Who to focus development on?

Hard core vs. mainstream

Published on July 16, 2006 By Draginol In GalCiv Journals

Writing strategy games is a lot like playing strategy games. Lots of choices on where to put scarce resources.

As the primary developer of the computer AI for Galactic Civilizations II, I have to make a lot of tough decisions on who to focus effort on.

And by "who" I really mean the general population (95% of gamers) or the hard core 5% who are the most vocal but small as a percentage of buyers.

During development, we focused on mainstream gamers mostly. We wanted to make a game that strategy gamers would enjoy and to do that, we had to be careful where we spent a lot of time on.

For example, we put a lot of time on having really cool ship design. Players can literally create any kind of ship they want both functionally and visually. There's Babylon5 ships out there, Star destroyers, etc. But at the same time, we didn't put in multiplayer because the cost of doing it would have meant the game would have to list for $50 instead of $40 and that was something we wanted to avoid.

Getting back to computer AI -- how tough the computer players are -- one of the game's claims to fames is that the AI provides a challenge to players without having to resort to cheating. The computer players play the same game as the human players do. In most strategy games, the computer players have to be given very significant resource advantages to provide a challenge. In many strategy games, the computer players aren't even playing the same game. One popular RTS I've been playing has its computer opponents able to essentially conjure buildings and units out of thin air in order to provide a real challenge.

But a challenging AI isn't the same as an unbeatable AI. So the result is that some of the hard core players -- as well as top level strategy gamers -- are able to defeat the game not at the highest levels but at the levels where the AI isn't cheating. As a result, they demand ever increasing amounts of time to be spent on making the computer players smarter as opposed to just bumping up the difficulty higher to where the computer players do get some extra resources.

Those players are able to list a myriad of "dumb moves" that the computer players made during a game in question. The problem is that all computer players will make "dumb" moves. Sometimes dumb moves are the result in a logic flaw, other times it's merely a difference in opinion on what is a "good" move, but most of the time it has to do with large-scale, multi-turn coordination which humans are much better at doing than computer AIs.

Right now I'm working on v1.3 of Galactic Civilizations II. In terms of computer AI work, I'm trying to provide something to both groups. I want casual users to be able to have more tools to beat the AI with if they are clever but want hard core players to see smarter AI. I'm also going to put in some additional difficulty levels so that players have more graduated difficulties (the game has several already but more is good IMO).

The issue the developers of GalCiv II are running into is no different than other games. Players of World of Warcraft have long seen Blizzard have to juggle time between the hard core players with their level 60 characters and massive guilds vs. casual gamers who just want to see more missions and more solo-oriented content.

So that's the conumdrum that developers regularly run into when deciding where to put updates.

Comments (Page 1)

samit-

on Jul 16, 2006

I think stardock has done good job. In the future a fantasy based turnbased strategy game could be cool. Like now we've seen MoO 4 and we'd also like to see MoM 2. What comes to writing the AI code and designing a strategy game, I think the ye olde game of chess can teach one what strategy and tactics mean and teach how to take all things into account.

Meglobob

on Jul 16, 2006

No matter what u do, a minority will complain about it. But the majority love the effort u put into the game. Looking forward to 1.3 and Dark Avatar.

The only aspect of the AI which is really poor still and if made better would improve the game considerably is the AI's planetary development. Its terrible to poor at developing planets. I play at suicidal level, I do not use cheesey exploits or any other dodgey tactics. The main reason I can still beat the AI is because of its awful planetary development. The 200% bonuses are undermined by mid to late game because its planets r not developed, well mine are manufacturing, economic and research powerhouses.

crow_t_robot

on Jul 16, 2006

I would want the AI to use cheesy tactics on the Player. Alot of people can resort to cheap tactics like allying with a race and slowly assimilating them into their culture. Or, they stab the AI in the back by lining rows of invasion fleets around their planets and then strike without the AI knowing they just got backstabbed.

SO,if the AI could do these things to the Player, it would make you more aware of your friends and foes. Also, it would add some realism to the game.

Another idea, how about assassinations? I had an event where Netromancer got killed and they declared war on me AND got ships from uknown sources. Suggesting someone did thid on purpose to get them to attack me. If you could do this, and the AI as well, it would make the game more in-depth. I read somewhere the Drath are manipulative...it would make them stand out more if they single-handedly engulfed the galaxy in war by assassinating leaders...

Mike693

on Jul 16, 2006

I suppose I'm in the "hard core" category.

