Title says it all. I don’t get it. I win like 10 matches in a row and barely rank up but when I lose once or twice my rank plummets. What is being measured here? Does anyone know? It is very frustrating. A simple straightforward ranked based on wins and losses or your position on the leaderboard might make things a little more enjoyable? IDK its just another area where I feel I’m being manipulated by the developer.
There are lots of threads on the ranking system.
Essentially if you win you rank up. Lose and you rank down.
How much you rank up or down depends on the average rank of the other team. If you want to rank up you need to beat teams ranked above you and not lose to teams ranked below you.
There is a slight weighting for kills per minute (hopefully just in Slayer).
If you win 10 games in a row against teams ranked equal or below you then you shouldn’t and don’t rank up (your internal MMR stays the same) but your external CSR slowly moves up.
If you eventually lose to a team ranked below you the system brings your CSR back to the MMR (it assumes you didn’t deserve the increased CSR). If you beat a team ranked above you then the system moves your MMR up towards your CSR (maybe you did deserve at least some of it).
The problem with matchmaking is that you are more likely to get a couple of bad team-mates (who your style doesn’t gel with) and lose to a team ranked below you. So you fall into a treading water pattern around your rank ceiling.
The very, very, good players just don’t lose - so they end up grinding out CSR’s into the 2000’s and beyond.
1 Like
Wrong. Now, by reading the rest of your post, I see that you understand the system better than that, but others may get the wrong impression. Winning and losing is a factor, but it’s not the only factor.
Without considering the technical details of how it actually works, the aim of the ranking system is to assign a rank to each player according to their actual skill relative the rest of the population in some abstract, ideal sense. But of course, 343i has to use some sort of objective criteria in some sort of formula to determine your rank. How well that that formula can make a judgment in your skill could be a matter for debate, if we were allowed to know all the details of how it works, but remember, everyone is playing under the same system. Suffice it to say that winning and losing is far from the only factor that determines your rank.
Now, it is a fact of life that there is always someone better than you. In fact there’s probably thousands or tens of thousands of players better than you, unless of course, you are one of the few actual pro level players.
Suppose there are 4,999 players better than you are, as determined by some Halo God that is able to perfectly discern player skill. So you’re ranked in 5000th place. Well it doesn’t matter if you win 500 games in a row. You won’t move up to rank 4999 if the existing person in rank 4999 maintains their skill level and you don’t actually get better than them.
I’m not telling you that I know the rank system is working perfectly and therefore the way you’re moving up and down in rank in relation to your game performance is perfect. What I’m saying is that you or I don’t have enough information to determine if it’s working correctly or not, because your rank is partly deterrmined by the performances of all the other players in the population.
The rank tiers are also assigned on a curve. Even if everyone playing Halo were all excellent players and the skill gap between all players is small, some players would still be assigned to Bronze, some to Silver, etc etc. The rank system is not an experience system. You don’t get to climb it just by playing, even if you play well. You only climb if you actually prove that you’re better than the players currently ranked above you.
The mistake with this is thinking this only happens to you. This happens to everyone. The players that are super good are still able to rise above and prove they deserve to rise in rank.
2 Likes
I find myself in the opposite end of the spectrum. I can lose 5 games in a row and my rank will not change in the Solo/Duo playlist, then I win one and I gain 10-15 CSR.
I’m currently ranked G4 and am constantly matched against Platinum and Diamond players. Regardless of win/loss I am generally competitive in my games.
Simple. The ranking system is BROKEN. I also find myself in a situation where I “walk up the stairs, but fall out a window.”
I am 5-2-1 today with an win against a D6 and loss to an Onyx. I have LOST xp today. I have been top in lobby in nearly all games. On the losing games, I am nearly always top on my team.
Something happened on the 19 JAN update, IDK if 343 is just trying to get ppl to play Attrition and/or go for rewards, but ranked seems like a futile effort at this point…
BaconAndEggsUSA
1 Like
Bingo. These guy defending 343 works for them. It’s obviously broken. They use words like “your skill”. How do you determine someone’s true skill with issues like servers, Desync, cheaters, wall hacks, aim bots. It’s sad, but it’s definitely broken. What’s funny is he is on every forum that talks about the ranking system which lets me know who works for them. The system sucks.
