Also unrelated note since it seems like you’re in a similar place: have you noticed how awful the game performance has been recently? I’ve started clipping every time something outrageous happens–magic melee lunge, phasing through, no hit reg–and I’ve realized it’s genuinely happening 2-3 times per game. I’ve got a great clip of me shooting an unshielded opponent twice point blank with a bulldog the other day and not getting the kill.
Man does it get old trying to play competitively when things are so unreliable.
Do you think this could be one of the drivers of the CSR behaviour you are describing?
With the low population the makeup of the squads is far from ideal. And it probably hurts the top end of town the most.
You invariably end up being the top dog grouped with a bunch of average players.
And critically - your MMR is going to be higher than the average MMR of the opposition.
We know the drivers of increasing your MMR/CSR are;
Winning.
Your MMR vs the opposition MMR (need theirs > yours).
Having an MMR > CSR (to pull it up).
KPM > expected.
I put them in that presumed order. But we don’t know the weightings. And we don’t know how the algorithm scales; eg. over the gap in MMR between you and the opposition team. It may not even be linear.
In the case of someone with a really high MMR being put in low MMR game pools; you win (great) but your MMR gap over the opposition looks like it tends to over ride the other factors.
It really only leaves your KPM as a way to push your rank up. Which is probably why you describe that feeling of your option to slay - a lot. And again, we don’t know how the algorithm scales. So meeting your expected kills and deaths will minimise your ranking down in a loss - but we don’t know how high you have to push it to actually rank up.
And you’ve probably locked yourself into a situation where the KPM required to over ride the MMR gap between you and the opposition is too high to do any consistent up ranking.
I don’t know how they can fix this.
Obviously they could have a regular check to see if someone’s CSR is lagging behind their average MMR. It would be easy to have a regular calibration to find and isolate those outliers.
As for those who find their MMR being locked. The way ranking systems work is that you are at the mercy of the population at hand (which is low). And combine that with the rank debt of people who drift in, get a rank, and then leave until next season. It’s frustrating.
To have the chance to rank up you need a consistent supply of matches where you are the (slight) underdog.
We need;
A clearer explanation from 343 about how rank / CSR works.
Tools to look things up for ourselves (MMR over time, KPM per game mode, etc).
Incentive for players to keep playing ranked though the whole season (more XP, military ranks, rewards for ‘x’ number of wins at a rank, etc).
Yup, at end of day the finer points of the ranking system really aren’t relevant at this point. There aren’t enough players to sustain healthy ranked matchmaking, period. No system works well when the population isn’t there. Heck, some people can’t even find ranked matches at all.
it is what it is. Turns out, can’t release a game where 75% of matches are on three maps for an entire year and expect people to keep playing, to say nothing of performance and all the other issues.
I have no idea what is causing the CSR behavior I’ve been describing other than “something is broken about how this system is designed and/or implemented and it’s incredibly frustrating”
I think this is a case of someone at 343 getting too fancy for their own good, and the whole system has collapsed under the weight of its own complexity and supposed cleverness.
A simpler system that is more transparently focused purely on either MMR or CSR would be better than this current one where the game attempts to use black box voodoo to match games based on a hidden MMR that is being adjusted by a different mechanism than the visible rank shown to the player in their CSR.
I kind of feel that most of the issues people are having circle back to low population.
I haven’t been able to find a ranked match for months.
That.
And the combination of no incentive to keep playing (just get your rank and then stop until next season) plus toxic grinders of CSR (eg. using squads to manipulate the MMR of the other team).
I think it’s quite the opposite.
It’s quite a simple system. To the point that it over relies on quality match making for it to work (ie. a big stable population).
To use a programming analogy. Garbage in, garbage out.
But it’s not unique to Halo.
If you made a Tennis tournament where the #10 partnered with the #150 in a doubles game vs #80 and #81 - how the Hell does the system make a judgement on how much the #10 should rank up with a win?
The system is quite simple.
The MMR vs CSR is just poorly explained by 343. And we no longer have someone like Josh answering ad-hoc questions here on Waypoint.
All they need to do is make the info available to everyone. It would be very simple to have graphs of MMR over time. It should show clearly our MMR for each game mode (Slayer, Oddball, etc) , the average MMR, and then show how our CSR relates to this.
The MMR doesn’t need to be “hidden”. And simply letting us see how it is changing over time would expel most of the conspiracy theories doing the rounds.
I don’t get why we can’t look up our Social MMR. Most people don’t care - and for those that do - it would dispel a lot of angst.
And KPM. We know it’s the metric. MS published it in their paper. So let us know our KPM vs MMR in each game mode. They already publish the data via the expected kills and deaths - so make it easier for us to access.
I disagree. It’s two related but different systems (CSR and MMR) functioning in a way that is nearly impossible to discern. That’s not simple.
Simple is Halo 3’s 1-50 system. Win = go up. Lose = go down. Fewer XP/games played = faster rate of change. Simple is chess.com ELO rating.
