of those GDC talks touches upon exactly the same experience. They know that they need to feed the right range of matches to allow your rank a chance to breathe. Especially early (placement needs to be volatile - even to the point you have to hide the rank) and late (don’t let sigma lock your rank entirely ala H3).
Your sole direct citation that was more recent than 2011 didn’t show that. Vague references to various undated unspecified talks do not constitute citation. Regardless, it doesn’t impact my understanding of the matchmaking, because that’s not the value they use in matchmaking.
I’m sorry you can’t understand that.
There’s a difference between reading for comprehension and noting that a few words and phrases are used.
Citation needed.
The one from 2011, you mean. Funny, they have a new system and they put out a paper on that in 2018. See my earlier citation, it doesn’t support your claims.
Are you mentally handicapped? Serious question.
The reality is that people’s skill at various things changes all the time as they fall in and out of practice. Improving is mostly a matter of practice and study, in other words time and effort.
Just because someone has played a lot of shooters doesn’t mean they are equally good at any of them. You could as easily say that because someone is very good at basketball, as a veteran player of ball games, they are therefore very good at volleyball or football. The games share similarities, but they are nonidentical
And it’s equally asinine to assert that there’s no room for improvement just because someone is already skilled at something.
Ranked is THE place for people who want to put in the time and effort to improve, and therefore it should be designed to make that experience satisfying to players. I’m sorry you don’t seem to get that.
But I’m done talking to you.
A citation with an older date is better than no citation at all.
I’m not sure what you mean by vague references though. They were GDC talks on ranking. One of them was given by the guy who implemented TrueSkill and TrueSkill2 into Halo 5.
Microsofts own webpage for TrueSkill clearly states it as mu - 3x sigma.
Surely the onus is back on you to prove it has changed?
And why would they pin everything on the mean? When your curve is low and wide the probability of your skill level being that value is way too low.
I’m lost.
The page has a picture of a player’s rank represented as a curve. It describes that curve with a mean and a standard deviation - which if I’m not mistaken are used to define the shape and position of a curve.
They talk about belief and uncertainty in a player’s rank.
It specifically says; “the ranking system does not yet know exactly the skill of the gamer”. And how maintaining uncertainty is important.
And you read blah, blah, blah - they just use the mean.
And furthermore I provided a link to TrueSkill maths (Moser) and a GDC talk (Menke) where they show some of the algorithms used to calculate rankings. I admit I don’t understand the maths all that well - but I do see a lot of calculations describing normal curves (with lots of sigma values floating around).
The basic maths underlying TrueSkill wouldn’t have changed.
And the TrueSkill2 clearly concentrates on applying values to the weighting parameters.
What’s the point, even if I had one. You are just going to go with your gut feel.
Are you resorting to offensive insults. Serious question?
The reality is that your skill ceiling won’t change much.
You can step away from a while and come back a bit rusty - but that’s ok, TrueSkill2 has got you - there is a weighting for time away.
But if usually doesn’t take long to get back to around where you were. But the reality is that you will still hit that ceiling. Hard.
It’s a fact. Not everyone can be Gold, let alone Onyx or pro.
Sure, if you are super dedicated you can and will push your ceiling as high as it can go. But we are talking hundreds, if not thousands, of hours of dedication. TrueSkill should move slowly with you, but there are always the seasonal resets if you have made a jump.
The skills are clearly transferrable. Controller use. Hand/eye co-ordination. Basic strategies.
Your sporting analogies are missing the point a bit. Halo to Destiny is a bit more like soccer to futsul.
And we’re not talking about having the same skill ceiling in both games. Some one who plays the Violin will hit their skill ceiling on the Cello a lot faster than someone who doesn’t play music. But for whatever reasons, they may always be clearly better at the violin.
I do get that.
And it is satisfying to improve. I have put in a huge effort to get from low to high Platinum.
But it has taken months.
For those who want more instant game to game gratification… or a slow build up… then XP rank is the way to go.
Thanks.
Since the other guy stormed off, can you respond to my post instead?
1 Like
Then the game should reflect on how I’m out shooting/playing other players that are a way higher than I am.
All observable evidence suggests this is exactly what’s happening.
You can raise your mmr by performing very well, but unless you win your csr will not rise, creating a substantial gulf between your observed rank and the average observed rank of your lobbies. Similarly, if you perform very poorly, your mmr will drop, but you won’t lose csr for winning and the same effect is seen in reverse.
In theory this should be corrected for by even a 50% winrate; if your csr is way below your mmr, you should get +15 csr for every win until you get close to your mmr, while losses are -1 or 0 change. And again, the opposite should be seen in reverse. But it doesn’t seem to be doing that very well.
