RoboTrannic
Spacefaring
haunting deviantart
Posts: 1,005
|
Post by RoboTrannic on Jul 23, 2016 20:18:25 GMT
so i was just lurking around the dev forus like the golum i am a saw a few things that tjwhale (the guy who works on the cpa) has no idea why there happening so i may have some ideas as to why there happening
1 why increasing the nutrients needed to create a nucleus causes extiction,maybe because they arnt reproducing enough to keep up
2 why turning predotors off causes population collapse,nothing to keep creatures from over poupliating causes a huge spike followed with immeadiate drop from lack of food
*ill update as there thread gose on but for now hoped this helped*
|
|
|
Post by tjwhale on Jul 23, 2016 20:49:20 GMT
Interesting ideas, thanks. 1. You may well be right. With a more expensive nucleus they have trouble reproducing so then they have trouble making enough compounds to get a chance to reproduce. This is a reasonable idea. 2. You're quite right in a real population dynamics system this kind of thing is possible but in this system it isn't really too viable. The equations, the way they are now, balance how much you are eating with how fast you are dying in quite a smooth way. Thanks for the ideas
|
|
RoboTrannic
Spacefaring
haunting deviantart
Posts: 1,005
|
Post by RoboTrannic on Jul 23, 2016 20:59:06 GMT
no probs
|
|
|
Post by fourfire on Jul 25, 2016 16:53:00 GMT
I had a series of comments for the differentiating microbe biomes thread tjwhale
I'll outline a couple of them here:
> Problem: The CPA system requires a "predation matrix".
I have a couple of suggestions here, the challenge is to make a system which works well enough without being too complex. I am a fan of hybrid solutions which perform better than all of their component methods, but this often leads to overly complex solution implementations.
Use client games to 1:1 simulate fights and use the data to generate sparse lookup tables of outcomes between species variants, each successive version of the game contains the composite lookuptables from players of the previous version. I'm pretty sure there's some way to do this without messing with netcode, maybe just get people to manually upload their generated gameplay files somewhere at first.
> Constraints: The big constraint here is that we need to be able to produce a result for ANY possible species that auto-evo might generate.
No, you can get away with sparse lookup tables; you run the expensive testing simulation on some fraction of all possible organisms, and then simply extrapolate an average fitness for/against each tested competitor based on how many mutations distant a nontested variant is from a tested one (if I didn't explain that concept well enough, let know and I'll try to illustrate it).
>Moreover in the later stages we need to be able to produce it for ANY possible species in the whole game, Which makes the space of possible creatures huge.
Yes, but with smart optimization processes it is doable, and becomes more doable every year as the average player's computers improve.
> So if you can easily beat a species yourself then your species should easily beat it, if you lose they should lose etc.
I'll get to this later but the solution involves a different local client file.
I'll have to complete this post later.
|
|
|
Post by tjwhale on Jul 26, 2016 22:27:47 GMT
My concern with a lookup table is the size of it. So say each multicellular species can be made of 5 different types of cell laid out in a 10x10 grid. This is a pretty restricted setup compared to what we would really like to have.
Well the total number of possible configuration is 6^100 (5 different types of cell or none in each of the 100 locations) to there are 6.5 x 10^77 different possible organisms. Now maybe a lot of them are functionally the same so maybe group them into groups of 1,000,000? This still gives 10^71 different groups.
And then each group has to be tested against each other group and you get a matrix with 4.2 x 10^143 entries. Which is a bit big.
|
|
|
Post by Moopli on Jul 28, 2016 4:43:07 GMT
Another problem is, small mutations can have large side effects, so you can't really rely on some form of mutation-distance locality to reduce the number of necessary pairwise species comparisons.
|
|