"Improvements" in ranked?

The “improvements” should be that Ranked playlists are the only playlists with MMR.

At the very least, if there is going to be MMR in every playlist, each playlist should have independent MMR from every other playlist.

There is absolutely no reason why how I perform in Tactical Slayer should affect who I play against in Fiesta. There is no reason why how I perform in Bot Bootcamp should affect who I play against in Big Team Battle. All the different playlists require different play styles and abilities so performing well in one playlist doesn’t mean I will perform well in another playlist.

Plus, if their Trueskill 2.0 (or whatever they are using) is as good as everyone seems to claim, it should very quickly find your MMR in each playlist without having to crossover MMR from other playlists.

7 Likes

No. Social needs some sort of SBMM. Much looser and more team based of course.

But no to random.

I don’t have a problem with shared data between the MMR’s. It makes sense in terms of coding efficiency. And the principle of having a shared core and then offsets for different playlists seemed to work ok in Halo 5.

But yes, they need to function as separate MMR’s.

I’m not sure what has changed for Infinite.

And it’s especially important that the MMR’s stay clean between ranked and social.

TrueSkill2 is good. It can ball park you in a handful of games - but needs closer to 50 to be super comfortable with your rank (for a 4v4 playlist).

So it makes sense to borrow a seeding MMR for new playlists. To speed things up.

And OK. We’ve got the botcamp anomaly. Which is actually quite funny. But who is it hurting? Those handful of players who spend hours in botcamp before entering any other playlist will have an artificially inflated MMR - but unless they can compete in their placement games it will come crashing down very quickly.

3 Likes

I think everyone universally agrees that social MMR and ranked MMR need to be separated.

5 Likes

It isn’t just the Botcamp “anomaly” which implies it’s an outlier. But this is happening everywhere with every playlist. There is another thread showing that you can purposely lose games in any of the non ranked playlists and it will give you easier games in ranked. This is ridiculous.

1 Like

I agree but I also think games like Tactical Slayer are completely different from games like Fiesta. There is no reason that doing well in Fiesta should make me play against more difficult players in Tactical Slayer.

1 Like

I wouldn’t say getting rid of the CSR Onyx levels that it took everyone to Grind out with the trueskill 2 system (love or hate it) isn’t an improvement to me.

They’re going to reset the CSR every season anyway. This season already got extended to 6 months instead of 3 months so it isn’t that big a deal if the CSR gets reset halfway through.

You have to separate two issues.

The first is the MMR being taken over to a new playlist (ie. one you have NEVER played before). This is definitely influential. But pretty much only happens on new accounts. Once you have played and earnt an MMR for that playlist it doesn’t happen again.

And for most people their MMR’s are going to be pretty similar across playlists. So having an MMR that is so out of whack with the others (eg. Botcamp) is definitely an “anomaly”.

The second issue is your MMR in one list possibly influencing your MMR in another. The evidence for this is much less clear. We know that the various playlist MMR’s all share some common data but we don’t know their weighting or use.

For most playlists this probably isn’t a huge issue… but for other’s it could be very important (eg. between ranked and social).

It is very important not to conflate these two issues. The first has a large influence but is not significant (it only happens the first time you enter a new playlist) and the second is a softer influence, but may (and I stress may) have some significance.

The problem is that people are seeing examples such as the Botcamp scenario and then declaring that this is happening to them all the time across all playlists (it isn’t) and that their rank must be wrong (it’s not).

This is a mindset.

With TrueSkill2 you shouldn’t be “grinding” out a rank. You should be just playing and letting the system do it’s work. The only grind should be your quest to get better.

But what we have now is an uneven top end - where Onyx players are “grinding” out wins against lower ranked teams and earning extra CSR. They are essentially grinding for XP points and the more they do it the less it actually reflects on their rank per se.

When you consider that the ceiling for MMR should be around the 1800 mark - yet we are seeing players with high W/L ratios heading towards 2400. They have definitely lost control of the right hand side of the curve.

And it hurts all of us when they start doing it in a toxic manner (eg. smurfing or manipulating team stacks).

