Why is it that even if you come out with a positive kd or high score on the team and you still lose, you get docked points on your rank? Many games i carried my team by either playing the objective or smashing kills left and right just to get my rank docked. It is a little bit frustrating.
1 Like
While I’ve never played ranked, it sounds like 343i wants it to be a team effort. They don’t want you to carry your team, they want you to assist your team.
1 Like
The problem with the ranked system is that CSR and MMR are completely different things.
All CSR does in its current form is cause problems. The game does adjust your MMR based on individual performance, but not CSR. The SBMM tries to make even 50/50 matches based on MMR, not CSR.
Either CSR needs to go, and make everyone’s rank their MMR, OR the system needs to change to match games based on CSR, not MMR.
1 Like
It’s a team rank. Not an individual one. Or at the very least it’s your rank reflecting your team performance.
Doesn’t mean a lot.
The K/D is not that helpful. And in some cases abused toxically.
The relevant metric for your MMR is kill and (to a lesser extent) death rates. ie. KPM and DPM.
You get a feel for how you went by looking up your expected kills and deaths post game on waypoint. They tell you how many kills and deaths you should be getting in that game mode for a game that went that length of time (ie. your KPM and DPM).
CSR doesn’t look at individual performance. Just the result (relative to the skill of the team you beat).
Scoring isn’t a useful metric.
When MS were building TrueSkill2 they looked at a whole bunch of stats; deaths, kills, K-D, K/D, KDA, assists, objective scores, medals, etc. And none of these added any extra ranking information above the result of the game. ie. if you are doing well in your expected metric then you probably scoring the W anyway.
The only ones that helped with the ranking independent of the result where kill and death rates. Which make sense when you think of them. You can fake a K/D or KDA against better opposition. But you can’t fake your KPM.
You need to carry your team to victory!
It depends on how tightly you knit them.
Your MMR reflects your current skill. And it seems 343 have left it a bit volatile to pick up changes in form (and stamp out smurfing).
The downside of your MMR jumping around is that you need to smooth the journey for the player. Hence the CSR.
If you think about your CSR as a “line of best fit” for your MMR then it reflects YOUR rank pretty accurately. They are not different things. Your CSR is just a value - that chases your MMR within the rules imposed by 343 (has to go up with a win, and can’t change by more than 15 points).
The problem is that your MMR can move away from your CSR. So you can be drawn into games that “appear” too high or low for your CSR. But they’re not because you are being matched on your current performance.
But people can’t cope with this as they try to micro-analyse who did or didn’t let them down at the end of the game (spoiler alert, it’s never themselves).
Some of the problem is that people don’t like the thought that;
a) Eventually their skill plateaus - and it takes a lot of time and effort to move it.
b) There is a natural form swing. Even day to day.
c) The skill gap isn’t as big as they think. An average Platinum player should beat an average Diamond player 25% of the time.
It does.
But the MMR pulls on the CSR - so in the next game the MMR will be even higher than the CSR and pull harder to drag it up faster.
So the CSR does respond personal performance. It just lags behind.
Which I personally think is a good thing.
The MMR is a more accurate reflection on your current performance.
Your CSR is slow to respond to fluctuations in form - so while matching on CSR will give closer matches on paper - they won’t be matched on current form. This will have two main problems; a tendency to mismatch teams, and because you are no longer matched against harder teams when your form good, you won’t be able to rank up as fast.
They could definitely do that.
But they would have to strangle the volatility of the MMR. Like they do with chess ratings. But just keep in mind you then can’t respond to day to day fluctuations in form, it will be slower to rank up or down (and if they are not careful we could end up with H3 type rank locking), and you lose a weapon vs smurf accounts.
Sound’s like this is where they are heading.
We’ll have to wait and see the devil in the details…
I imagine they will still use the MMR to guide placement. And then to control the rate at which your CSR climbs or falls. And finally, hovering in the background to detect Smurf accounts.
Easy. Because you lost, and that’s what should ultimately matter most. They could possibly make the highest performer lose the least amount of points if they do well, but they should still lose points because their team still inevitably lost in the end.
1 Like
Nah, I got fed up seeing absolute sandbags in Diamond 1-3 when I’m the same CSR as them and consistently the best performing player on my team, yet rank locked in mid-Diamond because I’m constantly needing to carry players who should be performing well but are relatively much worse.
