Features vs. AI

Philosophies on development priorities

Like many people who play strategy games, we have a lot of ideas for cool features to have in a strategy game.  They come up all the time.

Space mines, orbital bombardment, tactical combat, heroes, special agents, asteroid fields, special powers for races, and dozens I won't mention right now. ;)

It's not that these aren't good ideas -- they are.  I want many of those features too (except orbital bombardment <g>).  For me, PC strategy games revolve around having good opponents.

A good opponent means either another human being who is decent at the game or a computer controlled opponent.  For most of us, another human being who is at our skill level who is readily available to play a game that is going to last many hours is not a realistic option. That's the drawback on turn based strategy games (which isn't to say that we should shun having a multiplayer feature someday, only that it's not at the same priority as computer players).

The features I describe above could be added to the game in a month or two. Features are easy. Computer players who use them effectively are not.  On paper, having a computer player who "uses" a feature effectively seems easy. But in practice, there's a lot involved.  One mans genius strategy is another man's doom when played on a different map or in different circumstances. 

So like I said, adding features is easy. You can create a bullet list of great features but in my opinion, if the computer players don't use them effectively, then it's largely a waste unless you have a really solid multiplayer component and then it's fine -- for the people who play multiplayer. 

When we do the expansion pack, the features we choose to put it in will have to budget not just the time to implement those features but the time to write a computer AI that does the same thing and does it well.  And that can be tricky. 

And so next time someone wonders why "Feature X" isn't in the game, the answer may be as simple as "Because the time to have the AI use it well was longer than we could budget."  If the response is "Game Y had that feature" then ask yourself - how well did game Y make use of that feature for its computer opponents?

30,332 views 23 replies
Reply #1 Top
....Yet another reason the AI in the Space Empires series isn't so hot. Note: I'm not knocking the SE games. I'm just using them as an example of how a feature-heavy game can make it very difficult to program AI that's decent. GC2 is relatively simpler and more streamlined, which helps make it easier to program AI that's actually competent. Of course, Brad's "733t programming skillz" doesn't hurt either.
Reply #2 Top
Hi. Sorry for my bad english.


I say : Just freeze something, and we can mod.

For me, just add :

- Explain how the mod folder work (anf IF he work).

- No thief/no trade tag for techs.

- Negatives abilities.


The feature / ia debate is, i think, another thing and can go... after a working modding ability.

Reply #3 Top
Draginol, an excellent point well made. Having done my thesis and degree in Artificial Intelligence (some time ago now!). I have some understanding of how difficult it can seem to get a computer program to act in a consistently intelligent manner for what seems to humans an incredibly simple task.

As Martok points out the AI decency in GC2 is largely due to one the 'problem space' being simpler, henceforth making solution descriptions magnitudes easier (a extra feature here and there and can easily destabilize a currently well performing AI). And two a focus on matching appropriate AI development to the areas of the problem space which can approach a realistic solution.

I hope that didn't sound too much like gobbldegook

Thanks for the continued open discussion Draginol on the development of GC2, I know it is appreciated.

Have a good day all.

AR

PS This Split-Personailty stuff (Draginol/Frogboy) is making my head spin!!!!!!!
Reply #4 Top
Features are easy. Computer players who use them effectively are not.


It's refreshing to hear this. I played Civ 3 for hundreds of hours but the people who made it didn't seem to concerned about this fact. eg the AI didn't have a clue how to use artillery/armies and many other things. There are so many great stategy games that have fallen foul in this department.

Expansion packs are big culprits, fleece some extra bucks from hardcore fans with new features that the AI hasn't a clue with.

Although there is plenty still to do in Gal Civ 2 eg colony ships are still being built with a movement of 2 with 30 odd colonists. Gal Civ 2 will hopefully set a new benchmark in TBS AI that other companies will simply have to follow to sell enough units to survive.

Lenius.
Reply #5 Top
Good article, I understand what you mean. I think, any new features added should be relatively simple, and more thought can be given to them later for a sequel.

