So SBMM is designed to decide when you lose…?

Again, the end point is the same.

But we are assigning two very different algorithms.

In reality the match maker tries to balance teams so that either team can win. Preferably with the same chance. The mix of any uneven games is random. So your win loss naturally falls into a 50:50 rate.

The fantasy option is that the match maker specifically watches out for any player that might be improving and deliberately sabotages their win rate by giving them worse team mates, increased lag, assigning desync, setting unachievably high expected kills, and generally sending bad vibes through the interwebs.

Average team MMRs.

I am more than happy to gather evidence to back my claims.

But this is your claim.

When I wanted to prove social games didn’t affect opponents in ranked I recorded a couple of hundred games, set a rolling average for the last eight games, and looked for patterns leading into and out of my weekly social games with friends.

And then when I wanted to show that post game CSR wasn’t correlated to K, D, K-D, KD, KDA, KPM, DPM, damage, score, or objective play - I mapped over 500 game results and showed that there was no strong relationship.

Yes, I have a strange fascination with spreadsheets…

The behaviour of TrueSkill and TrueMatch are well documented. 343 have categorically denied any interference.

The key to ranking up, as it has always been, is to get better and start beating better teams.

If the stat says one team has a 33% chance of winning - and they win 33% of the time - that actually validates that the ranking system has it right. It would be great if there were more games closer to 50:50. But if you want to start proving to the system that you deserve to rank up then win more than 33% of the time.

I don’t understand how it’s a conspiracy to suggest playing on a bad server influences your win/loss. If you are playing with 150ping vs players on 30 ping, doesn’t matter if you are a high-ranked player if your shots are significantly delayed.

2 Likes

It’s not a conspiracy to say ping affects the way you play.

The way you gel with team mates.

Or desync.

There are lots of factors. And not all of them are in your control.

But if you want to say all these things are deliberately controlled by 343 specifically to decide who wins and loses. That’s a conspiracy theory.

Of flat earth dimensions.

1 Like

It is most likely not intentionally controlled by 343. However, the games I am placed in vs more skilled players and also in high ping servers, those games don’t accurately gauge your skill level. What makes them worse is you lose a decent amount of CSR which is extremely frustrating.

I’ve seen this number change match to match, and even in my own team of 4 friends that numbers total amount drop over 100 points into the next match “after” winning a match. Anecdotal, sure, but…the noticeable thing was the ping of the match where the average mmr was less was very high .

That lead me to think even the network data gathered in the match maker is used to generate that number. We had won both matches, but I had a huge question mark after it.

Why does that number fluctuate even to the left after winning? I mean…I can guess about it…I can try and just assume its cause although we won, perhaps no one passed their kpm benchmark? But…all I can remember is we generally all played well.

And what does the second set of numbers represent? Cause sometimes its like

Average mmr 1225.899812 (arbitrary numbers) but I’ve never understood that. The numbers after the decimal point. I get like…fractional values and such but…is that all that is? Or something else?

  • yes. The game has been reinstalled for the 125th time. Halo is the abusive ex gf that I just seem to always go back to. We fight and make up too much. I love the halo forumula too much for my own good. Lol

That’s not a fantasy, that’s what it’s doing to create a 50% win/loss. If it was specifically just building even teams and that’s it, then it wouldn’t be also creating a predicted kills and deaths, because your kills and deaths should be the same in an even match every game. It also wouldn’t be giving a % chance to win that isn’t close to 50, when it’s almost always skewed heavily in one direction or the other. Funny how after a few wins it suddenly gives you a game where you have a low plat teammate against multiple onyx. That isn’t a coincidence.

So you admit the game is forcing losses if the win rate is that low, good that you’re finally on the same page as everyone else.

That number is randomly generated and nonsensical, it will change dramatically between matches when your team is the same set of people.

You haven’t played much ranked apparently.

And it’s interesting points like this the community should be discussing (with 343’s involvement). Not huge threads on bizarre conspiracy theories like this one.

The important part is how much your MMR changes.

A chunk of CSR lost usually implies that your MMR is already below your CSR (from the games before this one). So it’s frustrating (even more so when you lose it on a 3v4) - but it was probably going to happen anyway.

