Page 3 of 9

PostPosted: Thu Mar 09, 2017 8:40 pm
by BROseidon
Thankyou.gif

PostPosted: Thu Mar 09, 2017 8:44 pm
by zoraster
I know it's both, but is the goal of the game to win or to finish the highest average position?

If we're talking about encouraging risk-taking behavior, the latter is not the way to achieve that.

PostPosted: Thu Mar 09, 2017 8:46 pm
by Vijarada
Yeah that discussion has been going on for years.

PostPosted: Thu Mar 09, 2017 8:50 pm
by zoraster
In the context of "oh no the meta is wimpy. i quit!"?

PostPosted: Thu Mar 09, 2017 8:52 pm
by BROseidon
Not specifically, but this has been discussed.

PostPosted: Thu Mar 09, 2017 8:54 pm
by xofelf
I had the highest score on the board by a long shot, but I was on the inactive list. But it works with recent games more, yes. It isn't an average of all the times you played. But for my game history, the 25th is actually ridiculously low in comparison. I think the other lowest place I've ever had was 11th? I've had ever placement from 1st to 7th I believe. It's rarer that I miss jury than it is that I make it. So I'm not super surprised this tanked me.

PostPosted: Thu Mar 09, 2017 9:06 pm
by hiplop
In post 51, zoraster wrote:I know it's both, but is the goal of the game to win or to finish the highest average position?

If we're talking about encouraging risk-taking behavior, the latter is not the way to achieve that.

ELO is kind of a joke that we all know doesnt mean anything but kind of as a community embrace despite its stupidity. I don't think anyone plays for ELO

for example I was top of the ELO for like a week and I suck at this game

PostPosted: Thu Mar 09, 2017 9:11 pm
by zoraster
I mean, I doubt anyone purely plays for ELO, but the method people are measured against each other tends to have subtle and often unconscious effects. I actually think ELO is a pretty good system overall. But if the concerns of some people is that there's not enough risk-taking, well, this system at the very least doesn't help with that.

I don't have enough knowledge to know whether or not that's actually a problem and not just people finding something new to complain about, but...

PostPosted: Thu Mar 09, 2017 9:30 pm
by xRECKONERx
In post 46, BROseidon wrote:It heavily biases towards recent performance.

is this true? I thought elo basically took everything into affect equally

PostPosted: Thu Mar 09, 2017 10:14 pm
by zoraster
it tends to be a feature of elo ratings. Which is generally a good thing: the goal is to show current skill. https://elliotnoma.wordpress.com/2015/0 ... o-ratings/

PostPosted: Thu Mar 09, 2017 10:15 pm
by BROseidon
In post 58, xRECKONERx wrote:
In post 46, BROseidon wrote:It heavily biases towards recent performance.

is this true? I thought elo basically took everything into affect equally


IIRC it weights more recent performances more heavily.

PostPosted: Thu Mar 09, 2017 10:21 pm
by hiplop
bros right

PostPosted: Thu Mar 09, 2017 10:24 pm
by zoraster
One of the curious things about doing this is that I assume (maybe I'm wrong) that your elo is adjusted as a one v. one competition with each person in the game. So for example, xofelf beat kdowns and Annadog40, but because she had a 1557 rating vs. a 1085 and 1200 rating, respectively, she didn't get many points for that. But THEN she lost to everyone else, including 986 me and 952 Trollie, both of which on their own would decrease the Elo score considerably. And that's why her Elo went way, way down.

So, and CC can correct me, rather than think about each game as a single iteration of Elo, think of each game as X number of games scored all at once, where X is the number of players in the game minus one.

But, depending on whether CC readjusts the elos EVERY time, it could be worse than that. Because xofelf would face every single person as a 1557 elo, whereas if she did it as 27 single games, she'd readjust her elo and it'd suffer less each time she lost.

