winrates9

Sheffrey5mon 1d

i am referring to overall game/role winrates (e.g. in Sandbox, Detective has a 55% winrate*) not individual winrates users have on certain roles

Right now the only winrates that are visible on this site are based on alignment, and we can only see these winrates on certain setups. Granted, the most popular setup is still 12p closed random roles here in the box of sand, which to my best understanding is fairly townsided overall.

Would having separate role winrates be useful? Would the randomness factor of typical sandbox setups make this a fruitless endeavor? There's also the issue of roles that convert that affect winrates; e.g. you start the game as Cop but eventually get Culted and win as Cult, should Cop's winrate go up in a situation like this, or perhaps adjusted or even not taken into consideration?

Among those questions listed above I would love to hear your thoughts/ideas on role winrates.

*not a real statistic

350visits
0viewing
6
nepentheAug 19, 2020
+1
afaik winrates are only calculated based on ranked games, so they would be completely meaningless in sandbox. but even if lucid decided to add unranked winrates, roles' value still changes from setup to setup, so a sitewide or even lobbywide role winrate would still be meaningless - maybe having them separated out in the setups themselves could be nice idk
SheffreyAug 19, 2020
nepenthe
afaik winrates are only calculated based on ranked games, so they would be completely meaningless in sandbox. but even if lucid decided to add unranked winrates, roles' value still changes from setup to setup, so a sitewide or even lobbywide role winrate would still be meaningless - maybe having them separated out in the setups themselves could be nice idk
Are there not roles that are strong in most setups? A few that come to mind are Surgeon (saves kills and stops conversions), Detective, Demolitionist (puts the game on a timer for maf if demo isn't killed early on).
banimeAug 19, 2020
I think as you mentioned there's are too much randomness in Sandbox to make it useful as a measure of a role's difficulty. Though maybe still interesting to see if you considered the variables:

- There's the conversion issue you mentioned. How would admirer's winrate work?
- Certain roles would have better or worse odds depending on the of the setups they are more typically found in. For example, deputies may get an unusually high winrate if they often win in shoot-em out games.
- Does sui invalidate wins? If so, some less liked roles wouldn't see as many wins.
- The number of players in a game also make a role's winrate less meaningful if there are many (smaller effect on the outcome) than they would in a 6 person game.
SheffreyAug 20, 2020
If the setup managed to remain consistent (e.g. only tracking role wins in 12p closed roles standard sandbox setup), would it help?

To my understanding a role's power doesn't increase when going up in the number of players, only down (as you mentioned in 6p games), so would that give us a good enough idea?
spaceAug 20, 2020
+1
idk if that helps the model much bc the truth is the casual approach of sandbox means any stats you pull from it won't account for any effort put in by the users playing as those roles. there are plenty of games where people don't necessarily play to win (which is perfectly valid!) but that kind of participation is regular enough where it could screw with your stats. and like others have said, how do we account for losses due to suis/sui races or vegs? i feel like the inherent nature of sandbox would create a margin of error too extreme to account for how likely a person is to win as a certain role even if they play to that roles strengths.
spaceAug 20, 2020!
+1
because of closed role setups, games don't follow much strategy at all outside of mylo/lylo situations.

also gingerbreadman's stats would be crazy because that's one of those roles like tree that you can coast through
asyncAug 20, 2020
I could calculate these if you're interested but it would only be for games within the last ~2 months since lucid deletes games after 28 days and i only started scraping the site a month or so ago
SheffreyAug 22, 2020
we need an exact % winrate of voodoo words by word

how powerful is the c word


what is the true voodoo meta

these are the questions that lucid refuses to answer.
KalvonAug 22, 2020
+1
all i know is out of the millions of games, siren has won maybe like 5 times