So, I decided to test this nonsense, and I can prove to you that this is not what the system does. The behavior you describe is ONLY true if CSR is roughly accurate to your MMR. If those values diverge, this is no longer the behavior of the system.
I played 35 games to get to bronze 1. Zero CSR bronze 1 placement and it seemed like it was estimating my MMR around 100 or so. After a single game playing normally, it jumped me up to 1250 MMR low diamond games. Played a bit more and it was clear - CSR wasn’t going to move out of Bronze 1-2. Completely rank locked on CSR.
I then decided to test how fast it would allow my MMR to decay. I figured hey, MMR is volatile and I’ve not played much, maybe it will swing wildly around. Nope. Playing like a mid-upper diamond jumped my MMR by over 1000 points in a single game, but playing like a bronze 1 decays my MMR by about 40-50 points per game (or at least, the average MMR of the match).
A few things can be concluded from this.
- No, the system does not converge CSR to MMR in any general sense. This is only true if the two values are very close together to begin with.
- The system is clearly designed with smurf detection as its #1 priority, and is excellent at it. Nobody is currently smurfing in ranked in any meaningful way.
- The MMR system is tuned to be far more sensitive to good performances compared to bad performances. I believe this leads to many players having over-estimated CSR/MMR values, causing the behavior I’ve seen for months that there are constantly players around my CSR division that are much worse than me. The game assumes that if you perform like a Diamond 5, you are not a Platinum 4 who just played the game if your life, you are a Diamond 5. However, if you are a Diamond 5 who has a bad game and plays like a Platinum 4, the game thinks you are Diamond 5.
- The system in its current implementation is not designed to move your CSR very much at all. 50-150 CSR over your placement rank max.
It needs a re-work. CSR based matching while using MMR to guide the rate of CSR gain/loss would do the trick, I think, and make the experience much more enjoyable and less frustrating.
It’s frustrating when someone doesn’t perform “as advertised”.
And it drives me batty.
But honestly. If you go back through with neutral glasses - it’s more often or not a case of observational bias. You only noticed those players when you lose. You don’t tend to make a note of the games where you were sandbagged and won… or those that you may not have performed as well in (win or loss).
People have off days. Or have game modes that they don’t do well in (remember your CSR chases your Ranked MMR which is an average of all your game mode MMRs).
I just hope this wasn’t your 8,319th Smurf account 
I’m not 100% clear on what you were testing here.
You drove your MMR down. Which is fine. And then started playing relatively normally… so your MMR climbed quickly (attracting better opponents).
As you said. This is what it is supposed to do.
Your MMR would have been very volatile. First it was dragged down from 900 to 100. And with only 35 games. So TrueSkill2 would not have been very confident in your MMR at all (remember it takes closer to 50 games for edge or inconsistent ranks).
I can’t tell from HaloTracker what happened. It stops on Dec 11th after 35 games - so I can’t see how it behaved from there.
But if your CSR is now way below your MMR it should go up. Pretty much close to 15 points at a time (the maximum).
There is no way you are getting locked at Bronze 2 with only 30 odd games under your belt.
I’ll check HaloTracker again in a day or so and see if it plays nicely.
Match averages are hard to use off the bat. You don’t know what other multipliers are being added (squads, form, time away from the game etc).
And you can also swing with the mood of the match maker. I imagine your MMR curve is wide (which defines it’s volatility) so you would have a fair degree of variation in your games anyway (almost like placing again).
The system doesn’t converge them in a final sense.
The MMR swings with form. Josh once commented it can swing by 150 points in a day!
And when you look back at old graphs of MMR over time you can see they have a very up / down saw-tooth pattern.
The MMR follows along, moving up and down with the result, chasing the MMR in baby steps of 15 points a time.
It tries to converge… but the way the system works it never catches it. But nor is it supposed to.
But if you draw an average MMR over time, the CSR follows this quite closely. Essentially like a line of best fit.
The system is good at closing down Smurfs. And it needs to be as a priority. But I wouldn’t class it as the #1.
I’m not sure you’ve demonstrated this definitely with just a few games.
But it is true.
It helps find players skill faster. And given you haven’t played many games (or shown consistent skill), it’s no surprise it’s erring on the side of ranking you up.
Honestly, it ends up being an observational bias.
I’ve been there. Done that.
But if you go over their playing history it often shows better performances.