I think the current combat system is tactical enough considering how you move your ships on the main map. Other features that you mentioned also sound very complicated to add. I'm not totally sure what you mean regarding asteroid fields, but I think some patches of impassable terrain would make things interesting.

Regarding orbital bombardment, I know you've explained your reasons for not including it in the game, but I think it can be made to work. I was thinking that your citizens wouldn't be too impressed with the attitude you present by burning worlds, so a morale penalty might be a reasonable disadvantage (either a small permanent one, or a large one that lasts half a year or something i.e. 20% per bombardment). Eventually your economy would suffer if you kept that up, and your government might even collapse! Diplomatic penalites seem like a realistic consquence too.
Reply #6 Top
Totally understandable. Spend effort on improving the AI, adding other templates for the minor races... the more defferntiatied the AI, the better the single person game is.... the AI needs to use custom race logos and pictures... we should be able to have more then 9+player+minors in the game... let us add up to 20 races (custom as well) so more Ai is necessary.. let us inhabit different types of planets... each race needs to look completely unique in it's ship design, it planetary style.. etc.

People need to start making different tile art...

I've said it before, and I'll say it again... I love this game.
Reply #7 Top
this is true
there's no point to adding a feature if it completely breaks the game
in a way that means the player is overpowered whatever the difficulty level (up to intelligent)

and I don't like having to resort to cheats for the AI to improve the challenge given
Reply #8 Top
Agree 100%. I'm also one of those gamers that plays games to escape from my real world life so the AI is very important. Nothing shelves a game for me faster then when I figure out what the AI is going to attempt to do and can counter that on like turn 2.

Honestly the game mechanic is pretty good as it stands right now. What I would really like to see are just tweaks to some existing screens. For example the Colony tab in the Civ Manager. It would be great if it showed what a planet focus was set to if it is set. I mean I've played enough games now that I can sort of guess what the setting is by looking at the numbers but not always. Besides even if I'm right I can't change it right there.

I do think that the AI must use all new game mechanic features. If it can't then you are just wasting your efforts. I really do like the fact that the AI plays the same game as I do. It's been a wish list item of mine for a very long time.
Reply #9 Top
I agree completely. There is no point adding a rule to a strategy game without telling both sides!
and i can certainly appreciate that 'teaching' the AI a rule and all its implications is very difficult indeed.

I once tried to make a chess AI, and adding in little rule 'tweaks' like en-passant and castling were such a pain!

I really enjoy this game and would not trade the challenge of the AI for a few features if I could then destroy the AI every time using them.

one gripe. I agree with lenious - many of the races start with ion drive (or must get it fairly soon) allowing design of colony ships with base move of 4, so why does the AI not build these (or the fastest it can)? It makes it very easy to out-expand the AI even on v. high difficulty. The AI plays cunningly after this, but is in such a disadvantage after the start.
It seems like it would be easy to change (and I apologise if it isn't, or i'm missing something!)

otherwise, great game, and these dev journals are really interesting!

Mark
Reply #10 Top
In my experience, the AI does build colony ships with at least 3 if not 4 points of speed. 
Reply #11 Top
You can create a bullet list of great features but in my opinion, if the computer players don't use them effectively, then it's largely a waste

Well said.
Reply #12 Top
In my experience, the AI does build colony ships with at least 3 if not 4 points of speed.


They do eventually, problem is a half decent human player has already colonised many of the planets the AI's should have gotten to because they have been producing movement 2 ships. The colony phase is the most important part of the game and currently the AI's can't cope against a decent human player.

In my current beta 4 game large map about 75 planets, by the time i had researched impulse drive (making 2 turn 500 colonist 7 speed colony ships) the AI was still making speed 2 colony ships with 30odd colonists on board i managed to colonise 15/60 of the first colonies v 5 opp's even though i was in a corner sector and i had big range problems getting to lots of potential planets.

