RE: Matchmaking Blog
Just to get this out of the way, because it is the crux of the problem with Dedicated Servers: 150 ping is not a "good" connection. If your stated goal is a "good connection", you need to be aiming for 50, not 150. For reference, from the East Coast of the US, 150 is about what I'd expect in a worst-case-scenario from -Europe bordering on Russia-. In a game where the chase is all about positioning and the tiniest misstep will lose you that hit, or that safe pallet drop, 150 ping is, personally, an unacceptable metric for a "good" connection.
Secondly, an Elo system for an asymmetrical game makes no sense, especially when repeated iterations to the visible rank system have never worked to correctly determine skill, and you even admit so in the post. So the solution is a hidden Elo system for something that you haven't been able to accurately quantify over many iterations? Perhaps it should be considered it's not because the systems you are trying are not working, but because "skill" in this game is not something you can translate into a static numbers system. I love DBD, but it is a casual game with a large amount of RNG and imbalances (maps, offerings, perks, etc) which drastically affect results, in many cases even more than raw player skill, and until those problems are addressed it will always be fruitless to try and make any kind of accurate ranking system.
Lastly, I can't see anything stated that will address the primary cause of bad queues, which is the September change of "highest SWF is your MM rank", and I assume this will extend to "highest Elo is your SWF MM".
"Batch" matchmaking sounds like it will only make things worse, because instead of waiting for the game to find 4 Survivors and 1 Killer, players will have to wait for what, exactly? It to find 50 players? 100 players? That all match even and drop them in all at once? I truly don't see how making the game find more players when it already struggles to find 5 will help queues times. You say at the start of the post that "Matchmaking wait times have exceeded our targeted limits" and yet go on to later say "we hope to improve the fairness factor of matchmaking while also keeping pings and wait times low."
I'm assuming it's just a wording issue, but wait times haven't been low since September, 150 ping is not an acceptable RTT in the majority of situations this game puts you in, and the fairness factor of games has been terrible across the board. Fog Whisperers with 6k hours beat up on rank 12's daily because they have a single red rank friend in their party, and my rank 16 friend is getting bodied by rank 6 Survivors so frequently an outside observer could easily assume there is no matchmaking in Dead By Daylight at all (since anyone at his rank is being forced into MM with their high rank friends leaving low ranks starved of players).
The old rank-average matchmaking was far superior in wait times and fairness, a hidden rank system is not a good idea for a game with such egregious imbalances, and dedicated servers need a better goal than 150 ping (come on, guys, 150 is what I used to get on old shooters playing on a DIAL UP connection in the early 2000's).
Comments
-
I wholeheartedly agree. Not only this, but I feel that making the matchmaking system 'hidden' while not implementing any balance to ranking hurts the game severely for a number of reasons.
1) Many people, myself included, use the ranking emblem system as a way to find areas we need to improve in. For example, if I only get bronze in an emblem, I can see what aspect of that game I was lacking in.
2) This entirely ignores the problem that people had of skilled players not only going against unskilled ones, but treating them poorly. If anything, this will make the toxicity worse because people won't see that the person they've just played against isn't a player with as high of a skill level as them.
A hidden ranking system does nothing of benefit for players. It simply leaves them in the dark and, honestly, feels like the development team have heard our complaints and simply don't care. The game is unbalanced, it's not fun to try and learn to play, and it's just becoming not fun in general.
By the way, I also echo the sentiment that 150 is not a good ping measurement to try and reach. A ping of 150 gets me rubber-banded back to a window I've already vaulted and gets me hit by hatchets through a wall. I understand not all players have good internet, but 150 is not a good ping for the dev team to be aiming for.
0 -
An ELO system would work as long it's not comparing your performance to the killer but to a baseline generated using some algorithm that uses statistics of survivors at a similar rating to yourself.
0 -
What algorithm do you suggest? What algorithm could determine a survivor's 'skill' better than the current system? What's considered a good survivor? How would the algorithm handle, for instance, being face camped, not being able to do generators, add ons like the BNP or keys and maps? How will it handle the fact that we statistically have evidence that survive with friends makes you appear more 'skilled' (that is, makes you win more and perform better in all categories) whereas killers suffer from survive with friends? On the killer side, how will we handle killers who are very clearly balanced differently, i.e. clown vs freddy?
0 -
Having a lot of RNG in the game doesn't necessarily disqualify it from having a functional ranking system. Hearthstone, for example, I've heard is enormously RNG-based, and still has a ranking system that works, because RNG evens itself out over time.
0 -
I'm not sure exactly how it should be done, but imo it would be best if the primary stats used for this rating were during your interactions with the killer. IE. Look at the average amount of time people at a similar rating to you run from the killer for, if you do better than other people then your rating can go up. What I'm thinking of would be a LOT of work as it would require tracking a lot of different interactions of weighting them based on importance to generate an overall score and on top of that you would ideally have different scorings and averages for each killer, so running for a really long time vs say a bubba isn't as much of an accomplishment as running for a long time against a spirit. The system would also need to be able to account for games where you get carried and never really interact with the killer much as well as games where you don't do terribly but just end up getting face camped first hook. IMO camping would give the survivor points since they are still distracting the killer same as they would be if they're getting chased. This means even if you die the system would still give your chase time + 2 minutes of killer diversion via camping.
If you never really end up interacting with the killer (ie. have very little overall score and score is primarily from doing gens, opening doors, leaving,) you won't really go up or down.
In terms of the weighting. Any direct interaction with the killer is extremely important for determining skill so it would be weighted the highest. Healing is very important for team contribution so would have middling weighting. Generator repair, while important, involves little skill and thus has the lowest weighting in terms of the skill rating calculation.
Just to clarify this is just a bit of a brain storm I had and none of this is really concrete and would require a lot of work to not only create the system but also to get it working right.
0
