Page 2 of 2

Re: Uzebox Tactics (WIP)

Posted: Wed Jan 23, 2019 5:35 pm
by Strela_17
This looks great, it's like Advance Wars's father and I can't not love it!
I'll give it a try and give you some feedback, but you're going on a great first impression.

Re: Uzebox Tactics (WIP)

Posted: Wed Jan 23, 2019 8:50 pm
by Kilo
Thank you!
You can find the actually version on the Wiki.

It is almost like Advance Wars but with some different gameplay and graphics.

Re: Uzebox Tactics (WIP)

Posted: Thu Jan 24, 2019 5:08 pm
by Strela_17
Thanks, I've found it; I've already noticed the first differences with the AW franchise, but that was the first association which popped up in my head when I saw the screenshots.

Re: Uzebox Tactics (WIP)

Posted: Tue Jan 29, 2019 3:32 am
by uze6666
Awesome, will look at it and post a new game alert on it! :D

Re: Uzebox Tactics (WIP)

Posted: Tue Jan 29, 2019 10:46 am
by Kilo
Yeeeeeehaaa.... 😍😁👍

Glad that u r back, uze 😊

Re: Uzebox Tactics (WIP)

Posted: Mon Feb 04, 2019 6:18 pm
by Kilo
I will upload a new version this or next week of Tactics on the wiki.
I made up one more level (map) and I will change the credits behavior and an ending screen for the winner.
So sad that it is only 2 player. I am not really able to build in an AI for playing against the Box. 😔

Re: Uzebox Tactics (WIP)

Posted: Mon Feb 04, 2019 11:01 pm
by D3thAdd3r
Nice! This game definitely deserves some attention and it's nice to see work put into finishing it out after all these years 8-)

It depends how interested you are in AI, and what your expectations are. I would estimate the amount of work for decent AI in this game, is somewhere close to all other work combined so far. I don't have a deep grasp of how the game works so maybe I am off, maybe it is just not feasible on Uzebox, or perhaps there is a simple "dumb but cheating AI" approach that would be reasonably fun and "easy". You could spend a lot of time and try evolutionary AI, where you give it some "sensors" into game state as well as "motor outputs" that trigger preprogrammed actions, then let it play versus itself. The genes are connections/inhibitors between neurons(sensor to sensor, sensor to motor, etc). Keep the best genes and mutate new generations to play against the old(to see if randomly, and unlikely, the mutation is more successful than the old weights, let it run 100,000 generations over some weeks of turbo no vsync play). It is extremely interesting in my opinion and I have had some success with this method, but it takes considerable time.