Must admit i max out sensors on abilities which helps a lot, but until the AI is focusing on making 3speed range colonys or 4 speed colonys from turn 1 and getting impulse drive quite quickly its going to have big problems.

Putting 30 odd colonists on planets is a big prob as well, it means their economys take forever to grow.

Terrans and Torians were guilty of these behaviours in my current beta 4 game. Didn't see other opp colony ships.

Theres surely many ways to improve the AI's colonisation efforts. Wouldn't be too fond of say giving speed/range bonuses on higher levels but is an option.

Lenius.
Reply #13 Top
Great post Draginol, I completly agree with you.

Nothing more annoying then a AI which is unable to understand certain aspects of a game and fails miserably because of that.

I really think that the approach you're taking towards AI developement is unique in the current gaming world and issomething which makes this so game great.

Just continue they way you people are doing things, imho your on the right track.
Reply #14 Top
I almost never see speed 2 colony ships by the time the AI and I are competing for the same planets on tough gigantic maps. They're usually up to speed 3 or 4 by the time we make contact. I might have a speed advantage, but it won't be a large one. By the time the colonization phase ends, the AI colony ships will mostly be in the 4-6 speed range, with some races having speeds as high as 8. By that time, my colony ships are usually speed 7-9, though I've paid for it by trading a military presense/research for drive research.
Reply #15 Top
I'd be happy to send brad my current save games if he likes just need an e-mail address to send to. On Gigantic maps compared to large you will prob meet another race later into the game so thats why you are not seeing as many slow colony ships because they have had more time to research impulse drive for example.

Lenius.
Reply #16 Top
Lenius, I'm not saying that they start off building speed 3 or 4 colony ships. What I am saying is that they're almost keeping up with me, and almost only because I haven't done any military tech research at this point.
Reply #17 Top
Very, very good journal entry. I am glad to see that Stardock has such open and honest employees. Honest companies are the only ones worth being loyal to.

=$= Big J Money =$=
Reply #18 Top
This is the same thought I have with the endless 'lolz MoO2' threads on the forums. MoO2 had plenty of features - for better or worse - but the AI appeared to have no clue how to use most of them. The race customisation, for instance, meant that the AI was generalised and almost never played to the strengths the particular abilities of a given race provided. Tactical combat was a joke, because the AI couldn't design a hole in the ground and couldn't use all the 'trick' weapons properly in combat.

I'm a proponent of orbital bombardment, even in a limited sense. I don't like that you have to land troops to mess with someone's planet. I'm not interested in exterminating worlds, but blowing up factories, disrupting production, crippling infrastructure, or rendering planets more vulnerable to trade/culture/etc would be fun. However... it would doubtless be a nightmare to build an AI that can intelligently choose bombardment goals.
Reply #19 Top
Stardock's post makes great sense to me.
The only thing I lament about this sensible post is that the crew at Stardock cannot implement these great features AND make the AI work wonders with them! Tant pis, but I sure hope they will try...
Reply #20 Top
Thankfully we have you Frogboy to "keep it real" for us. A good majority (myself included) hasn't the foggiest what is needed to bring about adding in feature x or y. Just feature x or y sounds cool and so forth. Nor do we know the implications of imbalance.

My own opinion is that I reject and or refute any ideas based on proposition that some game made before is far superior to this one. So many posts that say MOO2 was uber, GC2 needs this this and this... fall on deaf ears. I think it has to deal with attitude and tone. If someone says game x had this great feature maybe GC2 should have it.. I would read/listen to it. I guess its the arrogant posts they should go straight to the rubbish bin.
Reply #21 Top
I would certaintly accept that proposition. Taste being subjective, I'm sure some peole consider other games to be better, but to seriously claim that another game in the same genre was vastly superior to GalCiv2? I would think that hard to justify.
Reply #22 Top
It's the people who honestly think GC2 is ever going to be like MoO2 that worry me. The focus of the two games is *totally different*.
Reply #23 Top
Yes you would think that anyone who had played both games would realise that, especially if they claim to be huge fans of one of them ...