If you feel like MMR system is not good, how would you change it?
I just feel like people are pointing out a lot of errors on the mmr system, but at the same time i don't really feel like there's even any better skill calculator that don't be like "escape=mmr up" "kill=mmr up".
Don't get me wrong, im not saying escaping or getting 4k's means your good, neither saying mmr system is good. Im just saying people are like "this is wrong" and not much of a "it could be better if it was like this". How would you change, and how could it be better for you?
Comments
-
This is a party game.
It doesn't need MMR.
Post edited by Pulsar on6 -
The perfect method would be for MMR to be solely dependent upon how long you can loop a killer/how soon can you down a survivor. Obviously, this would involve incredibly complicated math that had to take map, RNG, pallet use, Killer, addon, perk, SWF etc into account. You would need to calculate optimal chase time and balance teams around that so that every player has a roughly 50% chance of running the Killer long enough that they can all get out/down survivors fast enough to 4k.
But considering how catastrophically bad the current system is, I say just bring back the emblem system and make ranking up significantly more difficult than it used to be. Also, just get rid of monthly rank reset and perk tiers while at it since they are bad relics of the past.
1 -
I prefer this to what it was before by a country mile, but it needs improvement.
I'd use the emblem system and match people up with others who score similarily in each individual category (e.g.: people who have an average of 4,000-4,999 BP in Survival and Altruism are put in matches together). The killer emblems roughly align, so they can do the same (e.g.: a killer with 5,000-5,999 BP in Devout is matched with survivors with 5,000-5,999 BP in Unbroken).
The aim would be to match people on average, based on performance. All things are considered, so kills or escapes are still important, but so too is altruism, defending gens and so on. I'd say this would accurately be called SBMM, even though it would be a complex system to put in place.
The alternative is stop insisting this is SBMM, and just say it's a basic MMR. Doesn't solve the problem, but does remove the fogscreen of "skill".
2 -
You need more stats to be considered in the MMR up/down. Things like safe unhooks, heals, chase time, escapes, gens done, and other things that affect your team. Being a "good" all around contributor should count for more than stealthing around and escaping because the other three members of your "team" did everything. Stealth should count though for a portion. Things like staying in the TR without being in a chase.
Killers... kills pretty much sums it up. Although you could say chase time (short as possible), actual hooks and not just hook states, time for generator completion, anything else that would be "Killer did their job". Kills though feels pretty good. 2k2e should be 0, no change. Anything above/below that up/down. Although 2k2e + 4 hooks... have a good argument that maybe they should go up a bit.
1 -
I have to ask- are you implying the game should just toss together completely random players that just happen to share a region server...?
1 -
Revert it back to ranks and improve the emblem system
7 -
That's what it does currently, so no.
2 -
So, what alternative is it that you're proposing?
0 -
I'm not a Dev, nor am I arrogant enough to say I know what works.
I do know what doesn't work though and it's the current system.
2 -
Use the emblem system. Easy fix.
2 -
Primarily? Take into account team results over individual ones.
If we were completely revisiting it, start with the question of "what makes for a fun game?" Kills and escapes are nice, but what makes the game actually exciting for people to play? Don't just go straight to the emblems, but the fact there are so many scoring events is a nice starting point. Is a fun game one where both sides are scoring lots of BP consistently over the course of a match?
People are going to work out and game your MMR system either way, you might as well use it as a tool to encourage fun gameplay.
2 -
Revert it back to the emblem system or its old predecessor.
3 -
It's fine, I like grading more because I don't need to sweat to get bonus BP.
0 -
I think MMR is working ok for the most part in that if you have a match where all five players are about the same MMR it will probably be a close match.
The devs have already commented on how a lot of mismatches occur because of people leaving lobbies and people with outlier ratings causing the system to be in a position where it has to choose between having people wait a long time for a closely matched opponent or putting them in whatever slot it can find at the moment. That’s not a problem with the MMR being inaccurate but rather an issue with finding the right balances between waiting for a “close” match and wanting a “quick” match.
I suspect where the MMR could possibly use some tuning is:
- New player initial ratings might be or have been too high. The devs already indicated this is an issue in their update report last month so presumably they’re tuning down the initial rating for brand new players.
