Topic: News about AI

Tibor wrote:

Well, no in-depth analysis is needed here - just overall performance.

Why not? Before you test something, you should know why it's worth to be tested!

E.G pick map, tribe and run two AI-only games for 3 hours with my list and your list and evaluate, findings might be:

• with my list average land was bigger

• with your list 1 of 4 players got stuck, but none with my list

If your lists should just prevent AIs from becomming dead-locked, they should look different... Do you know how or does someone have to tell you?

Then you might decide to tweak your list for better performance.

When I compare trunk vs. my branch I look after overall performance, I dont care about individual buildings, their count and when they were built...

All this testing would take a lot of time so this is why I am not eager to test other variation of lists...

So maybe you should save your time and don't test your lists but different lists?

GunChleoc wrote:

How about having multiple levels of economy, rather than just the basic one? This way we could experiment with multiple stages. If all required buildings have been built for one stage, move on to the next stage. So, the AI could require 1 sawmill in stage 0, but 2 sawmills in stage 3,... of course, this could cause a problem when it's impossible to build all buildings for a stage on the current map, so the rules when to move on should be somewhat relaxed, or be a list of building priorities or some such.

Tibor wrote:

This would be quite different to how AI is built now, basic economy is only a small help when IA should know what buildings to build first and which to avoid on the beginning... As I can see, even for such simple concept it is difficult to find consensus and volunteer testers

It depends what your concept is exactly about. Just avoiding deadlocks? Or a little bit more?

And why is it so hard to find consensus?

Wanted to save the world, then I got widetracked

Top

Quote

I am willing to do some testing but I can´t compile your branch, because I have only 2GB of RAM and the workaround "swapspace" which I used in Ubuntu 16.10 doesn´t work with 17.04 any more

Is there a way how I can get a compiled build (PPA or something) for the branch so I can test the lists?

Top

Quote

I have created a PPA. Once it's finished building, you shoule be able to get it with:

sudo add-apt-repository ppa:gunchleoc/widelands-ai-test
sudo apt-get update

If you have any other Widelands PPAs (e.g. the daily one), there might be conflicts - you can remove the old PPAs with

sudo add-apt-repository --remove ppa:widelands-dev/widelands-daily

Busy indexing nil values

Top

Quote

@Nordfriese

the output to console is bit verbose and can be usefull for you, you can find there info like:

basic economy not established, 1 buildings still missing, f.e. barbarians_lime_kiln
Player has achieved the basic economy at 001:50:41

Now I see, player number is not there, but I can add it, if this will be useful for you...

Top

Quote

Daily is sufficient, thanks

Top

Quote