PostPosted: Fri Mar 10, 2017 3:07 am
by CuddlyCaucasian
You're right zor, it's calculated as (in the case of ned's) 26 1v1 games against everyone else in the game. I think it does skew too heavily towards recent performance, although I could make it more stable by lowering the k-values used for everyone. It's nice to have a dynamic ranking system, but obviously none of them will be anywhere near perfect in something that's so subjective like Survivor and other large social games.

PostPosted: Fri Mar 10, 2017 9:57 am
by zoraster
In post 63, CuddlyCaucasian wrote:You're right zor, it's calculated as (in the case of ned's) 26 1v1 games against everyone else in the game. I think it does skew too heavily towards recent performance, although I could make it more stable by lowering the k-values used for everyone. It's nice to have a dynamic ranking system, but obviously none of them will be anywhere near perfect in something that's so subjective like Survivor and other large social games.



Is it calculated using static ratings or dynamic ones? In other words, was each of my games judges as if I had a 986 or for each victory did mine inch up a bit until I went out and then every loss incrementally reduced my elo?

PostPosted: Fri Mar 10, 2017 10:18 am
by mv_gimmick
It's caculated using static, similar to a chess/go tournament. Ratings don't change until after you have the performance of one full event.

PostPosted: Sat Mar 11, 2017 3:53 pm
by Malkon05
In post 56, hiplop wrote:
In post 51, zoraster wrote:I know it's both, but is the goal of the game to win or to finish the highest average position?

If we're talking about encouraging risk-taking behavior, the latter is not the way to achieve that.

ELO is kind of a joke that we all know doesnt mean anything but kind of as a community embrace despite its stupidity. I don't think anyone plays for ELO

for example I was top of the ELO for like a week and I suck at this game



HEEEEEY I'm pretty proud of my 7th place :?

PostPosted: Sat Mar 11, 2017 7:56 pm
by BROseidon
Elo is a measure of how far on average a given player gets. Getting far and winning are two different things though so yeah.

PostPosted: Sat Mar 11, 2017 8:34 pm
by xRECKONERx
Yeah I think "as long as I get to FTC I will be happy!" or "as long as I make jury I will be happy!" are pretty toxic views on the game tbh and that has a lot to do with elo

perhaps if there was a way to scale elo to match the Survivor payouts?

Winner: $1,000,000
Runner-up, $100,000
$85,000
$70,000
$55,000
$45,000
$35,000
$27,500
$20,000
$15,000
$10,000
$7,500
$5,500
$4,500
$3,500
$2,500

that's what i found on a quick googling... so like, do that, but scale it based on # of players and keep track of "winnings" instead of elo?

PostPosted: Sat Mar 11, 2017 8:34 pm
by xRECKONERx
look @ me having great ideas

PostPosted: Sat Mar 11, 2017 8:36 pm
by xofelf
That sounds awesome actually.

PostPosted: Sat Mar 11, 2017 8:37 pm
by xRECKONERx
like okay, winner gets a mil, runner up gets 100k

then you total up the remaining pool (~386k) and space it out over the rest of the contestants

then winning has a much higher value to this ranking system everyone pretends to not care about but secretly does

PostPosted: Sat Mar 11, 2017 8:39 pm
by Cheery Dog
How do we compare when compared to people that have played double the amount of games though?

PostPosted: Sat Mar 11, 2017 8:39 pm
by Cheery Dog
I guess average winnings, but that sort of hits the same problem

PostPosted: Sat Mar 11, 2017 8:41 pm
by xRECKONERx
i mean i think it's cumulative

and i think that gives people a reason to go for the "big win" to boost their rankings

perhaps we could curtail it to more recent games and exclude older ones in order to eliminate the games from before anon survivor became like a "legit" thing

or yeah average winnings could work too

no reason we couldn't keep both statistics handy.

i think both avg winnings and cumulative winnings gives people more reason to play for a win than elo which gives a minor difference between winning and coming in 3rd