I played two matches tonight one complete and the other where the entire team quit out and I gained no rank for the match and at the end of the match it said Rank unavailable for all players.
1 Like
I donāt think itās unusual for quitting players to show up as āRank Not Availableā.
1 Like
Hopefully itās just a bug I just won 4 games in a row and gained nothing. All the games I went positive and normally Iād get at least 3-5 csr.
https://imgur.com/a/xdE9gCt
I did see this in another post.
Itās in the known bug list:
āIn Ranked Arena playlists, completed matches occasionally do not count toward your rank and will show no stats at the end of the match.ā
This happened to me a handful of times before the reset in Season 1. Iād win a match but get no credit for it and all of the stas were Unavailable.
I think so. Iāve played since and have gained/lost rank so it mustāve just been a momentary thing that only impacted a handful of matches.
1 Like
The answer to the title is Yes. Until they overhaul TrueSkill, decouple or weaken the affect of casual matches on ranked MMR, and impose some type of limit on ranks queuing together ranked will continue to be dysfunctional.
Note, I havenāt kept up with 343s blogs so they may have addressed some of these issues. But given the pace they set in S1 Iām pressing x.
TrueSkill2 is still solid.
Any system can always be improved. But I would have thought weāre talking ātinkering around edgesā as opposed to āoverhaulā.
If youāre talking about the drama people get into over K/D vs playing the objective etc. Itās not worth worrying about.
Iāve collected data on over 100 player games comparing CSR change to K, D, K/D, KDA, K/min, total score, and objective score. So far there is no statistical correlation. Iām aiming to get the sample size closer to 1000 to be sure.
I really donāt think there is much.
That Reddit post re: seeding a new ranked account with a Bot MMR really freaked people out. But thatās a unique one off scenario.
Anecdotally Iāve seen no influence in real play.
And my (brief) research into average opponent MMR hasnāt shown a consistent effect after 2 weeks.
Plus we can now see our separate (non global) MMRs in at least the three playlists. Iām 1250+ in Ranked. 850 in Rumble Pit. And (donāt laugh) 300 or so in LSS.
This is a big one.
I think itās the number one problem today. People grinding rank using smurfs and sandbags.
1 Like
If they gave us the option for BR starts in other playlists, some of us wouldnāt play ranked with such a broad CSR range. Weāre not all smurfs and sandbags. Some of us just prefer BR starts.
What about wins/losses in casual play affecting the skill of opponents you match against in the short term? For example, if you lose a bunch in casual play and then switch to ranked, will you have an easier game?
Just anecdotally, it seems that once I have an established MMR in a playlist, then Wins and Losses seem to be weighted more heavily in determining my opponents from game to game.
What Iāve been looking at is the rolling average of opponentās MMR.
Going from Tuesday night chill with mates back into ranked doesnāt seem to make much difference when you take into account the way the MMR dips with runs of L.
Still need a lot more data.
Anecdotally I agree that W/L gets more important as you go. I get the feeling that placement in particular favours kills. A weighting change of some sort? And if this does exist it would be interesting if and how they reduce it.
I did notice that when I was playing the occasional game of RP to convince myself that it and LSS have separate MMRs - I was on a run of āLā for my ranked games. But while my opponentās were softening in Ranked my MMR in RP stayed rock solid at 850 to 852.
The easiest way to be sure is to play and deliberately lose a heap of social games - while tracking team MMRās before and after. But I canāt bring myself to deliberately lose. While I am happy to play social with a different intensity and mindset, Iām a little too ethical to deliberately lose. Weāll see what dominoes fall.
1 Like
Where can we see our Ranked MMR? I see one on HaloDataHive for FFA MMR, but no ranked.
My logic kind of goes along the lines;
We know their is a match making weighting for āformā.
And itās probably part of the shared data across different playlist MMRs
And we know match making canāt be too perfect otherwise everyone gets locked in on a straight line graph.
So, you could use āformā to flag when to nudge opponent quality and give the player a genuine chance to rank up or down.
Being a shared, global value, I could see this effect rippling between playlists. And standing out more in Infinite because there is only one ranked playlist.
But the important part is that it would only be a fleeting effect as losing to the tougher team (or beating the softer ones) would reverse your āformā weighting.
And set up this way your actual MMR isnāt changed at all (if results fall as expected).
But anecdotally I havenāt really noticed it. It may be there. But if it is itās subtle.
1 Like
My personal feeling is that it is leaning towards being too perfect but having more ranked playlists would probably go a long way to mitigate that feeling. (My feelings are way more unbiased than other peopleās feelings 
)
1 Like
There was some forum of discussion - youtube or forum post - where the critique was something to the tune of TS was too rigid from a design standpoint. Effectively the argument was TS is overturned for balance and fails to take into consideration some Players arenāt looking for a perfectly balanced match in non-competitive modes. Itās been a while since I consumed that specific piece of information so Iām very fuzzy on the specifics but I remember their being a variety of very good design philosophy points against TrueSkill from a former 343 or possibly Bungie dev.