I’m not saying I want to go back to Halo 3’s 1-50 system, but I don’t agree at all that the current system is simple. It’s an ELO system combined with an ELO approximation system that doesn’t function the same way.
IMO the MMR system should be taking into account the result of the game (win or loss), the players total kills, the players total objective score, and the players kill/death spread.
Maybe it’s already doing that and the weighting is just bad and way too far on the KPM stat, I don’t know.
What I do know is that if you need a TED talk to understand a game’s ranking system, that game has a bad ranking system.
If you haven’t read it - the MS paper on TrueSkill2 is illuminating.
They looked at a whole bunch of metrics to see if what increased the speed and accuracy of ranking.
Spoiler alert - it’s only KPM and DPM.
Apparently they looked at K, D, A, KD, KDA, K-D, Objective scores, and medals. And probably more I don’t remember. But none of these increased prediction over the win. And this makes sense since it is already implied that you did well in these scores on the back of the win.
KPM on the other hand is a different beast. If you think about it correlates with your ability to win a 1v1. And it’s only going to be high if you are the big fish a small pond. So, it works to rank you up faster. But not necessarily higher - as your KPM vs people of your skill level will naturally equal out.
Covered by KPM.
Like K/D - a metric that is easily abused. And would encourage toxic play.
Not a good metric. At all.
When I play with my kids I have to hold the Oddball. A lot. But it doesn’t mean I should rank up on that basis.
The real skill in any game is winning a 1v1.
If you rank people up on the basis of objective score or assists - then they just end up getting thrashed because they are out of their depth slayer wise. This frustrates them and their team mates.
Over the years I’ve done a lot of placement games with my kids. They are better than me and invariably come out a division or more higher than me. If you look at the stats I have higher objective scores and heaps more assists. And my K/D or K-D is usually as good, if not better than theirs (I hate dying).
The only stat they consistently do better is total kills, or essentially, KPM. So they deserve to rank up higher.
So it seems to work from that perspective.
And interestingly, I played a custom warm up game with one of my Sons the other day. One v One. I am mid-Diamond at best. He is definitely mid-Onyx (although never really tried to rank). He was up 10 to 3 when we had to stop… which pretty much matches the expected 75:25 that you would expect from two players a division apart.
No TED talk required.
Just better communication from 343 and stop hiding our MMR (it just fuels a bunch of conspiracy theories).
And for players in general to stop trying to micro-analyse their rank on a match to match basis. It’s not how it works.
A bit of a bad patch before that, I’ll give you that.
But going back over the first page of results we’re about 30W and 40L.
So it would be pretty even if it wasn’t for whatever happened on Dec 22 (you went 2 & 11). Maybe it was just being a bit rusty for not having played Ranked for a couple of weeks?
I think it’s just human nature to reflect disproportionately on the bad days. It colours your perception of the big picture.
Since the rank reset on the winter update.
I’m gaining more CSR with wins than I lose CSR with losses.
Which is much better compared to Season 1 and 2 where losing tanked your CSR.
However I am playing currently in Platinum 3 for this season can’t seem to budge it past P4 due to poor matched games. Whereas before in previous seasons I would fly out of Platinum and get stuck around D2/3.
Which is what should happen early in the Season. When your MMR is likely to be higher than your CSR.
Sadly I haven’t been able to get into a single ranked game this season - so I don’t have a feel for what the spread is like.
But lower population is likely to shift people to the left. So it may be a bit harder to match your rank of previous seasons.
And the rank gaps within teams seems to make it harder for the better players in each game to rank up (personal MMR > opposition team average) and lower ranked players to rank down (MMR < opposition).
I imagine there will be pockets of rank that tend to be more affected than others.
As good as TrueSkill2 is - it can only work with the data it is given. And the old analogy about garbage in…
I kind of take it back about the mismatch… if I play early the population is low and sporadic, but I’ve just had a few games at peak time and the matching was good. The highest I played was a D1 and lowest G6 but I think they partnered up. The rest was all around my CSR.
But those morning lunch time games it’s like the Wild West. Might stick to BTB in the day from now on.
Why is it that someone with the same rank as me does not play as well as me, but loses less points in a loss? That right there does not make sense as well.
Your skill vs the opposition team skill (upsets earn more).
The pull of your MMR (your CSR resists moving away).
Personal performance makes ZERO difference to that game. But your KPM and DPM do affect your MMR and can affect your CSR in the next game
And because 2 and 3 are different for each player - the change of points is different.
You may have the same CSR - but you could have quite different MMRs.
It makes sense. It’s just woefully explained by 343. And because they insist on keeping everyone’s MMR hidden - nobody really knows what’s going on. And this just frustrates people and leads to all sorts of conspiracy theories.
They could fix it by;
Giving the option of looking up your MMR vs CSR over time.
Not showing everyone else’s CSR change (that was a very silly move).
- so I don’t have a feel for what the spread is like.