…That’s literally what we ARE discussing in this thread…

And that the interesting part. The match making system expects you to lose this match.
Think about this. The match making system decides on which server you play (ping), which players are your team mates and which player you play against.
On the same evening, 1 match it expects you to get only 6 kills, but the next match it expects you to get 16 kills.

When does the match making system decîde which match i shouldnt win? Because when its expected that i get only a few kills in this match, i cant help the team to win.

2 Likes

If smurf accounts mean spending time doing bad to play bad… You’re literally forcing the experience on yourself that you blame True Skill 2 for all day long.

You’re spending (wasting?) your time achieving ~<45% win rate while keeping key performance metrics low; KPM and DPM.

Surely negative boosted accounts are being sold for 800ms points like the old days no? I would never waste my time building a match history several hundred games with poor performance just to play against noobs.

1 Like

Just for context of how bad the SBMM is, i got most of my challenges done today, except for the ultimate, and of all the matches i played i won maybe 5 out 25.

Granted i didn’t do so hot on some but still this yoink system was pairing me with people going afk and having a 2 and17 kd. But hey at least they went for that broader audience instead of you know, making a fun matchmaking experience and fixing desync.

1 Like

The MMR would fully self correct in a few games, likely in less than a handful, and the money spent on buying that poor performance account would ultimately be wasted. But, as the saying goes, “a fool and their money are soon parted.”

1 Like

Not necessarily. I can think of valid reasons it could be better to match players against a spread of competition. Some games where they’re favored to win by a bit, some where they’re unfavored by a bit and some where it’s very close to 50/50.

One reason would be to provide more checks and balances to ensure a player belongs where they are ranked. You would not only see performance in “balanced” matches but also against competition a bit below or above this rank.

Seeing how a player performs against competition a bit above their rank would be especially valuable. Simply because the best way to ensure a player can handle the competition at the next rank, and should be moved up to it, is to observe their performance against it.

I am sure someone will come along and say, using arbitrary ranks, beating diamond 3’s with enough consistency shows you deserve diamond 4. Alas, this is a flawed assumption. Beating D3’s says you can beat D3’s. It does not say you can beat D4’s enough to warrant a D4 ranking.

Even if we do assume beating D3’s enough means you can hang with D4’s what happens if you’re moved up to D4 and can’t? You’d end up getting slapped right back down to D3. Conversely, if you must show you can beat D4’s with enough consistency before being moved up it’s based on observed data instead of a flawed assumption.

Yes, this would mean the matches aren’t perfectly balanced all of the time. To this I would respond by pointing out it’s a competitive shooter. Players donning their carebear suit and screaming at the moon about the tiniest inequality seems like a “if you can’t handle the heat get out of the kitchen” moment.

It’s not a conspiracy to suggest pulling the Microsoft Windows Azure Clown Servers: Powered by Linux short straw is going to present… challenges. Lower your odds of winning.

It is entering tin foil hat land to suggest the matchmaker is deliberately placing you into this position until you can prove otherwise though. This applies to win rate, teammate quality, the way the matches are balanced and everything else.

A game here or there where something happens proves absolutely nothing. This is just as likely to be caused by randomness or other factors as deliberate intent. The only definitive way to “prove” the MM is rigged would be to look directly at the system as it’s been applied into the game and be able to understand it.

Since this is probably not a realistic goal the next best option would be to look at a large number of past matches and identify suspicious patterns of repeating behavior. Curiously, the barometer for proof always seems to revolve around isolated, handpicked games and “I said so”.

Yet, over the span of about 50 ranked crossplay games you had a platinum teammate like twice and a gold teammate once, maybe twice.

One would think if this behavior was not coincidental it would pop up more frequently. If it happens rarely and with no apparent rhyme or reason then… it kind of seems like a coincidence.

Have you tried giving performance above the expectations a whirl?

If the game expects a certain game result and performance out of a player, they are ranked based upon this and consistently hit those expectations right on the money it’s reasonable to believe they’ve been placed correctly.

On the other hand, if the player deviated from these expectations it makes sense to reevaluate their placement. Deviating from these expectations with enough consistency suggests improvement or regression.

In short, you help the team win by outperforming the expectations.

1 Like

why would the matchmaker give people in onyx lobbies gold and plat teammates then

since you have all the answers

Would be awesome to capture some data.