Not familiar with the reddit post; There was a post somewhere on this forum relating to someone sandbagging in casual and seeing how it effected the aggregate MMR of their ranked matches. I am unaware if the posts are related/clones of each other, itās been a while since Iāve read it, and I donāt remember how long or how many games it took to get easier matches.
My opinion is sandbagging to get easier ranked matches shouldnāt be feasible even as an edge case. My conjecture is if MMR is shared across queues and TrueSkill2 is built off of Probabilities and Infinite doesnāt have a solid anti-cheat and has poor reporting mechanisms then Players can influence competitive matchmaking by throwing non-competitive games with minimal risk. This is something the paper on TrueSkill 2 confirms in Section 7:
ā¦each player receives a small increment to their skill rating after playing a match, regardless of the outcome. These increments are small relative to the difference between winning and losing a match, so skill rating will still go down if a player loses repeatedly
I took a long break and therefore am uncertain where Infinite currently stands with an anti-cheat solution; In any other game I would say botting to lose games would be detected and banned either through playtime analysis - no one can game 24/7 - or through reports of inting. Given Infiniteās track record in Season 1 Iām not brimming with confidence. Disclaimer, Iām not suggesting people do this, Iām explaining why I think TrueSkill2 requires a strong anti-cheat/reporting mechanism to shore up itās explicitly stated weakness.
Also to be clear, Iām not an expert on TrueSkill2 and the paper makes reference to a tunable param for calculating skill decay (see section 4). I also only skimmed the paper.
The Paper: https://www.microsoft.com/en-us/research/uploads/prod/2018/03/trueskill2.pdf
1 Like
Itās tricky.
Most SBMM advocates are very happy with strict (on the individual) and broad (on the team) in social.
And for a wider tolerance for mis-matching in Social.
So you are pretty much trying to create 50-49 epics in Rankedā¦ but not so much in Social.
In the end it depends by what the particular players mean by āarenāt looking for a perfectly balanced match in non-competitive modes.ā
The problem is that a large proportion of the vocal anti-SBMM players just want ārandomā matches so that they āchillā all day stomping average players.
If weāre thinking about the same posts / discussions the end point was that we should be thinking about designing and creating different experiences to make the game feel less competitive.
Which makes perfect sense to me. Weāre taking hyped up competitive players and then putting them on the same maps with the same game structuresā¦ and expecting them to play differently!
The classic Reddit posts had heaps of people trashing their social MMR and seeing if it could help them ranked up. Lots of people didnāt see a difference.
And I would have thought by now weād have seen posts or videos of average joes showing off their Onyx medals if there was really any substance to this. But we havenāt.
Sadly itās true. Iāve noticed lots of them. In the mid S1 reset I even hit the same team five times. And I followed their trajectory on Halo Tracker.
They had;
- 2x Onyx players grinding for rank.
- 1x Gold player. Obvious Smurf.
- 1x Silver sandbag
Itās a pattern Iāve noticed even more prevalent in S2.
Apparently there are Discords around where Onyx players can find sandbaggers to play with.
Shouldnāt be. Thatās not how TrueSkill2 is set up. Itās not the way H5 worked.
Itās been harder to prove in Infinite because there is only one ranked playlist. But now you can visit sites like Halo Hive Mind and look up your MMR in the last FFA game you played (I assume they can extract the average team MMR for your team of one).
I have three clear MMRās. Last Spartan Standing = 300 or so (I suck). Rumble Pit = 850 (I play chill and go for challenges). And of course my sweaty Ranked MMR of around 1200.
But there are shared data between all the MMRās. Like form and time away from the game. These can have cross playlist effects.
Yep. You can think of your MMR curve as a probability curve. And the maths involved in matching, working out the expected result, and allocating rank change - all involve maths that produce result (probability) curves.
With you 100%.
While I think everyone should read the paper - the problem is that it skims over a lot of things - and expects you to know the maths inside out. But the maths is hard (google Moserware for some resources if your math skill is better than mine). And they deliberately donāt discuss too much detail about the weightings and their strengths - I guess to stop people from manipulating them.
The best take aways from the paper are the discussions on Squad weightings and Kills/minute.
1 Like
Today the game has decided to put me against a guy who peaked at onyx 1900. I am not onyx 1900, not even close. I shot 69% accuracy and went 3-18 
Ranking system is pathetic. How do I get lobbied with him
Looked like a tough game. 75% in their favour. Ouch.
But how it happens is that they are lobbied up. And at some stage the match making has to find them a game. You may not have been waiting long. But they probably were.
The only consolation is that you wonāt have lost MMR on the result.
But important to differentiate. Not a problem with ranking. Issue is with match-making and squads. Leave TrueSkill2 out of it. 