Main page

You're taking your time, aren't you? I'm on a similar turn number as you on my Klackon hard game and the only planet in the entire galaxy that is not yet under my rule, by my choice, is the Meklar homeworld Meklon II, as I first want to finish my project Gaiafication of the Galaxy, before dealing the final blow. But, yes, even on hard the game seems too easy for me to justify the name of the difficulty hard. You should especially see the web I weaved over the galaxy, as each system sans 3 (two with wormholes and one without planets or asteroid belts) are connected by Jumpgates. It is an impressive web to say the least, however, it is also somewhat visually distracting.

 

Please Use a spoiler when you are  putting this insane amount of pictures in one post.

 

 

 

 

i have just spent the last 100 or so turns hitting next to see if the ai will actually do anything XD, and to see how autobuild handles the 30 or so of my 3rd wave colonies (not well)... it doesn't seem to understand the concept of terraforming. it just built EVERY structure and then switched to trade goods on barren planets.....

Well it is not much better on Very Hard either... Had some big battles yes but overall i crush the AI. 

 

Autobuild is just dumb AF, building all the defense building first, as if the AI was gonna attack in the next 10 turns... I suppose this is why the opposing factions are so slow : they dont go full eco. 

Edited by Arktheon, 04 April 2016 - 09:44 AM.

Hi Guys, 

 

It seems the development team have some work to do on the AI autobuild, the problem looks like it has a set plan for each type of world and builds too many defence structures, slowing the progression. It looks like there is work to do on the progression logic the AI uses. I think we can agree that phase 2 is better than phase 1, so it is going in the right direction!

 

Nice Meklar walkthrough Deathstalker! But I am going to say I think you were helped on this one by having most AI start opposite you hehe :-) , so they all compete whilst you have unfettered expansion. If you play on Large Galaxy , Very Hard with 8 players it feels more challenging, it does attack if it senses weakness so you have to spend time building Destroyers and Cruisers, whereas I prefer more colony ships. I am not ahead on everything, I am neck and neck with Mrrshan for Excellence and Military but 4th place on science. The Meklars, my closest rival are angry with me for not allowing them to have free travel through my space to make colonies within my borders, it will be war soon. Fortunately for me their plan is to put all but one of their military ships at their home-world and sit there doing nothing. 

 

AI Logic could be something like 

Q - Am I bordering a rival system - if yes, build automated factory, does the system have outpost blocking entrance, if yes then missile base, then starbase, if no, Space Factory, Missile Base, Star Base.

If first answer is no, do I have large population , if yes , do I have abundant or above minerals , if yes build production focused world , if no build research world etc etc 

But also adding algorithms to make sure one world is always focusing on expansion. So it could have designations, border world / research world / production world / colony ship world / junk world and make sure the AI aims to keep a ratio. On junk worlds, build automated factory, biospheres, hydroponic farm, starbase, research + structures, spaceport, colonial revenue services - all things that a fixed plus. 

 

In any of your games, did you try gifting Battleship, Titan or Doom Star tech to an alien race , maybe this way you can get them to build them? 

 

The one thing I am not enjoying is how when you start to get a little ahead on real estate, the races lagging behind all hate you, rejecting all proposals. For example, I have a trade treaty with the Psilon that ended, but I created two more worlds since the creation of the first treaty. Now if I offer to renew it, they want 500+BC and 137 BC per turn to have a trade agreement that won't produce anything like that. This belligerence may make sense for prideful races, but makes little sense for intelligent races like the Psilons, but all the races behave the same, even close allies will keep the alliance but refuse all your dealings unless you agree ridiculous demands :-) The algorithms need a bit of fine tuning , also whilst criticizing constructively, the human representative says friend too often , like every other sentence! lol

Was just going to add, that in my first very hard game I am know curb stomping the AI. The Mrrshan did take advantage of my war with the Meklar, which was good - but once destroying the orbital defences of my world, even though they could bomb, they choose to blockade rather than bomb the hell out of me. Then once I had a cruiser heading to one of their systems, the whole fleet got out of dodge (it was about 50K strength, 65 frigates, plus about 5 destroyers) This left my system completely in tact. They probably should have sent about 12 Frigates to go after it and continued the attack until my fleets go within range. 

 

Eventually the bulk of my forces hightailed it up to the Mrrshan conflict zone and I had 41 ships engage the Mrrhsan's 65 , which I won 65-7 , as they had slightly worse ships and all frigates against a much broader range of my ships, it was something like 1 Battleship, 9-10 cruisers, about 25 destroyers and the rest Frigates.. The tactical combat map broke for an engagement this size and the grid vanished. I was on the wrong PC to be able to send a screenshot. 

 

I am miles ahead on Excellence, Conquest and Diplomatic victory conditions. 

 

I am behind on Science to the Sakkra and Psilon. One thing that is odd is that when you go to exchange tech, they never have any tech to exchange - they are ahead of me, so they have something I don't have Once you get past the first 3 levels of tech, the AI never seems to have anything new to trade or even demand?

I don't know, I do not think there was a good AI in the history of gaming AI-s

 

Of course the game's AI needs to be tuned and made better, after all It's still just early access.

 

 

Most AI in games is hard because They have some ridiculous advantages compared to the player.

Like in Civ V the Deity difficulty is hard as hell because the AI gets an army, two cities and a plus era on you.

In Total War It was extra unit recruitment, and being able to sustain a full army with one territory.

 

However a bad AI does not necessarily mean a bad game, for example Civ V is an excellent game despite It's bad AI.

  

I was thinking about a film I saw where a novice player draws with two grand masters in a simultaneous chess tournament. In this case they had the novice copy the moves of the grandmaster.in each game, so they in effect played each other and she won. A bit of cheating was involved but the principal was sort of there.

 

Now I was thinking how could the MOO AI make use of this, then I thought blatant copying. It can record what the human player does in previous games and if the human player was winning. If winning = yes , then shamelessly copy every build decision in the next game. The bigger the win, the more the AI chooses to adopt the human choices as its own. The same could apply to space combat and ship designs. 

 

 

Taking that further, thanks to modern day internet, you could make it so when you win a game it uploads your score and game to a leaderboard on the internet to win prizes. However, uploading your game gives the server new data, which it uses to program the AI. So if one player totally owns the game as Terran, those game choices are uploaded to the server with the game score, then on the next update, the game adds this algorithm to the AI - who then has a percentage chance it will use the same build against you in the next match if it is playing Terran.

 

You could be really clever and make the computer look at the real estate after say 20 turns of exploration and then decide which successful game closest matches this architecture for this race. Then adopt the choices the human made from that game. 

 

Effectively you then cut out the computer from the decision making, it just needs to search a database of human games and find the one that closest represents your game, then find the most challenging set of moves humans made and throw those at you.