TWG 2016 Ranked Test Thread

[Xiz]

holy shit im bad at math

[choof]

Originally Posted by Xiz

See spreadsheet. I'm working with something that MIGHT work.

It factors in Win % to games played with total points. Hopefully by doing this it prevents "most games wins" and "play 1 game, win with 100% never play again" logic.

I don't think games played should matter too much, if at all tbh, aside from a minimum

[danceflashrevo]

aww yeah red blaster GOAT

[Xiz]

Someone who knows how to math should help me lelelelel

[Red Blaster]

Originally Posted by Xiz

Someone who knows how to math should help me lelelelel

YoshL is a spreadsheet wizard, he can probably figure it out

[Funnygurl555]

not the only thing wrong with your spread, xiz

i need like way more points

[Xiz]

Originally Posted by Funnygurl555

not the only thing wrong with your spread, xiz

Fixed. You should see an appropriate adjusted point system for yourself.

[Xiz]

Originally Posted by thesunfan

EDIT: I think to avoid players making the game by playing a lot, you would have to make win % a significant factor, while also decreasing the significance of signing up for games/replacing into games.

I feel as though replacing in should still be given 1 point minimum, to help encourage replacements. So instead perhaps increasing win points to compensate?

[Red Blaster]

Originally Posted by Xiz

I feel as though replacing in should still be given 1 point minimum, to help encourage replacements. So instead perhaps increasing win points to compensate?

What about if they get 1 point for signing up as a replacement that they get regardless of if they get into the game or not, but if they actually get called on to sub in they forego that point and get whatever points they would normally get based on how the game goes?

[Xiz]

Originally Posted by Red Blaster

What about if they get 1 point for signing up as a replacement that they get regardless of if they get into the game or not, but if they actually get called on to sub in they forego that point and get whatever points they would normally get based on how the game goes?

If I miss a game then i'd just wait to be like 3rd or 4th sub so I know I most likely wont get in and keep that point. I think that could be abused.

[DaBackpack]

Originally Posted by Xiz

Someone who knows how to math should help me lelelelel

I have suggestions when I get back to a computer

[DaBackpack]

so a few things

I'm viewing this under the lens of somebody who will try to game the system to maximize points

1) The (mostly, given the meta of this site) optimal strategy for signing up is to sign up as a replacement --- given the proclivity for many members on this site to replace out, signing up as a replacement gives you a very good chance of having a raw point advantage over those who legitimately signed up (2 pts and 6 pts vs 1 pt and 5 pts). I suppose it's a tradeoff in that there's a chance that replacements will see 0 points, but it might shift signup behaviors because the expected value of signing up as a replacement (estimating here) is higher than signing up as a regular member.

2) The -5 penalty for being replaced incentivizes inactivity. When a person requests a replacement, a lot of the time the member realizes that they will not be able to commit to the game as they should. But there (currently) is no penalty for being inactive and making a handful of junk posts per game. The best option for "somebody who can't play" is to sign up and post minimally in order to avoid the -5 penalty. (So long as they don't go into modkill territory.)

3) I considered the notion that it is harder to win as some factions vs other factions (say SK vs town). I thought that the serial killer or mafia should be rewarded even more for winning considering how harder it is. Imagine a scoring system that scales a base value according to the number of players that also lost during that game. So a serial killer winning gets him more points because he bested everybody else, vs town who bested only 4 people. But then I realized that the current system actually accounts for this: the serial killer winning only introduces 5 + (number of other players) points into the system, vs a town win which introduces (5 * number of town players) + (everyone else) points into the system ---- so the serial killer winning means that everyone else is denied points, so it balances out (mostly).

Again this logic is predicated on the notion that certain people would try to game the system in order to maximize points. I think this community as a whole is trustworthy and respectful, so I don't think these concerns are worth fussing over. But they might be interesting from a game theory perspective.

[DaBackpack]

Originally Posted by choof

I don't think games played should matter too much, if at all tbh, aside from a minimum

A suggestion: do not factor win rate into point values at all. It's dangerous to mix the two measures imo.

Instead, just change the "all-star game" criterion to:

The top 13 ranked players that have above a 50% win rate will enter the game.

Keep raw point values but also track win-rate separately.

EDIT: The limitation with this is that you must manually specify a "win-rate" threshold. In your proposed method of having the win-rate affect the scores, it bypasses this measure by implicitly comparing player vs player winrates instead of player vs manual threshold winrates.

EDIT EDIT: The "manually-specified threshold" could just be the average winrate of each player in the rankings, or something. "You need a lot of points, but you need to win more often than average in order to win." Or something like that.

[_Zenith_]

Originally Posted by Xiz

If I miss a game then i'd just wait to be like 3rd or 4th sub so I know I most likely wont get in and keep that point. I think that could be abused.

It would be abused, but I do think being a replacement is worth one point.

Make it so signing up and participating as a replacement awards one point, and other point variables will be accounted like winning, being modkilled, etc on top of the point.

[Xiz]

Originally Posted by DaBackpack

A suggestion: do not factor win rate into point values at all. It's dangerous to mix the two measures imo.

Instead, just change the "all-star game" criterion to:

The top 13 ranked players that have above a 50% win rate will enter the game.


EDIT EDIT: The "manually-specified threshold" could just be the average winrate of each player in the rankings, or something. "You need a lot of points, but you need to win more often than average in order to win." Or something like that.

I really really like this

[Xiz]

I'd say more like 40% or something. Also because as of now, there are not even 13 players with above a 50% atm hahah.

(Because it's so early in the year and literally 2 out of 5 games have had SK wins, those have fucked everyone else)

[Wineandbread]

#DIV/0! win rate best win rate

[DaBackpack]

Originally Posted by Xiz

I'd say more like 40% or something. Also because as of now, there are not even 13 players with above a 50% atm hahah.

(Because it's so early in the year and literally 2 out of 5 games have had SK wins, those have fucked everyone else)

yeah the specific number itself doesn't matter all too much

just that the entry conditions for the supergame should account for both participation (which is included in the point value, alongside some notion of success) and success (which a win-rate captures)

the difficulty with specifying a win-rate manually is that we have no idea what the final win-rates will look like in the end. And we can't manually decide the win-rate at the end of the year (because that allows for bias & possible cheating)

might do some math to back this up later, but I think the win-rate threshold should depend on the average win rate of everyone in the tally (or some top % of participants) --- it covers the problem of the SK winnings as well. (If the serial killer wins a lot, then the win-rate required to enter the game drops because more people lost routinely, for example.)

[Makilaz]

does it count as a win points-wise if you were on the winning team but didn't survive until the end of the game

[DaBackpack]

Originally Posted by Makilaz

does it count as a win points-wise if you were on the winning team but didn't survive until the end of the game

yeah pretty sure