I haven’t seen those threads, but this seems to be a fairly common experience judging by the number of complaints. As you said, some is noise and some is signal.
Ken has a much better eye at reading people’s game history than I do.
But yes. There is a flaw in a system that tries to put a positive bias on every win. People struggle with the concept of not seeing the needle move for a win. I really wish that they would just bite the bullet and say, “no points for you” (bonus if they do it with a evil Disney villain for the voicing).
And it comes down to match-making. If the server can only find games vs lower ranked opponents you will get long runs of meaningless (in terms of MMR) wins.
I think we agreed that your case was a bit of an extreme one though.
Was it all purely CSR drift on token wins. Probably not. Was some of it actual MMR you won (on beating good teams) and then lost (on losing to bad teams) - very likely. Did you get a bad hand in regards to some crucial 3v4? (we all think 343 got it wrong here).
And in the end, for all we know, there may have been a badly weighted parameter that allowed some anomalous CSR v MMR drift. They certainly lost the shape of the population skill curve (skewed to the right). And if so, I hope they fixed it in the recent updates.
There are always exceptions to the rule.
Certainly it’s not the way my account has worked. My drift seems to be the usual quarter to a third of a bar - which is consistent to the guys in my social group. There will always be a bias on forums such as this for when the system glitches a bit.
Have you noticed any difference in it’s behaviour for you post the mid-season refresh?
Yep. No system is perfect. And any system is going to have exploitable bias.
But a lot of the noise is just that. People unhappy with being told their rank and finding it very hard to get better.
I still have a lot of questions for 343;
- why do they persist in the MMR number?
- why do they HAVE to give CSR for every win? (FFS, just rip off that band-aid).
- does K/D have any weighting (above and beyond K/min)?
- how do they temper K type weightings for objective games?
- why don’t they have a breakdown for CSR gained or lost?
There is no reason why they can’t be more open. I understand they can’t put hard numbers on the weightings involved - but by telling us nothing we just get flat-earth type conspiracy theories about hidden MMR’s being used to trap players in lower ranks than they deserve.
It should.
If your CSR is lower than your MMR (eg. post placement) it will be reflected in the better quality of your opponents.
If you show consistent performance (ie. wins) at this level your CSR will come up to meet your MMR very quickly.
For some Onyx players, initially placing in low Diamond, they ended up back where they were supposed to be in as few as five or six games. It’s just a matter of stringing together a run of good performances in the games that matter (placement and immediately after).
If you’ve hit a snag - bad team-mates or the dreaded quitter to 3v4 - then your curve may take longer to narrow. It can take closer to 40 games if you take the long road.
Yea, it should, but it doesn’t. I should immediately be put in diamond if I’m beating diamonds. Forcing you to grind like in H5 is stupid. This is no different than any other game that uses this garbage csr with “resets”. It causes nothing but problems. I’m getting tired of forcing myself to go negative to keep matches fair.
If you play consistently for any period of time you will find your rank settles.
It can’t be immediately. No ranking system can be that volatile. Just because I beat an Onyx player doesn’t make me instantly Onyx - and the same in reverse - they shouldn’t be penalised to Platinum.
There may be a bit of a grind, but it shouldn’t be too harsh.
The other way to think about it is that each division (Bronze to Onzy) is structured to be a standard deviation on the population curve - and they aim to have a 3:1 performance gap for each division.
That is a Platinum player should beat a Gold player 75% of the time, but Diamond players only 25% of the time. But just because they lose 1:4 to Gold doesn’t make them Gold, or the fact they beat Diamond 1:4 doesn’t make them Diamond.
On one hand the resets are a marketing tool - to re-engage the players regularly.
But they also serve a purpose to refresh the rank and give you a new chance to prove your worth. Ranks do tend to narrow and become resistant to change over time (but not as bad as H3). The new season gives you a chance to flex your (new) muscles.
Why would you do that? Just play to your best ability and let the ranking take care of itself.
HaloTracker has your K/D around 1 and your W/L close to 50%. I’d say the system is reasonably comfortable in your rank.
A platinum should beat a gold 100% of the time and a gold should NEVER play with a platinum to begin with. But that doesn’t happen in H5 or infinite because 343’s ranking system revolves around one or two players being forced to carry. This was always a part of their scheme; to make it easier for a bad player. It’s the “everybody wins” mentality. Resets only make it worse. You’re playing against the worst or the best. There is no order. If I was diamond one season then get dropped to gold in the next, that is a broken system. I’m not “losing skill”. It stays with me forever no matter how many accounts I make. It’s not the nfl where I’m taking an off season of 12 weeks and somebody probably surpasses me. You grind and grind to get your rank and then it’s gone just like that.