2 Likes
Weird you post on every ranking forum. Just fix the game over there at 343. Address the cheaters and halo hacks.net. Then I’ll consider what you are saying.
Trying to keep it simple.
All we really have to go on is the discussion paper published by Microsoft.
This was in 2018.
Sadly we don’t know if any more parameters have been added, or if their weightings have been changed.
But the “tone” of the paper, if you like, is that wins are the important part. And any “tunable parameters” are a weighting to improve on the efficiency of TrueSkill (which was purely win/loss).
I would have thought the number, for me personally, was in the millions.
I agree that this is the crux of the problem.
The system can’t rank you up if you aren’t playing (and beating) teams that are ranked higher than you. It literally has no useful data to work with.
Another analogy would be a Tennis pro who spent a couple of months playing in local amateur events. They would pretty much smash it out every match - but nobody in their right mind would expect them to use those results to move up the ATP rankings.
The problem in matchmaking is that there are too many games where the result is expected and not useful. And this is not a criticism or necessarily a fixable problem. It just reflect real world populations separated into local (server) groupings.
Yep. And the problem is they have lost control of the right hand side of the curve.
If you continue the division/tier points into Onyx the skill ceiling should fall at 1800 (Onyx 6 would be 1750-1800).
Of course, it’s an open ended system, but only a handful of players should venture past 1800 - and definitely not well into the 2000’s as we are seeing.
343 know they can’t control the mushrooming of MMR. Matchmaking just simply can’t provide the spread of games that stop the top players/squads from going on winning streaks. The have tried, artificially, by limiting the amount of CSR gained from wins (in Onyx) and weighting against losses… but they just end up frustrating the bulk of good players who get caught in the middle “treading water”.
Ideally what they would do is normalise everyone’s results back to a normal distribution from 0 to 1800 on a regular basis. Daily or even weekly. But I don’t know how they would explain / sell that to the general population.
I don’t think I ever insinuated it as such.
You can see from the multiple threads on ranking that are popping up shows that it’s happening to a lot of people. Particularly around the high Diamond to mid Onyx level. Which is co-incidentally the group that is most likely to care about the exact value of their MMR.
And yes, the cream will continue to rise… but that actually just highlights that the system is not working at the top end (as a number to grind).
1 Like
What can I say. I’m at work and bored.
Plus I’ve always been fascinated by ranking systems.
As far back as Halo CE I used to have spreadsheets and grids that worked out who killed who in our LAN parties, trying to make ranking tables. My favourite was one which took the last five games vs each player and gave you a point if your W/L was better.
Sadly no. Don’t get me wrong… I actually do a lot of programming as a hobby. In a different life I probably would have been in IT. But no, I work in an Emergency Department… and as you can see I get way too much admin time. 
It’s important that you separate the ranking from the game issues.
All of those issues suck big time… but they don’t make the actual ranking system better or worse.
So how can you truly say it’s about skill then with all those issues? Something is clearly off and halohacks.net is a thing.
The system is doing the best it can with bad data.
I don’t know how to fix what people are experiencing… other than abandoning the MMR as a visible stat all together. I would move back to simple division and tiers across the board (including Onyx) and give people an XP system to grind (weighted for wins, medals, and objective scores).
The ranking system is about data. Pure and simple. You feed it data and it spits out your rank.
What you are describing is “bad data”. The influence of exploits and flawed game mechanics.
The ranking system is completely independent of the “data quality”.
You can complain about the game issues (quite rightly) and the ranking system (that’s up to you). But you can’t conflate them.
Even if you assume desync, cheating, etc are widespread problems, in order for us to think the ranking system is broken, then you would have to show (or assume) that these problems are affecting certain kinds of players more than others. Otherwise, it would be more reasonable imo, to assume that the problems affect all kinds of different matches/players and their affects on rankings is sort of averaged out. That doesn’t mean the problems don’t need fixing. What I’m saying is, it’s improbable that the problems occur just so that they only conspire against you to make your performance worse and everyone else’s performance better.
1 Like