For me, the heart of the game is the AI. I enjoy having to develop a sophisticated strategy for every game. While I love GC2, the situations where the AI does "dumb" things detracts from the experience. I presume that the priority AI improvements that require less computation time have been addressed, and that to get smarter, the "large-scale, multi-turn coordination"-type problems would need to be solved. I like the idea of letting the AI use more time to make better decisions or solve a harder class of problem. Some chess programs let you set how long they are allowed to think during their turn. That might be nicer than setting an fixed target duration for AI turn length.

If you decide to tackle AI problems that take more CPU time, would a slider for max AI turn length be possible?

Shadow

on Jul 16, 2006

The Galactic Civilizations II team has done a very good job balancing the various demands of the players. I believe I read somewhere that you (Draginol) improved the planetary AI so it would be good, but not perfect. I think the planetary AI needs a bit more work or integration with the rest of the AI -> prefarably, this would entail designating certain planets to be one-purpose (Research Capital, Economic Powerhouse, etc.) instead of going for the 'jack of all trades' approach (though this may be because I play at somewhat lower difficulty levels), and recognizing the tile bonuses and using them (or at least realizing that the tile bonuses in effect give the player, human or AI, 1, 3, or 7 extra planetary tiles if they choose to match up the proper structure with that tile). Aside from this, the AI is very well balanced.

crow_t_robot

on Jul 16, 2006

I also saw on the forums that many people would like more minor race tweaks. Things like starting out as a minor and working your way up and minors becoming part of the UP (if they get powerful enough of course.)

Also, the espionage on minors needs to be fixed. Some might say it is a waste of money, but some minors stock up on techs and I want some of them

... but I can't because there is no tab. Also, there is no "break alliance" tab. All in all, minors need an overhaul in general.

Cuddlelump

on Jul 16, 2006

People keeping mentioning planetary development. I would like to make note that in my recent games, the AI has infact done a pretty good job of developing their planets, proper tile bonus usage and all. Main problem I find is that they havent upgraded their various social buildings via research at all, which proves problematic for their effectiveness, they just cant match my resource production.

Now, I do generally play on smaller maps, with only a few opponents and on the toughest difficulty pre-economy bonuses, so maybe that has something to do with it. I've found that overall, the AI does better in smaller maps with fewer habital planets.

MacsenLP

on Jul 16, 2006

I'm also going to put in some additional difficulty levels so that players have more graduated difficulties

I havn't played Gal Civ 2 since one of the 1.1 betas I think just before the new combat system came in. Basically because I couldn't find a difficulty level that was both fun and challenging. Crippling was way to easy and maso you could only win due to AI having awful map settings or through serious cheese imo.

Glad 1.3 sounds like it will address these issues, might even Gal Civ 2 another go time permitting.

Lenius.

theStormWeaver

on Jul 17, 2006

My biggest problem is that I am unable to refuse victory. Let me give an example of my latest map.

It was late game, I had the largest territory, and far more advanced tech. I was planning to destroy the Drath, since they were the weakest, and as of yet I hadn't really spent much time on actually building my navy. But just before I did I noticed they had allied themselves with the Iconians. Normally thats not so bad, an extra month of ship building and prep would allow me to take two civs at once. However, the Iconians were also allied to the Torians, who were stronger than the Drath and Iconians combined. That I couldn't handle. So I allied myself to the Arceans and the Thalan. The Thalan were about on par with the Iconians, and the Arceans close to the Torians. My plan, to divide the galaxy in half in a huge war. Then I realized that when I did that it would satisfy the conditions of a Diplomatic victory. But that would be boring. So I just allied myself with "the other side" and called it a day.

I think it would add to the fun if I could deny that victory and keep going. If I had known it would end up like it did I would have just disabled the Diplomatic victory.

sam0t

on Jul 17, 2006

I also think SD has doned a very nice job with GC2. One thing I would stress out is better or bigger beta tester team. The original had some clear UI flaws in it. I just keep wondering why did not the beta team spot these. Sometimes things are hard to spot, but I guess anybody of us could have noted the awful ship upgrading system the original had. I must also point out the extreme pain it is to give AI correct ships with diplomacy screen. Having to write down ship names on paper before opening diplomacy screen is hardly something I call "without effort"

Most of these issues have been ironed out, but I cannot but help feeling sad that it took two major patches to accomplish this.

On to the real subject which is AI development. AI has indeed gotten better along the patches but planet developing/civ expansion needs some more loving in my mind. Iam not sure if AI expanding too fast (getting broke) and because of this I see alot of half empty planets (tiles) in my games.

More depth diplomacy is in my mind the sugar that keeps games like these fresh and I really do hope that SD puts alot of effort into this with DA coming. With 1.3 patch I would hope that alliances would get some work. Its quite easy to get alliance going in 1.2 and AI never brakes them. After few alliances the game loses its appeal in 1.2. Alliances between Evil and other civs should be quite shallow in my mind, where both parties would brake up Alliance once it has served its purpose.

