The Nineteenth Byte

Shebang

4:00 PM

Oh wow

Nathan Merrill

I want to find teams that work well together

Shebang

200 points for a blank answer :P

Fatalize

I can't say I'm proud

Nathan Merrill

teams will initially be grouped randomly, but I'd like to slowly converge to teams that have synergy

Sherlock9

@El'endiaStarman My terms go late August to mid-December; short semester for research from January to early March; then late March to mid-July

6

A: Sandbox for Proposed Challenges

Capsa, a card game KotH! king-of-the-hillcard-games Capsa, known by many times, including the name Big Two in English, is a popular card game in East Asia and South East Asia, especially throughout China, Hong Kong, Indonesia, Macau, Malaysia, Singapore and Taiwan. There are many variations and...

El'endia Starman

4:01 PM

@Sherlock9 Ahh, okay. That's definitely a different term structure.

Sherlock9

Holy crap, I posted this before Christmas @_@

Nathan Merrill

@PhiNotPi does that make sense?

PhiNotPi

@NathanMerrill Are we trying to construct a single team with high synergy, or maximize the synergy of all teams?

Nathan Merrill

teams aren't distinct

so a player can be on as many teams as possible

but that said, we are trying to increase the synergy of all teams

PhiNotPi

But, we want each player to be on (at least a) certain number of teams, right?

Sherlock9

4:04 PM

@zyabin101 Please excuse my weird-ass controller (github link to it is at the bottom of the post)

Nathan Merrill

@PhiNotPi yeah, likely

actually, maybe some sort of evolution

PhiNotPi

Because, the "simple" way is just find the 5 highest-scoring bots and place them on a team, but that's a terrible solution for everyone else.

Nathan Merrill

hmmm, its actually not that simple

because scores are given via teams

and because the top 5 scorers may not work well together

PhiNotPi

Okay.... so we need to figure out a way to give individual players "scores" based on team performance.

Nathan Merrill

I actually think that evolution is the answer here.

flawr

4:07 PM

> I'll trade you, we have an in house php framework with no documentation and a designer that hates object oriented programming.

source

Nathan Merrill

create 1000 teams or so

then of the ones that win, randomly merge and mutate them

the problem is that teams are only 5 players large

PhiNotPi

Idea: we create a method to assign individual scores based on team scores (say, average team score), then, to maximize "synergy," we look for teams that score significantly better than predicted based on its members.

Nathan Merrill

hmmm, that's not a bad idea

PhiNotPi

Are "team scores" simply win/lose or an actual number?

Is there a concept of offense/defense?

Nathan Merrill

the problem is deciding how much of the improvement is due to synergy and how much is due to performance

3

A: Sandbox for Proposed Challenges

Battle of the Fellowships king-of-the-hill epic In this challenge, you are going to create a Fellowship with the goal of defeating all other fellowships in battle. Your group will consist of 4 characters. Character details: Attributes Available attributes are: Strength, Agility, and Intellig...

The best measure is win/lose

you could count the number of remaining players after the battle is over, but I don't know how accurate that is

PhiNotPi

4:13 PM

I think win/lose is the best.

But, that also complicates the problem.

Because now we need to find a way to assign numerical individual scores and team scores based on just win/lose.

Is there a version of multiplayer elo?

Nathan Merrill

the only stuff I found on multiplayer elo divides the scores

PhiNotPi

Like, players ABCDE each have elo scores, and FGHIJ also have elo scores, find a way to distribute them based on the result of the match.

Nathan Merrill

that said, you could assign an ELO to each possible grouping

PhiNotPi

Is it something we can invent?

Nathan Merrill

aka, AB, AC, AD, AE,.. ABC, ABD...ABCDE would each have their own elo scores

oooh, the problem

PhiNotPi

4:18 PM

that's a lot of data

Nathan Merrill

who is AB going against?

unless we pair them against every other possible combo

which is a lot of calculations each game

yeah

maybe we could limit it to pairs

that's 15 pairs a team, so 225 matchups a game

PhiNotPi

I'm surprised this isn't a problem that's been thoroughly solved already.

Nathan Merrill

I'm still leaning towards genetic algorithms

I mean, isn't this what this is meant for?

PhiNotPi

I assume each "genome" would represent a possible grouping of all of the players. If a set of teams A beats the majority of a set of teams B, then A is the better grouping.

Nathan Merrill

no, I was thinking that a genome is a team

randomly pair each team

PhiNotPi

4:24 PM

But, then we would have gravitation towards the single best team, right, instead of using all the players?

Nathan Merrill

if they win, they live on and breed

New Sandboxed Posts

0

A: Sandbox for Proposed Challenges

Generate ASCII TetraVex TetraVex is a tile-based edge-matching game that was originally coded by Scott Ferguson for the Microsoft Entertainment Pack 3 back in 1990. The game consists of a grid upon which square tiles are placed. The tiles have numbers or letters on each edge, and the goal is to ...

Nathan Merrill

mutation will ensure that a player never gets removed from the gene pool, but yes, good players will play more

hmmm, there is a problem with diversity though

Transcript for