For whatever reason, your team didn’t gel, and performance suffers.
At this point that is probably true.
You go into placement with your old MMR as a seeding point. You don’t “start again”. So you come out of placement half to a full division below your previous rank. And then over the course of a day or two you climb back to your skill plateau.
Depending on the make up of the population for the new Season of course. That can influence your final rank.
Sadly we don’t have access to all the old graphs that Josh used to post on Waypoint.
But here is a typical one.
Showing how MMR can jump up and down while CSR follows in a consistent way.
https://drive.google.com/file/d/1-lVUagJzS-8duR6FTeo14L0c2FIvvt07/view?usp=sharing
No. I can guarantee you that the CSR on this account cannot exit Bronze this season. It’s simply not how the current system works. I’m getting the standard ~50% winrate, without much noticeable in the way of CSR boosting. After playing normally for an evening, I was back at a CSR of zero.
No. It’s not. On my main account, I am 25-67 in the past 100 ranked games. My CSR has not moved (because it will usually give me +15 CSR for a win and -4 or -5 CSR for a loss) - Diamond 3 for all of it. 18-13 is my average K/D.
The game is placing people’s MMR near what the game thinks is their theoretical skill ceiling. Then, it is trying to have their CSR converge to that number, which works if the placement CSR is close to the theoretical MMR but doesn’t work if the placement is off by more than 1-2 CSR divisions. The system is not tuned properly and is not pushing people down in either CSR or MMR sufficiently for bad performances. The end result of all this is a forum full of complaints where people are all saying the same thing - some form of “this system seems designed to force me to take 15 Adderall and go into mega tryhard mode 100% of the time or it’s not fun”.
It needs a re-work.
1 Like
*25-67 in ranked in my past 100 games is what that should have said
That’s very interesting.
Is the account Not A Smuf8319 (not working on Halo Tracker) or a different one. I’d love to see it.
What should happen, even with your 50% win rate at your more accurate MMR, is that the CSR gains for a win should be larger than for a loss. So your CSR should gradually work it’s way up.
Not an ideal W/L over those 100 games.
But your CSR not moving much is probably more a factor of your MMR not moving much.
And what you are describing is almost the exact opposite of what everyone else experiences… where you go up by a few points with each win and then fall back harder on the next loss.
You’re an enigma.
TrueSkill is very good at that.
I don’t think we can use your Bronze smurf as definite proof for how the system behaves for everyone else. Especially in normal ranges of CSR vs MRR.
But still very interesting. I wish I could actually see your account on HaloTracker.
I’ve been involved in a lot of Ranking threads over the last few years. And your issue is somewhat different to everyone else’s.
Usually it’s the oscillation of slowly going up and then quickly going down that draws people’s ire. But that’s just them failing to realise the system is built to oscillate and the reality is their rank isn’t going to change at this point. People really don’t like that.
It does that.
I would leave the MMR alone. Let it do what it does best. Rank the player. But leave in enough volatility to deal with Smurfs and allow the system to adjust for form.
The CSR needs a smaller scale. 0-2000+ lacks any degree of accuracy. Especially if you are pulling players from across servers.
I’m not saying go back to 1-50… but maybe 1-100. Or the oft suggested 1-117!
And it needs to be capped. This open end rubbish is woefully inaccurate and just encourages toxic grinding practices. And for the love God please don’t bring back Champ ranks from Halo 5. Sigh.
I still think the divisions are useful. From a population perspective at least. Each one is a standard deviation of the curve - so you instantly know roughly how many people should be in each one.
To keep people playing I would assign points or Military ranks for wins at your level. So even if you can’t rank up (you’ve hit your ceiling) you can still grind for something.
And then present your final Season medal at the end. Congratulations Darwi you were a Diamond 2 Three Star General.
My entire experience with this game’s ranked mode has made zero sense. I played a LOT in season 1 and early season 2 and it had me around Onyx 1550-1600. Then, after they reset the ranks (when they told everyone they would go down a division), it kept me in low Onyx but I started consistently matched against Onyx 1800+ players who slaughtered me, but it wouldn’t rank me down. It was so aggravating that I just quit for 6 months and went back to MCC.
When I came back, the above has been my experience. It placed me into diamond instead of onyx, I assume because of MMR decay after 6 months. Now, I’m outperforming the vast majority of low-mid diamond players, but the game is giving me 30% ish winrates and not allowing me to rank out of low-mid diamond.
