Fixing AI should be a top priority

I'm generally enjoying this game and think that a lot of imbalance and hopefully the memory leak issues can be resolved quickly. What really needs work though, is the AI. I know the developers are going to be focusing on this, but something must be wrong with some of the basic algorithims and strength calculations that could be fixed quickly. These examples were on Hard:

1) Suicide sovereigns. They come at you with a modest stack and attack your cities, die and that faction exits the game. When they attack, the probability of victory isn't even close to 50/50. When I approach an enemy city with a sovereign, or any unit for that matter, I look at what is stationed in the city and make a reasonable guess as to whether I can win. If not, I leave. The AI doesn't do this. They attack and die. I'm thinking that either the AI doesn't "look" into the city or the calculation of relative strength is off. Maybe the AI doesn't factor magic ability into the relative strength of a stack? Or ignores ranged abilities like archers? Not sure, but I'd think it's a simple algorithim to write that says add up all these stats for the units defending the place I'm going to attack, compare to mine and if less than 60/40, don't attack, and if my sovereign is in my stack, odds need to be 90/10 or better.

2) Declare war then do nothing - lose war and refuse treaties. The AI loves to declare war and then do nothing for hundreds of turns. Then, if you ask for a peace treaty, it's like 50/10000 and there's no way to get it done. I, as a Kingdom, had a large wealthy city right on the border with an Empire that also had a large wealthy city on the border. They declared war, but did nothing - ever. My city was defended only enough to fend off creature assaults (figured I could teleport my sovereign if needed). Eventually, I sent a stack to attack the AI city and lo and behold, only a couple of units. I captured the city and hunkered down, thinking that the AI's army would be on its way....nothing. Attacked and razed two more cities (nice ones)...nothing. This was an Empire that supposedly had 4x my "power" score and twice my population and refused to sign a peace treaty, even after taking three of their cities.

Clearly something is fundementally wrong with both the diplomacy and the AI behavior. It's not even a matter of the AI using a predicatable strategy. It has no strategy! I realize that programming a good AI is one of the hardest parts of the project, but Stardock eventually figured out a pretty good AI for GalCiv2 so I'm surprised that this one is performing so poorly out of the box. It will limit replayability if not fixed, leaving only multiplayer to get a challenege. But of course multiplayer on a 4X game requires massive schedule coordination among players and frequently falls prey to the killer-strategy problem, which saps the fun out of it (for me anyway).

9,969 views 10 replies
Reply #1 Top

Look for more AI improvements in the upcoming 1.1 patch.  Brad has also said that he will be taking a sabbatical in the Fall in order to focus his attention on the AI and mod tools.

Reply #2 Top

I agree that this needs to be the top priority. I can forgive a lot of bugs, balance issues and wonky mechanics as long as I'm having fun. With the AI a complete pushover on ridiculous level, I'm not having much fun any more.

Reply #3 Top

I disagree. Oh I agree that AI is bad, but I think balancing and significant late game changes should be incorporated. Good AI can not be written without first setting the rules of the game, and right now, I suspect lots of changes are needed in the rules in order for the game become fun, THEN good AI is possible to write.

Reply #4 Top

Yep i fully can support this analyze. I started a game on ridiculous difficult and got a war declaration.

 

The AI did absolute nothing against my cities even a really bad defended was in direct neighbourhood of my opponent.

 

I for myself moved to his cities, the enemy stacks IGNORED me and i just conquered his cities one after one. All was bad defended and the big stacks seems to move outside around without any goal.

 

I also really like the game itself but the AI is really terrible at the moment!

 

 

Reply #5 Top

It might have been better for the book and the game if Brad had hired a ghost writer for the book and focused on the game exclusively. Eventually the AI should improve, although the quality of Destiny's Embers is now pretty much set in stone.

Reply #6 Top

Brads taking off to just work on the ai while the rest of the team works on other stuff.

Reply #7 Top

Quoting MxM111, reply 3
I disagree. Oh I agree that AI is bad, but I think balancing and significant late game changes should be incorporated. Good AI can not be written without first setting the rules of the game, and right now, I suspect lots of changes are needed in the rules in order for the game become fun, THEN good AI is possible to write.

Very true,

Should there be any changes made to how cities are managed, the talents added/fixed/removed, you can't start coding the AI for things that don't exist, and you can't use an AI that is trying to use a removed feature. Take note, ain't no coder myself so maybe I'm wayyyy wrong.

Although I'm more than sure that SD are completely aware that they can't just fix the AI and then start messing with features in the game, it's probably going to go hand in hand with the 1.1 patch. Although I'm really hoping for something to come soon to fix the shards and bottomless debt AI. Those 2 things are quite sad and affect game play in a significant way. And My 16 day vacation is coming in about 3 days... I wants to play elemental... D:

Quoting Mtn_Man, reply 1
Look for more AI improvements in the upcoming 1.1 patch.  Brad has also said that he will be taking a sabbatical in the Fall in order to focus his attention on the AI and mod tools.

Will the 1.1 patch be, early, mid or late september?

Cheers,

V.

 

 

Reply #8 Top

Quoting MxM111, reply 3
I disagree. Oh I agree that AI is bad, but I think balancing and significant late game changes should be incorporated. Good AI can not be written without first setting the rules of the game, and right now, I suspect lots of changes are needed in the rules in order for the game become fun, THEN good AI is possible to write.

I agree as well. While fixing the AI might give the most short term improvement, the AI will only become broken again once all the other fixes the game needs are implemented. The game needs an overhaul to it's rules, balance and gameplay first, and only then can work start on an AI that will use those new rules intelligently.

Reply #9 Top

AI is important, but a lot of people still can't play the game very well. Performance and stability, as ever, are the number one priority.

Then, bugfixing. Things in the game should be working as they are *supposed* to work, otherwise you can't make the most informed decisions about what really works, and what doesnt. A lot of stuff in the game is simply glitched beyond recognition, or not working correctly.  When I play the game, I am going to be more quickly put off by special abilities Ive chosen for my character that don't work and glitches that show the wrong city labels, and are apparent immediately, than by the AI weaknesses, which become apparent over time.

And all the AI in the world wont overcome the rampant balancing  and exploit issues.  The smartest AI in the world wont make the game more challenging, when you can buy "legendary" items for all your heroes for pennies in the item stores like they bought legendary armor on a bulk discount, and you can throw infinite + attack and + speed amulets on your characters for a 100 gildar a pop.

So yes, AI is important. But it's more like fine tuning an engine that is already running to factory specs.  The EWOM engine has a half a pound of sugar in the gas tank, and a banana in the tailpipe.

Reply #10 Top

actually, the ai doesn't bother me much at all. i think it will help more to fix the flawed game systems that have been picked up on so far, before the momentum is lost to change them. i'm mainly talking about:

- the entire concept of combat speed (kill it, kill it, kill it)

- the whole way spell damage is calculated, shards and attributes are used

- the map generater putting me in a corner of the map with no metal and only bordered by one civ

until these problem are fixed i will nt have the patience with the game to try and master it to the point where AI becomes an issue. there are plenty of better longer threads on how these problems can be fixed.

i've so far started about 5 games without ever enjoying one of them enough to get my sov to level 5. hell, i've only even conquered an enemy settlement once.