Suggestion; please make quickplay less random

[SEPHi27]

Hello,

Your quickplay system is wearying over time. Until proven otherwise, it’s all based on the randomness.

Tell me if I am wrong, the actual algorithm does something close to :

Query available matchmaking;

with conditions : strict machtmaking (not begun game yet) if possible and with a free slot for your current character

• region lookup is first processing
• the nearest available game has been found
• try to connect

else :
change condition : ask for the player to change character

• repeat the same process

else :
change condition : already begun matchmaking

• repeat the same process

else :
enlarge region range regarding the user’s network settings

• repeat again

else : create and host your own game

Maps you already played previously may not affect the next result.

I think that letting the statistical map distribution to the full randomness itself is not good.

And I feel it bad, because of some facts :

• Each map among the 15 has arround 6.6% equal chances to occur in each trial. p = 1/15.
• Each gets 4‰ (of 1000) chances to occur twice in a row regarding only 2 trials (but read bellow for 10 trials). p = 1/15^2

Let’s play 10 quickplay; (a standard session, made with mixed wins/fails [even though if it does not affect the outcomes, just to say a not so long session])
• Each map gets 14% chances to occur twice spaced out or not over 10 quickplay.

Indeed this experiment follows a binomial trial (aka bernouilli), outcomes are success or fail (“I get a specific map or not, and repeat it…”);

Which the binomial distribution for n = 10 trials, with a probabilty of one occurence : p = 6.6%, gives 14% for at least 2 times occurences expected).

Always in this case :
• Each gets 50% chances to not occur at all.
• Each gets 36% chances to occur once only.

While 10 among 15 is still 66.6% of the map list.

However there are multiple different combos -> actually 3003 arrangements (for Binomial coefficient (15, 10));

Binomial coefficient (15, 10) - Wolfram|Alpha

However again, most of them are made by multiple occurences of the same few maps, spaced out or not.

Let’s go further :

Well, the best scenario we would like to see should be :
getting 10 exclusively different maps among the 15. No double.

That’s here the main point of friction ! For it’s falling to between 3 and 4 of 100,000 chances ! We can say it’s very rare !

For X = x = 10; P = 0.36^10

And chances to get the same map twice in a row (always regarding 10 trials) at least once is:
No need to paste again one more picture with the binomial applet, it’s 1 - (224/225)^10 = 4%
(for info, 1 - (1 - 1/15^2)^10)

Which can leads to a realistic example of a possible random 10 trials sample : 1/ the pit; 2/ skittergate; 3/ the pit; 4/ convocation of decay; 5/ righteous; 6/ righteous; 7/ hunger the dark; 8/ into the nest; 9/ skittergate; 10/ athel yen lui;

Counting : 2x the pit (spaced out); 2x skittergate (spaced out); 2x righteous (in a row); only 7 different maps.

Can be worse or better over time; after repeating it every day it’s becoming heavy for the brain ! We are made like this, we feel it like if we always play the same few maps. I know it’s hard to give diversity with only 15 maps. What I want to tell you it’s not that.

Just please try something else, at least enough to prevent many repeated occurences in a small amount of time (I mean a daily playing session) as much as possible you can ! The previous map(s) played should be taken into account on the next algorithm.

Like someone posted : https://www.daxthink.com/think/2017/2/27/apple-made-itunes-shuffle-less-random-to-make-it-more-random

“Apple Made Their Shuffle Feature Less Random, to Make It More Random”

I know you have lot of work on many things, and it’s not a high priority, but you should give a try on improving it this year, to make the game even better to play, to make the quickplay funnier.

Thanks for reading.

Keep it up.

[Ricordis]

First: Nice work you did there.

I did not do any objective observations but sometimes it feels like there are “this week’s featured maps”.
10 games, 3 times ussingen, 3 times athel yenlui, and 2 times each festering ground/skittergate.
Another week (I can’t play on a daily base) I get more or less the same distribution but with other maps.

If the game would consider already played maps that could have interesting outcomes with the option of not only playing QP games. I’d like to simulate that.
But until then I’d like to blacklist up to 3 maps which would work as a filter like the chosen hero. If no games are found you gotta play a blacklisted map.

[Angerblaze]

@SEPHi27 Correct me if I’m wrong, you are saying that one of the conditions for the “next map determination” should be the actual maps that you’ve already played today? If that’s the case, that’s a good idea. But given the small pool of players and low amount of active sessions that condition would often hit a negative result.

[UnLimiTeD]

First of, nice math.
I am reasonably certain that there are “daily featured maps”, roughly 3 of them, maybe 4, that have a significantly higher chance of happening.
First was the case in closed beta when a new map came out and 50% of the games were on there.
Just looking at the map distribution in the game search suggests as much, as well.
I’ve assumed in the past that the featured maps would hog 50% of the probability, but it’s probably flexible.
More than double the expected average chance, though.
Those featured maps seem to change daily, and maps are more likely if new, or if there have been changes to those maps that fatshark wants to test. This will change up the maps you play over a week if you play only 4-5 games a day, which is probably a reasonable assumption for the majority of the playerbase, but if you play a lot, or binge on the weekends, it notably increases your chance for doubles.

The idea of factoring in your past played maps has come up in the past, and I still believe it is a good one.
There are some challenges with that, though:
A) Quickplay goes after available games. Any game that is already started will have a map that completely ignores your previous games. If a host picks a specific map they want to play, that’s still Quickplay to everyone who doesn’t join through the lobby browser - again, ignoring your past playeyd maps.
B) If you and some other random join two people that have already played 2 games, you now have three logs of maps played, one weighted double, that would best be stored on your computer (but, in the interest of the Anti-Cheat software, few things are stored there, so a solution would likely increase server load).
So what if you all have been playing different maps before? Someone will get a double, soon.
Unlike the example you posted with the shuffle, the “turn off/on” reset happens a lot with Vermintide - Everytime you switch groups.
C) Of course, if no one in the group has the DLC, no one can pick the new maps, and I think they won’t come up if that group goes Quickplay - not sure about that one. Which means that, as soon as someone who has would come in, the chance of the new maps being chosen would be higher. That is oppsed to Fatsharks business interests, is it not?
Ok, ok, now I’m really playing devils advocate. I like the idea, always did. Still, implementation would require some extra thought.

[SEPHi27]

Hello,

First, thank you for the interest you have shown on that post.

I got new informations, from credible sources; some famous modders like bIbIbI(rus) and Grimalackt (indirectly).

Thus, it’s pseudorandom like some of you expected. Indeed, I have been explained that in backstage the algorithm was also querying another lookup table of your own map completion. It means that win rate of each maps will next affect their own occurence frequency. In other words, you are given the maps you have less completed !

If it’s true, that’s for when the last DLC has been released, the amount of the new both maps has been increased.

When you are hosting along with people in your lobby, it’s an average mixing of everyone’s completion.

There are some penalties on the last map played won or not (I have no more informations).

It’s not 100% sure, only devs know exactly how it’s working behind.

But one thing is obvious and shared by many people, we want an improvement !

[Abysswalker]

Some days ago I got the same map 5 times in a row in Quick Play… Also getting only 2-3 maps daily while playing for 5-6 hours is just annoying. The chances are…? So the games randomizing system is really screwed up somewhere.
Nice post tho!