Something about this system is not working in a way that makes any sense. It feels like a system that is designed to work well for the middle of the bell curve at the sacrifice of everyone on the ends of that same bell curve.
Ironically, if that’s the case, it isn’t actually any better in practice than Halo 2 or Halo 3’s systems despite being a LOT more sophisticated.
After moving everyone one division to the left everyone seemed to settle half a division or so to the right.
It was weird.
There is some weird stuff that happens from Season to Season.
I was wondering if a lot of people who came back to Season 2 somehow skipped the rank redistribution and took their inflated ranks into placment.
All anecdotal stuff. But that Season seemed to take a long time to settle.
Probably. And don’t forget you place at least half a division lower than you should on average (wide MMR curve) - and there is a maximum starting point (it was D1, but I think it is now D3?).
It’s always going to struggle out to the tails. You just don’t get enough quality matches with similar opponents on both sides to help you rank up and/or down.
But mid-Diamond should be fairly consistent.
I was pretty happy with my last Ranking session. Despite the 200+ average pings I somehow still made Diamond.
But since the last reset I can’t any ranked games at all. Zilch. Times out everytime.
I’ll anecdotally add to this–I’ve had a similar thing happen. I was around 1600 in season 1, then consistently high diamond in previous seasons. I placed in at like mid platinum this season, and I’m definitely not playing my best Halo. But until last night, I was on a 3-17 streak. I gained max CSR for every win, but just was not winning.
You especially get into nasty cycles because at lower ranks, people have less and less idea how to play objective, and even if you slay well, you can still lose (and then you start overplaying objective and end up with poor individual performance, yadda yadda).
I’m not as good as I was, but I’m better than my CSR. I just can’t win enough to get it up, regardless of how much CSR tries to compensate by giving me loads of CSR every time I win and relatively little loss when I lose. The volatility of matchmaking doesn’t help, as again, in these ranking levels the game is often decided by which team is saddled with a player who’s going to go -15, -20.
I also see a number of Onyx nameplates or whatever in these games, suggesting that other people are in a similar situation where they’ve placed in way below expected and can’t win consistently enough to get up, despite individual performance.
2 Likes
I just think the game currently operates with massive error margins between player’s CSR and their average MMR.
Just like the MMR of a population follows a bell curve distribution, so also an individual player’s ability follows a bell curve. Everyone has good games, typical games, and bad games. The system currently expects everyone to be playing at their “good game” level all the time. I suspect this is done primarily to prevent smurfing. I also think this was a contributing factor for why so many players placed Onyx in Season 1 - the system was even more overtuned than it is now.
I’m also pretty sure MMR is only tied to your kills per game, so there’s often no incentive to actually play objective.
Well, you still want to win since that affects your CSR. But yes, you do also end up in the nasty conundrum of, early into a game, realizing the team has no chance of winning on objective–so, do you just decide to give up and focus on kills to prevent greater MMR loss?
If I see three teammates going negative AND doing worse than me on the objective count, which happens in about a third to half of my ranked games, I just play slayer.
I play ranked to get even games where all the players are roughly similar in skill level. The game modes are ALL designed around that kind of competition. But the SBMM is so atrociously designed that this just doesn’t occur in a great many games, and I’m tired of it.
1 Like
I should be more like you. I’m a stickler and usually still try to play the hill, cap a zone, play for ball. Just furthers the spiral though, as you end up with fewer kills, more deaths, and worse MMR in the long run.
Nothing more fun that playing King of the Hill when you’ve got a teammate or two camping on the wrong side of the map!
1 Like
It’s just exhausting to feel like you have to constantly carry teams in ranked. If the game wants me to carry games in social, who cares, but I expect to not always be the #1 or #2 player on my team in ranked. I get really excited for those rare games that I realize I’m #3 or #4 on my team and I have to really try my best to not be the reason we lose - that is SO much more fun than the typical experience of “oh, my team is all trash again and I need to go +20 for us to win… how on earth are these people only 50 CSR below me? ughhhhhhhh”
Yeah, that’s actually a fair point. As far back as my Halotracker goes on the first page (25 games or whatever), I’ve won exactly one game where I wasn’t positive–and it was a CTF game where I was -1. I know it’s K/min not K/D, but the story would be same if that stat were easily available.
Basically, I appear to have no slack.
1 Like