These ofcourse are only my ideas and as Iam no programmer I have no idea how hard it is to implent them as code. Most of the stuff Frogboy alrdy knows Iam sure. Hopefully 1.3 will be a success so we have something to mess with untill DA

Drengin

on Jul 17, 2006

Great game, great backup, & now great forums!! What more can we ask for??

Thanks for all your support and i for one look forward to 1.3, if only for the new levels (not being able to beat the comp at the higher level an intermediary would be fantastic for learning).

Roll on 1.3

Spiderhip

on Jul 17, 2006

Yeahhh, I want a Computer player which doesn't cheat on us.. I want an a.l. which can play like me , play with the same rules but think better.

I love Stardock, they really care about their customers, their game players.. I hope Stardock can earn good money to make more games, better games.

In these times everyone started to think 3-d super ultra graphical games.But I think good game means good programming and graphics and musics.I mean graphical situation is not enough. I think there are just few games which have a.l. 's as good as galciv2.

The other campanies just preparing games with ultra phsical graphics , but they do not have enough programming..

I'm sorry about my subnormal English knowledge.I can understand better but write worse.

But I think You can understand what I try to explain, tell.

AlphaAsh

on Jul 17, 2006

RE: Planetary development.

The problem isn't so much that the AI fails to specialise planets. Lots of manufacturing and research buildings will give output, regardless of distribution. The main problem is the AI gives up social development and concentrates on research and military spending once it feels threatened by ANY level of military build up. Seriously, I only need to get started on my military and wham, the AI drops social spending to practically nothing.

Even with a stuffed treasury, they fail to BUY super projects they could afford when they hit a certain tech.

This problem is highlighted even more in a mod. The one I'm working on considerably revises planetary improvements and adds many more super projects, and the AI just won't build them once it sees an increase in military might. It incorrectly believes that social production should be abandoned and all spending put into MILITARY research and production. Hundreds of under-teched ships will not ever match a strong productive infrastructure churning out sophisticated warships.

A player has the sense to improve their infrastructure, which increases their military and research production. When I look at a graph my lines rise far far quicker than any AI, because I keep building up my infrastructure. I make my economy stronger. I increase my population. I keep my population happy to increase taxes. All that requires SOCIAL expenditure, which the AI gives up allocating.

It doesn't matter how well the AI places improvements until it at least makes the effort to do this beyond the colonisation rush and the beginning of militarisation.

Improvement placement could be better still too, and specialising planets would make the AI even more competitive, but first thing is first, fix the AI so it doesn't abandon its infrastructure and leave tiles empty or the social queue empty or with NEVER in it. And make the AI BUY super projects, galactic achievements and trade goods when it can afford them.

Finally, to support mods, it'd be great if the AI could tell that 4 tp is better than 3 tp, or 25% morale bonus is better than 20% morale bonus, when deciding what improvements to build. Mods allow for multiple improvements - the AI could do with recognising that.

AlphaAsh

on Jul 17, 2006

Addendum -

If the AI didn't build ridiculous excesses of freighters and constructors, perhaps that expediture could be directed to social spending. Right now my games are lagged out by each AI often having over a HUNDRED freighters bouncing between planets, presumably just in case a trade route is destroyed. The AI should build ONE freighter on ONE planet to establish a trade route when it has ONE trade route free to establish. Also hundreds of constructors should not be bouncing between resources or starbases. It amazes me how many constructors are flying about yet the AI fails to upgrade starbases very effectively. This is a massive waste of military production which the AI could direct to social spending, and it would reduce the process overhead too.

On a tangent, the AI would colonise more effectively if it reduced focus on constructors and built more colonisers. In my mod I gave the construction module a tech requirement - and debug.err is full of messages of every single AI trying to build constructors rather then colonisers at game start. Even without the tech requirement, the AI puts out far more constructors than is sensible, when it could be colonising.

Kosty

on Jul 17, 2006

I see the AI still sitting a lot on lots of cash.

AI planet development is much better but often I see planets that re not fully developed still.

I would like to see the AI Pearl Harbor me, attack me without notice. Now I get a one turn warning in effect after he declares war on me where I can move units. I would rather have to watch the diplomatic relations slider and have a chance he would attack me first if our relations level get too low. Without warning.

Welcome Guest! Please take the time to register with us.
There are many great features available to you once you register, including:

Richer content, access to many features that are disabled for guests like commenting on the forums.
Access to a great community, with a massive database of many, many areas of interest.
Access to contests & subscription offers like exclusive emails.
It's simple, and FREE!