We know there are weightings for other things than squads - so it would be great to get an idea of what these could be.

Could there be a weighting for expected latency? That would be awesome.

You could, have course, dropped some MMR. It’s is probably a bit more volatile than I have been expecting. Or early placement. But you said it went down after winning - so a drop of 100 or so points is not going to be explained by that.

The other thing we don’t know a lot about is the possibility of different MMRs for each game mode (eg. Oddball, CTF, etc). It was mentioned in a Halo 5 post that something like this was happening and that you ended up with an average MMR for each playlist.

Could the change in average MMR be explained by that?

It would certainly be possible to drop a bit of MMR if you did very badly. But it would be impossible to play so badly that you drop 100 points average across the team and still win!

Nah. Just a weird UI choice.

1 Like

I think we could discuss this until Infinite Anniversary comes out. But let’s refresh first with some facts.

Diamond 1 is the top 33% percentile. Platinum 1 is the bottom third.
The people who are complaining about this system are almost exclusively coming from diamond/onyx players. These are the people pushing past 2000 games played already.

Now I suspect that the people who are ranking gold and platinum are either 1st time halo players or they are playing on Xbox One which caps you at 1080p 30fps. If you’re playing on xbox one and upgraded to the series X you will see a significant improvement. especially if you have a good monitor. max you can go is 4k 120 fps but 1440p 120 fps is good on a monitor under 30in.

For PC players you can run with a monitor that’s 4K 240 fps

Why am I talking about monitors? As mentioned with an upgrade from xbox one to Series X, you will see an immediate drastic improvement to your game. For any player those extra frame rates could mean you win more of your 1v1 battles and could with time push into the next rank.

It takes the player’s available - organises them into the evenest team combination - and then chooses a map and game mode.

Obviously the jigsaw is a bit harder when you have squads in the puzzle.

But I doubt there is anything nefarious in the selection.

The expected kills is simply your KPM vs that level of opponent multiplied by the time of the match. So a KPM of 1.2 would expect 6 kills in a 5 minute match and 18 in a 15 minute match.

It varies game to game depending on your expected KPM and the length of the match.

I doesn’t decide that you shouldn’t win. It just predicts who is most likely to win. And it uses those values to judge how much rank you should gain or lose on the result.

It’s telling you that if you put in your “average” effort that is how many kills you are expected to get.

You are more than welcome to perform above average. Both in terms of kills and of deaths.

If you are “expected” to go +0 and end up going +5 then you have taken a big step in helping your team over the line!

Use it as a bench mark of your performance. Both past performance and current one going forward.

Yep. I don’t understand the mindset at all. Just sounds like a double waste of time - smurfing your account and then a mere game or two of “dominating”. Your ego would need to be in need of some serious massaging.

Especially because TrueSkill2 starts pushing their MMR up as soon as they go bang on the trigger.

1 Like

Where did you get these numbers?

The population is supposed to fit a normal curve with each division being one standard deviation from the mean.

So Onyx should be the top 2% or so. Diamond the next 13%. And then Platinum 34%. And these three giving your the top 50% above the ideal mean of G6/P1.

There is a tendency to skew to the right, particularly given the open ended nature of the system. And this is one of the reason they favour Season resets - to bring the population back to fit the curve.

1 Like

the problem there is the game is almost dead, and the majority of the people playing ranked consistently are actually good enough to make it to Diamond, making Diamond 1-3 or thereabouts the most common ranks. The numbers are all skewed, with a small fraction of the population being below Diamond, and a very small fraction being in Onyx, but probably well over 2% if you include the large number of people that get to 1500 exactly or just a hair over then never play ranked again.

The ranking system is busted as a result

You’ve got to consider that the people who play ranked at all are a fraction of the already dwindled total population that play Infinite.

1 Like

I get that in different playlists there could be different mmr averages, but it was the 1 playlist we’ve had from launch. Ranked arena. So…although it does have different game variants within ranked arena, wouldn’t it still count as just 1 mmr score for the 1 playlist across all variants since its all contained in 1 playlist? Would be weird to have so many different scores in 1 playlist but i guess not completley out of the realm of possibility. It was fairly late into the season. I just thought it was so weird that it did that. That’s the 1 thing I couldn’t really understand. Have you noticed that when you play with a full team? I of course only see this post match though.