1 Like

I agree.

But let’s say your MMR shares data on time played and “form”. And that there are weightings in the algorithm for these factors.

If you are playing a lot of games in Fiesta and doing well… it makes sense to allow for your improvement in Fiesta will carry over into other lists.

It shouldn’t be literal, obviously. You should be able to be Platinum 5 in one list and Diamond 2 in another. But if you have just improved by 10% in one list then the system can improve it’s ranking expectation in the other by the same margin.

Why? It’s social, sometimes you stomp and sometimes you get stomped, it’s all good fun. If you aren’t happy in whatever lobby you’re in, re-queue to spin the wheel again.

Sbmm turns casual lobbies into sweatshops at anything above average skill, giving incentive for skilled players looking for casual games to start smurfing, ruining the experience for the unskilled players who now get to routinely go up against people from the high end of the skill curve instead of constantly going up against general population weighted towards the middle. Everyone loses.

You get essentially nothing for winning against teams with lower rank than you, no matter how hard you carry. Meanwhile, fail to pop off constantly and you will derank hard. The fact that so many people’s ranks aren’t accurately represented exacerbates the problem.

2 Likes

Because that is simply how random selections from normal distributions work.

The only people getting 50:50 (some good, some bad) is the bone average player.

Those at the top end of town are gorging themselves on -yoink- teams - which is why they like it.

SBMM turns all lobbies into equal skills. The same for everyone.

Whether you sweat or chill is entirely up to you.

The problem, I say, is more that many of the “better” players have simply forgotten how to have fun (without going on killing streaks that is).

The thing is that for the majority their ranks are represented. Their CSR just oscillates around it.

The problem is that matchmaking can’t produce enough 50:50 games to keep rank sorting.

It may, for example, run at 5 lower ranked opposing teams for every higher ranked one. So they have to nerf the CSR gains and weight the losses.

For some teams this works and they just take 5 steps forward and 1 step back. They at least have the opportunity to win that 1:5 game and rank up a little.

For other teams it’s still a little too hard. They occasionally lose one of the games they shouldn’t and their rank takes a hit… and they slowly have to work their way back.

And for other teams… they can break free of this cycle. They easily milk their five easy wins and win enough of the 1:5 (often with the help of smurf accounts and squad manipulations) that they can grind their CSR’s a bit higher.

There is no easy fix for this.

Do you cut the CSR gains even more and punish the losses even more? Do you limit which games actually count for rank? (I would). Do you implement a system where you regularly re-normalise everyone’s CSR?

When you think about it the upper limit of CSR should peter out at the 1800 mark. I know that ELO is an open ended system by it’s nature… but it should still reflect the population’s skill as a normal curve. Onyx should be all about jostling for position in the 1500-1800 range… not seeing how high you can go.

Regarding Bots and boot camp or what ever it’s called. My guess the bots will have a low MMR and will not have that much impact on a persons MMR if they play regularly in match made games. I think you will only see an issue on a new account who has only played bot matches.

Why the hell do you think people want SBMM removed from Social? This has been a hotly debated topic since forever.

The only issue with not having the social and ranked ones linked is typically most players will be capable of a certain level of play across all gametypes. Having it in place means that as soon as new playlists are introduced they are a safer place for players of all abilities to enjoy.

The other issue is ranked is better for it. If I’m playing at a fairly high level in social, then putting me in a game with Gold level players is going to absolutely ruin their time. If they weren’t linked, the unranked players would be far less reliably inserted into a game.

It sounds like everyone’s ranks are going to be more accurate. More players should be in Gold/Platinum than are currently. Will no doubt upset many players (“Ranked sucks I was diamond now I’m gold 5”. With these changes hopefully players will be more accurately placed, MMR can be separated more definitively so games should be more consistent and there should be more people at the same rank. Stop people having to play with Onyx players when they’re platinum because there will be more platinum players.

1 Like

I wouldn’t be surprised if the most upset comes from Onyx players… “Ranked sucks I was 2050 but now I’m 1750”.

1 Like