|*KoG*| King of Gores - general

iMilchshake

Hmm my generator doesn't work in a way that would allows you to 'reverse' it, to use it to classify map difficulties. But that is a really interesting side project idea. If you then let it run for all kog maps you might be able to show shifts in the star ratings of maps or smth :D

it it open source

@Bʟυɛ ♡

01:15

can you use cactus client (edited)

01:15

on kog?

01:16

because im arguing

@$Starboy

Aka'jkaem

@Bʟυɛ ♡

addwar aka'jkaem

1

!addwar

05:12

ALL WAR

hii

why this game feel 2x faster sometimes

hi alunia

Hi

Ho

ほ Marin

it it open source

https://github.com/iMilchshake/gores-mapgen-rust

GitHub - iMilchshake/gores-mapgen-rust: Procedual random map genera...

Procedual random map generator for the gores gamemode in teeworlds/ddnet - iMilchshake/gores-mapgen-rust

1

@iMilchshake hi

iMilchshake

I could use the generator to generate tons of synthetic training data for some ML model tho :D

imo there is no need for AI. The generator already looks competent enough. I haven't looked at the GUI yet, but if you have waypoints and a fill-in function. To be honest, I wouldn't know why you should still map. Then you build something here first and tweak the rest of the work. By the way, with fill-in I mean that you mark a certain area and it is regenerated. It would also be cool to be able to customize the parameters after each waypoint. So you have different styles according to certain sections.

2

Arylz

@iMilchshake hi

hello fellow cat profile pic enjoyer

guys, why register KoG&

08:15

?

Kero

imo there is no need for AI. The generator already looks competent enough. I haven't looked at the GUI yet, but if you have waypoints and a fill-in function. To be honest, I wouldn't know why you should still map. Then you build something here first and tweak the rest of the work. By the way, with fill-in I mean that you mark a certain area and it is regenerated. It would also be cool to be able to customize the parameters after each waypoint. So you have different styles according to certain sections.

yeah its waypoint based, the fill-in function sounds like a funny idea! Having different presets is also no problem from a technical perspective, as all the required features are allready in place. The main task/problem would be building the actual UI for that tho xd (edited)

building ui worst part of software development:D

yes :(

08:29

and having no UI is no option for smth like that D:

iMilchshake

yeah its waypoint based, the fill-in function sounds like a funny idea! Having different presets is also no problem from a technical perspective, as all the required features are allready in place. The main task/problem would be building the actual UI for that tho xd (edited)

It would be very awesome if you could specify the difficulty, but thats hard even with ML. We dont have enough data for a good working ML & we are not even sure if our ratings are correct. There might be some wrong ratings applied

08:33

The next problem is that maps difficulty change over years. So we would need some sort of reinforced learning capable of judging new incoming maps and adopting the ratings for existing ones each year for example

08:34

But in general its an interesting topic in ML/AI

Avolicious

It would be very awesome if you could specify the difficulty, but thats hard even with ML. We dont have enough data for a good working ML & we are not even sure if our ratings are correct. There might be some wrong ratings applied

my current procedual generator can do everything from 20 wide easy tunnels to insane maps, but yeah with ML approach it would be very hard

iMilchshake

my current procedual generator can do everything from 20 wide easy tunnels to insane maps, but yeah with ML approach it would be very hard

For ML you need data, clean data

i'd be interested in doing some kind of statistical analysis on all kog maps and their rating

Oh god, you'll find a lot of wrong ratings

08:36

esp old maps

08:36

compared to latest releases

it would be super interesting. i might be able to predict some confidence interval for the r1 time. i might be able to show a shift in star ratings or outliers. maybe it doesnt work at all. dunno. but sounds like a very interesting project :D

iMilchshake

it would be super interesting. i might be able to predict some confidence interval for the r1 time. i might be able to show a shift in star ratings or outliers. maybe it doesnt work at all. dunno. but sounds like a very interesting project :D

Someone did this on ddnet, no?

08:37

Like counting stars on a map

08:37

if they are infront of the moon or behind

hmm dont know that many tw related projects as i havent been active as long as most people here :D

Avolicious

Someone did this on ddnet, no?

you have a link by any chance?

@Avolicious hi

1

iMilchshake

you have a link by any chance?

Currently searching for it

ty <3

iMilchshake

ty <3

There you go: https://discord.com/channels/252358080522747904/746534464984973323/1238111472702193736

1

xddddd

08:41

okay thats funny

Arylz

@Avolicious hi

hey

iMilchshake

okay thats funny

Yeah, there is no "real" profit, but still impressive that people put that much effort in this type of things

yeah i'd do some pathfinding + distance transform stuff on the maps to extract some features by hand (path length, average part width, part width variance, .. idk) and match it to kog's star rating and current r1 time. Maybe i can fit a sane model to it xd

Avolicious

Yeah, there is no "real" profit, but still impressive that people put that much effort in this type of things

yeah i love that :D

iMilchshake

yeah i'd do some pathfinding + distance transform stuff on the maps to extract some features by hand (path length, average part width, part width variance, .. idk) and match it to kog's star rating and current r1 time. Maybe i can fit a sane model to it xd

Supercharged TAS

1

iMilchshake

yeah i love that :D

btw, no you cant say the words you tried to say

Avolicious

btw, no you cant say the words you tried to say

i see xd

@Vampy :3 I find ur all cookies TwT

1

hello, i can't create ticket because a too many already created

4

Shocker

Hi!

1

Avolicious

It would be very awesome if you could specify the difficulty, but thats hard even with ML. We dont have enough data for a good working ML & we are not even sure if our ratings are correct. There might be some wrong ratings applied

Aster

susenka55

Hi!

hello

Ih

Smash

Shocker

Congrats on pink name koggers

Xin

Congrats on pink name koggers

Its purple

Its pink

11:26

For little girls

Love you too

Avolicious

btw, no you cant say the words you tried to say

when manual review my ticket

11

Tnul

when manual review my ticket

1

Avolicious

Oh god, you'll find a lot of wrong ratings

@Krumplee demonic ehm

Not my problem if u cant play

1

do TUR servers have different host or something?

12:39

its much more laggy even tho i get the lowest ping there

guys