- There may be an inaccuracy when it comes to how the system calculates what the optimal MMR for a killer should be given a set of four survivor ratings that differ by a wide amount, like when a swf has a high skill player paired with a low skill friend for instance. I could easily be wrong but I suspect the system is just doing something simple like taking the average of the ratings and aiming for that, but in reality a more accurate estimate is probably a non-linear formula so a better way to estimate it would be something like a neural net trained to input four survivor ratings and output a killer rating that it thinks produces close matches. (The devs haven’t ever gone into detail on this part though so I’m just speculating. 🤷♂️)
1 -
MMR isn´t the issue.
Balancing is the issue.
The SBMMR just made the balance issues visible to more people.
0 -
This game will never be balanced with these devs
Just revert the system and let's go back to emblems
1 -
What would be achieved by that? More Rainbow matches like before the SBMMR?
1 -
Delete the code
3 -
Partygames are played with friends. Thus, SWF should have to bring their own killer.
0 -
Blops 2 had party games online.
Lord knows I did not make friends in Sticks and Stones.
1 -
Killers want easy 4ks. Going back to the emblem system will give them the 100 4k streaks they've been addicted to for years. Getting a guaranteed level 20 in every match with the old emblem system guaranteed 1k in the first 30 seconds.
I really think Killers would be happier with 4 AI survivor bots that somewhat played the game than having a challenge. Because lets face it, every argument about the MMR system is that it's "too hard" for killers and that killers should be the "power role". Basically meaning killers should have easy 4ks.
1 -
The game is not to hard. The game is simply unbalanced.
If it were to hard, then Streamers with thousands of hours like Otz would still manage to get their 50 win streak challenges. But they don´t. So its not a skill or being to hard issue.
0 -
No.
I just want to stop getting teammates who have 15 hours. Me and my friends all have over 4K hours. There is zero reason default Meg should be here.
2 -
The old emblem system had its failings but it did kind of take contribution into account.
Surely its possible to rank up and down based on escape/kills but adjust the total according to contribution.
The value shifts up and down based on escape/death but is then bracketed in how much it shifts by your emblem outcome.
Its possible to die and lose but contribute a lot so you only go down slightly, whereas its possible to do nothing all game and escape and then you only rank up slightly.
0 -
But that's exactly what you would get with the emblem system. The new MMR system is much less likely to get that. Most survivors in my games are pretty close to my skill level. Every once in a while there's one that is better or worse, but for the most part, they're pretty close.
Also, when I play killer, I can tell the survivors I'm going against are pretty close in level too. They all loop about the same.
0 -
I think the SBMMR system is mostly fine, except for one small issue: MMR is evaluated individually, as in via four simultaneous 1v1s, as opposed to as a 4v1.
If more than 2 survivors die, the Killer should gain MMR and all the survivors should lose MMR. If less than 2 survivors die, the Killer should lose MMR and all the survivors should gain MMR. If exactly 2 survivors die, neither side should gain/lose MMR.
0 -
Yep, that's my point. It's the same.
Also, it's not much less likely. We get sub-200 hour teammates constantly. We also get sub-100 hour Killers regularly too but occasionally we get a monster. We got a certain Blight main a while back with a Jake who had Iron Will 1 and 15 whole hours.
1 -
I ran into a really funny clip of a dev Q&A about a year ago when they were still ironing out MMR, and Patrick outright said that having it only look at kills and escapes was horrible and they knew it was bad. Oh, the irony.
I want it to focus more on chases both on the survivor and killer side. Kills and escapes are useful data to factor in, but look at chase times for both survivor and killer, and number of chases completed on the killer side. For survivors, perhaps even look at plays like flashlight/pallet saves. Detract points from every survivor who wasn't being chased/hooked/downed if someone progresses a hook stage and the killer wasn't close enough to trigger the proximity emblem event. The old emblem system's failure was that it put too much stock in gens and heals and not enough for actual skilled gameplay on the survivor side, which in turn told the game that survivors were more skilled than they were - but it was better than our current SBMM. And it rewarded killers more for 8+ hook games and less for camp-outs and 'everyone dies on their first hook' type scenarios, so players who did stuff like basement Bubba or couldn't get kills without NOED didn't shoot up way past their skill level.
