Mafia Games ELO thread

[Gypyx]

UNOFFICIAL ELO SHEET HERE (in the sense that it's not endorsed by the mafiascum mods in any way, this is my personal fun project)

to continue the discussion that was started in moderation discussion thread

i was thinking of making an ELO spreadsheet for mafia games

obviously it wouldn't be on every game, the way i was thinking of doing it is that if at least half of the players + the game mod agrees to it, without any objections, this game will count towards the ELO spreadsheet once game ends, bastard games being forbidden from making it onto the sheet, but otherwise any game can count

as for the actual math stuff, I was thinking of implementing it like this

take the average ELO of the winning team

the average ELO of the losing team

calulate it like a regular 1v1 system afterwards

[Gypyx]

obviously if people have better ideas for implementation i'm all ears, this is just my personal preference

[humaneatingmonkey]

would it potentially help me find the mafia team?

[Gypyx]

i sure hope not

[Black]

Ego

I'm working but I'll check this out later!

[Ydrasse]

this is a rough framework that i think may be more fulfilling than one point but not terribly sweaty

- rather than points, pick a set of setups that you want to run for this ranked system. 5-10, imo semi open or open are the best, or something to avoid swingier setups doing bad Things or closed ones etc

- determine what the average % win rate is for each team and weigh points (out of like 100) for each side accordingly to account for ev and how they tend to play out. adjust as needed

- run it blitz style to not make it sludge that people grow disinterested in because point accumulation is taking too long

this avoids the tiebreakers that may come with everyone just getting one point over and over and allow a slightly truer ranking of ability without it being terribly hard to figure out or sweat over since the core is still to win as your alignment versus optimizing the minutia of whatever role you receive

[Gypyx]

i don't think i agree with it necessarily being open, like, i guess i have a different view of competitiveness, but this ELO thing is not about pure skill expression, it's just a fun project for everyone to showcase their stats with

[Black]

Yeah without actually reading the first post, I don't think this needs to be some super specific thing to track who the best players are. Ranked team games rarely work like that. Using League as an example, sometimes you just lose because of your teammates, and that's fine because that expectation is already there

[Gypyx]

my only worry with the system i have implemented is that everyone's points would be weighted towards the bottom naturally due to scum winning about 50% of the time but not being 50% of the players

a fix to that would be (i think) to instead split the point gains between everyone? basically to continue with my 1v1 system, a 10:3 game that ends in a town win, with the town gaining 100 points would have each townie gain 10 points and each mafioso losing 33 points

[humaneatingmonkey]

ngl i just want to track an official w/l record on any alignments using quality, standard games

[Gypyx]

there's the normal games archive although that place could use some work by getting put into a format better for data analysis (i wanted to do that but i'm bad at comitting to stuff)

[humaneatingmonkey]

but without replacements!

[Gypyx]

Kids these days taking everything for granted

[Psyche]

if interested, i could implement a retrospective calculation using the archives given certain parameters. have already done a lot of data cleaning in that space.

making something people use would be a nice change of pace for me

[Psyche]

but like, let's think through what we really want first

this could be a thing like the SUPP where site culture overwhelms all the things that should be bad about doing this, but taking criticism into account might improve the final product

[Gypyx]

I guess it could be used for test runs to see if the numbers we reach using whatever system we choose look right ?

Kinda uneasy with the idea of implementing the "official" ELO retroactively but maybe people shouldn't care

pedit : I think the way we want this to work is mostly as something that's kinda personal / to encourage personal improvment? Like "let's go i reached 1500" instead of "cool go i won another game" which feels way less satisfying

at least i think it could make playing mafia feel way more meaningful

[ketchup777]

this seems cool
i feel available points should be staggered as well the longer you survive
say if a townie is NKed night 1 having them lose as much Elo as the people who miseliminated in ELO (oh no two ELOs thisll be confusing) feels wrong, as does rewarding a townie who gets limmed day 1 compared to someone who survives 5 days and wins, for instance

[Gypyx]

this seems cool
i feel available points should be staggered as well the longer you survive
say if a townie is NKed night 1 having them lose as much Elo as the people who miseliminated in ELO (oh no two ELOs thisll be confusing) feels wrong, as does rewarding a townie who gets limmed day 1 compared to someone who survives 5 days and wins, for instance

i agree but also i think i want to keep the system very basic to emphasize it's not really a thing meant to be competitive, more like, a funny statistic for players to enjoy

you could also argue that it punishes strong town players who make themselves extremely big targets for scum to protect PRs for instance, like, truth is that i think that having as few rules as possible is best

[ketchup777]

this seems cool
i feel available points should be staggered as well the longer you survive
say if a townie is NKed night 1 having them lose as much Elo as the people who miseliminated in ELO (oh no two ELOs thisll be confusing) feels wrong, as does rewarding a townie who gets limmed day 1 compared to someone who survives 5 days and wins, for instance

i agree but also i think i want to keep the system very basic to emphasize it's not really a thing meant to be competitive, more like, a funny statistic for players to enjoy

you could also argue that it punishes strong town players who make themselves extremely big targets for scum to protect PRs for instance, like, truth is that i think that having as few rules as possible is best

true that

[Psyche]

i wonder if preparing and presenting the data in a more raw form is the way to tackle the fear of hyperfixation on scores
or maybe actually presenting many measures of performance instead of just one?
i guess either way a data table is going to happen so will make that the first step

[Gypyx]

boutta win the title for longest time spent alive

[Ircher]

I think you should split the ratings based on alignment. Being good as town doesn't tend to equate being good as scum and vice-versa.

[KayJayQueue]

I think you should split the ratings based on alignment. Being good as town doesn't tend to equate being good as scum and vice-versa.

Can confirm.

[Black]

I think you should split the ratings based on alignment. Being good as town doesn't tend to equate being good as scum and vice-versa.

I like this

[usesPython]

there's the normal games archive although that place could use some work by getting put into a format better for data analysis (i wanted to do that but i'm bad at comitting to stuff)

If anyone does this for a large enough dataset make sure to also grab when the game was played since that allows implementing confidence decay (where the longer you go without playing the less confident the system is in your rating)