Improving the AI in TripleA

@P@nther - Thanks for the warm welcome and for creating this thread.

@all - So to give some background, I decided to create a new AI for TripleA about a year ago mostly since the existing AIs were well… sort of… yeah…

Overall, it has been an interesting journey and I can confidently say that TripleA has a much more competitive AI today than it did a year ago. As P@nther mentioned, most of the AI development conversations are over on the TripleA dev forum and in particular this thread is the core AI development discussion that gives a good overview and documents the past year: tripleadev.1671093.n2.nabble.com/AI-Development-Discussion-tp7585227.html

Currently, I’m the only developer working on the AI and there are a handful of users that pretty consistently help test the new AI pre-releases that I put out. I’m looking for assistance in both development of the AI (need to at least know Java) and in testing in the AI (need to enjoy beating up the AI and posting about it). I tend to put out an AI pre-release about every 2 weeks or so that adds or enhances certain features of the AI which I spend some time testing myself but my testing is somewhat limited since it takes a considerable amount of time.

Now that the core AI phases have been written and its released in TripleA 1.8.0.5, I’m looking to improve things based on the community’s feedback. I recently held a poll on the TripleA dev forum asking what users would like to see improved in the Hard AI: tripleadev.1671093.n2.nabble.com/Hard-AI-Feedback-and-Priorities-tp7587499.html. This ended with the following priorities that I’m currently tackling:
1. Improve Performance on Large Maps (complete for now)
2. Value Exposed Units When Attacking (complete)
3. Improve Air Units Landing Safely (currently working on)
4. Scrambling
5. Valuing Objectives

So previously, I haven’t really been involved with the community here and I’m not sure what percentage use TripleA. I’ve mostly been on the TripleA dev forum and the TripleA war club forum (which is pretty dead now). The final thing I’m looking for is a few potential PBEM games so that I can better learn some of the popular TripleA maps. I’m personally a WW2V2 (revised) player and tend to do a lot of the AI testing on it because I actually understand the map/rules and know the optimal moves. I would like to get better at some of the other popular maps so that I can better train the AI on them as well.

If your interested in either assisting with AI development, AI testing, or playing me on some of the popular maps (PBEM) then please post here or on the TripleA dev forums.

@P@nther - Correct. The pre-release is essentially the AI from 1.8.0.5 plus all the improvements I’ve made since that release (development in Jan and Feb). It can be a little unstable since its less tested with the idea of trying to find any bugs and get feedback on the latest changes. Eventually, it will become the next version of the Hard AI in the next TripleA stable release (1.8.0.6, etc). Users pretty much download it and replace the JAR that comes with the 1.8.0.5 download (I’d recommend saving the JAR that comes with the 1.8.0.5 download in case you want to revert). Then most users play a few games against the Hard AI on their favorite map and post on why they were able to crush the AI :)