1 -
I would have removed the SBMM and improved the ''old'' ranking system. It would make the game so much more fun in my opinion.
4 -
No it's not. I keep getting 6-80 hour Megheads despite having 1.2k hours in the game. And yes, I know time =/= skill, but when I'm forced to watch the Claudette crouch on the other end of the map while my Kindred is clearly showing Blight zooming to down the Dwight in 10 seconds from full health, it's not the exception. It's the rule. With the emblem system I had a 50-50% chance of having a decent game. Now it's 9 4ks for the killer at 3 gens remaining and then 1 match of 2 hooks and we are all out. Rinse and repeat. The balance is significantly worse than it was before.
Edited for typo.
2 -
I'm sorry but this is just a poor argument. Players on both sides will call for easy kills and easy escapes, trying to demonize one half of the player base is unhelpful.
I think the key problem with SBMM from the casual perspective is that it does a pretty good job of matching competent survivors together, you watch even some of those great streamers and they are often exploiting the weak survivor to gain a foothold.
That is becoming less of scenario so games eventually get sweatier and sweatier, the viability of chases generally goes down and you get a increase in playstyles that people don't like camping/tunneling/slugging etc because they become more efficient/rewarding than the typical spread damage and hooks playstyle.
There's this real culture in DBD of while its ok for survivor players to split up on gens and smash them out asap for maximum pressure, its not ok for a killer player to remove someone early to gain the same kind of pressure. Yet that's exactly what you should be focusing on if you want to secure a 4k.
Just as there is nothing wrong with people smashing out gens there is nothing wrong with people wanting the 4k, its their goal after all.
0 -
I think the MMR system is...mostly okay. The big issue is that there seems to be a deficit of killers, and that the system doesn't handle lobby dodges well - leading to problems.
To the lack of killers problem, an easy fix would be a 'Call To Arms' system that offers double BP to the side that needs more players at a certain time of day, until queues for the other side hit a reasonable level.
To the backfill problem, BHVR said they are working on it.
1 -
The mainproblem is not which kind of mmr is used, the old one or the new one. The main problem is that queue times are prioritized, meaning at some point mmr is thrown overboard anyway. Thats when the 4k hour team gets the 15 hour teammate.
I had taken a break from the game just before sbmmr came into the game. But my experience is this: With the old mmr, i had a survival rate of about 75%, and potatoes in a lot of my games.
Nowadays, my would guess overall survival rate is higher (this is actually just a guess. I used to keep stats of my games, but havent done so since i came back), but even though i see potatoes once in a while, i am mostly matched with people actually contributing, leading to way less frustrating games, win or lose, because even the games i lose have most gens done, all 5 a lot of the time. Since i came back, i never had a game with no gen done at all, which happend under the old system.
0 -
Widen the value range
Raise the softcap
Restricted que time override
Remove averaging for team scores
Score adjustments need to have factors from the " nuances and context" of actual gameplay in combination with the results.
In any other game if both sides "win" its called a draw, so the complicated math = an adjustment of 0
0 -
Remove it.
Making proper MMR system is not easy, and if this took them so much time then I can assume they are not willing to put enough time into making a proper MMR system. So just remove it.
2 -
Remove both MMR and the Emblem system. Keep bloodpoints and pips
Boom, PARTY GAME fixed.
0 -
just take hooks into account, Survivor escape without getting hooked, more MMR gain
Killer kills survivor with all hooks more MMR gain
0 -
Killer being based on kills is mostly ok. That's generally going to correlate with skill but it becomes problematic on certain killers like Leatherface.
Survivor needs way more nuance to accurately reflect skill. If you're just taking the result and not comparing all of the players in that match against that killer, you're missing the forest for the trees. It's very apparent that the player who ran the killer for 2 minutes is better than the teammates who somehow went down instantly to that same killer. But MMR sees that they all died. I'd even argue that the survivor who ran that killer for 2 minutes is better than that killer and that they died because of their teammates, not because the killer was better.
Survivors often die to vastly inferior killers because of super weak links on their team. You don't have much of a system if you can't account for that. Especially if you're just going to throw out SBMM as soon as peak hours hit and throw 50 hr players with 3000 hr players. The system we have now is somehow worse at matching players than old ranks were. That's wild.
0