Therefore, I will continue to keep doing what I’m doing at least until I eventually uninstall it. If it ruins others’ matches, oh well. Not my fault it isn’t Slayer to at least give me an incentive.
Not a good example.
A Gold 6 could be CSR 1199 and a Platinum 1 be CSR 1200. Of course they should play each other - and the games should be close.
As for the more general case of Gold vs Plat they are only 1 standard deviation apart on the population curve. Not huge.
And the rankings are literally structured so that each division is at a 3:1 probability vs each other. So, yes. An average Gold player should beat an average Platinum player 1:4 times. They literally use this as a system performance parameter and adjust things if it’s not happening.
The ideal of Platinum beating Gold 100% of the time is just not realistic.
I am solid Platinum. I have been beaten by my fair share of Gold players. And I have held my own, or better, against plenty of Diamonds. As long as I am roughly 3:1 vs Gold and 1:4 vs Diamond I can be comfortable in my zone.
I don’t get this narrow definition of 1 or 2 players having to carry.
Of course the best player in the team is expected to do better. But that’s not necessarily carrying the team. Their performance may be just adequate or below par despite having the most kills. It could very well be the lowest ranked player who “carried” the team by getting an extra kill, or even just because they died a few less times than they were expected to.
I’m not sure what this means?
There are plenty of lower ranked players out there. Hopefully all having fun despite not being Onyx. I know I am.
Resets just give the system a chance to breathe. Pretty much everyone comes out of the re-rank with the same rank they had before.
Except this one of course. They reset the rank distribution to the left. So everyone ended up losing a chunk of MMR.
If you are talking about this recent reset, then yes - everyone was kicked back a division. The population, for whatever reason, was skewed to the right. They needed to reset. Hopefully they have fixed the problem and it was a one off.
If you are talking in general. Yes, you can come out of placement with a lower CSR. But remember this is a conservative estimate of your MMR. It’s 3 standard deviations below the mean. You just need some consistency to narrow your curve.
Yes, and no. There is obviously a component of skill that is form based. Time off will dull that. But TrueSkill2 actually has a bit of weighting for time away from the game to help ease you back into it.
If you’ve been away a while you might be unfamiliar with new maps, game modes, or other sandbox changes. This may take some time to get familiar with. I know, personally, I struggle with maps and 3D spatial awareness. It takes me a long time to get moving around new maps optimally and my rank can suffer.
That’s really sad. And I mean your behaviour. Not the situation.
You shouldn’t be ranking up based on individual performance is my point. It should based off of wins and losses like every other game that uses csr for balance. You’re just describing the current system. It’s not that you’re wrong, it’s just that it doesn’t work. Most people are in platinum or high gold right now. I started in gold 5 and had to grind to get out of it, ie going positive and trying to win. It’s terrible. But after reading more into it, none of it is worth it. So now I play I how I want. I don’t think it’s bad behavior either. I’m simply adapting to a rigged system, and so long as I can control it by going negative once in a while, I can get a feel for getting matched up against other players around my skill level. You should be blaming the people that designed their only ranked gamemode around challenges to entice people to buy colors that were free in every other Halo. I don’t like playing like this anymore than you. Believe me.
You mostly don’t. You go up and down on the wins and losses. The rate of how you go up and down can swing a bit on personal performance. But the destination is pretty much the same.
Kills per minute is the classic parameter. It identifies the big fish in the small pond and ranks you out of there. It’s different to K/D and it’s ilk (which are usually reflected in the win). But as you rank up, the matching of your opponents in skill means that you simply can’t maintain the high K/min you need to make a difference. You can still play well and get a high K/D - but not the K/min.
Which is about right. Gold to Platinum is minus one standard deviation to plus one standard deviation (from the mean). Over 2/3 of the population should be Gold/Platinum.
Similar to this ideal distribution; https://i.redd.it/ql811ro73zxz.png
Isn’t that how a ranking system is supposed to work?
The “worth” is getting your highest rank possible. Not sure what else the system needs to provide to anyone.
This makes no sense. If you just play you will get games around your skill level. Above and below. By ruining the experience for everyone else you are essentially smurfing your account to get all your games ranked below your ability.
This has nothing to do with your rank.
If it’s having this effect, you may benefit from doing something else for a while.
Add me on Xbox and we’ll play. Just to show you I’m not bs’ing when I say it doesn’t matter how well I’m performing.
Im in platinum too.
In my head I see a bullish stock going up, then dropping slightly, then going slightly higher…. Repeat!
Until I stop investing my time then it will drop!