posted 05-27-99 02:03 PM ET        

1: A GOOD AI should always pursue long-term strategies (like building terrain- and base-improvements) because it WANTS to – and short-term strategies (like building military-units) because it HAS to.
Time and time again i see AI-controlled factions in SMAC doing it the other way around, and that just dont work against perfectionist players, like myself.

2: A GOOD AI NEVER let its military units wandering around aimlessly just showing off their presence (VERY common in both CIV-2 and SMAC). Any mid- and late-game AI-units (when most of the map is uncovered) should EITHER be garrisoned OR engaged in worthwhile battles. By “worthwhile” i mean launched military build-ups of at least (= absolute minimum) 5-10 units (often more; perhaps up to 15-25 speeder units invasion-armies).
I am also not pleased with the AI “cost-effective units only” strategy, which always means SLOW units. The game-AI should being able to adapt and exploit the advantages of offensive speeder units as soon as possible.

3: A GOOD CIV-AI should ONLY allow wars between AI-controlled factions if the human player is behind (or equal to) it in terms of POWER and SCIENCE.
If the Human Player (= HP) speeds ahead in above areas, the AI should immediately terminate all “internal” quarrels, and instead concentrating on competing with HP. By that i DONT mean that alliances/treaties with HP should be terminated, or that everyone suddenly goes to war with you. Some of the more aggressive factions perhaps starts a military build-up of 15-20 units against you, before launching it, while the more peaceful ones concentrates on building terrain- and city-improvements, besides upgrading their garrison-units. The latter more peaceful approach is absolutely vital for any late-game competition that goes beyond minor “controllable” border-quarrels.

posted 05-27-99 08:19 PM ET            

Just My 2 cents

posted 05-28-99 02:14 AM ET            

Speeders as an invasion force are not a cure all - I use cloaked hovertanks en masse, but if I didn't hapve to fix the walls afterward, I'd use a 12-guage for a fly swatter. The problems with speeders are that they are only speedy on certain terrain, they are more expensive to build, their limited advangtages can be partly offset by ECM defenders, and they only have an attack advantage in open terrain.

The SMAC AI does build too many useless units, and move them around for no particular reason, but if you look at chess programs for an example, they don't have any real mid-game strategic depth unless you give them huge amounts of time between moves.

A game like SMAC has a lot more depth (algoritmically) than chess - there are more "pieces" and the strategic rules vary structurally - terrain and balance of power, etc., far more than in chess.

A game like SMAC is a lot like the game GO from an AI standpoint - huge flexibility and lack of structure, so extremely difficult to design and code for. That's why a rigid but complex game like chess has computer programs that can play on a grandmaster level, but nobody has been able to design a GO program that can play worth a sh*t. I've played both chess and GO on a tournament level, and there's a world of difference.

If SMAC had great AI, but each computer faction took up to ten minutes per turn or more in the mid game, would you play it? The trade off in commercial game design is that the AI has to move hundreds of "pieces" and make a large number of production decisions very quickly and then return control to the human player.

posted 05-28-99 03:00 AM ET            

Worst of all at the highest difficulty, the AI cheats!
A good AI player is a skilled AI player, not a cheating AI.

posted 05-28-99 03:26 AM ET            

Or I might have to send over a few sisylian friends of mine... after I get their kids hooked on SMAC and then aggravated by the high level of "cheating". I'd rather not have my fun ruined so efficently. If that means a crappy AI for the Strategically Advanced, hey, they didn't promise you a fight against Sun Tzu or Genghis Khan...

And their simply IS no excuse for there not to be a good AI Go. With the state of processing power, data processing, and pattern recognition, we should be able to build a Go AI that only takes a minute or so in deciding its TOUGHER turns. And as it plays, learn to try and avoid certain patterns that lead to your EFFECTIVE patterns. Go does NOT have the complexity of options that SMAC has.

Where can I get info on this Go AI competion/challenge? And what formal rules of Go should it understand? Mr. Train claims its a million dollars, but I'd like to see that in writing or more official than just BBS gossip.

posted 05-29-99 11:20 AM ET            

I wish you would do it more.

So far I have heard two things that I am not willing to accept:

1. Firaxis can't get the bugs out because of OS problems. Blame it on Microsoft and WIN 95/98.

2. The AI can't be improved because it can't be improved.

Blah

TheHelperMonkey said Firaxis has to do something. Well, they don't because we "accept" the product as it is. Its a purely business decision for them. Make a lousy car, as long as they buy it, make some more.

God, I am gonna get flamed for this!

I love SMAC, but, ITS JUST NOT A FINISHED PRODUCT.

MICHAELTHEGREAT, I can understand the issue about the time it would take the AI to process turns. Is this going to be used by developers as a permanent excuse? Are we supposed to just accept this? I hope not.

posted 05-29-99 12:15 PM ET            

Meanwhile, since you're paying the development costs up front (or your loanshark, oooops bankers and venture capitalists are) you have all these people with suits crawling up your a** looking for your kneecaps, so they can use their tireirons to check your reflexes and ask you politely about the publication date and the schedule.

So the business reality is ugly, but look at the number of talented designers and companies that have gone by the boards, plus the normal capitalistic frenzy (all the acquisitions, etc.) Try starting up a game company, and look at the size of team and development schedule involved in the industry nowadays.

To be fair on a lot of you guys that bought the game when it first came out, (a) I knew better - I'm a developer, (b) I didn't have the time to play anyway (c) I didn't buy it until I saw the 3.0 patch had come out, and I immediately installed the patch.

Also, I never let the damned thing make decisions for me (autoforming, governors, etc.) it's more tedious, but if I wanter my computers to paly with themselves, I'd point the monitors at each other, continuously stream the Pam and Tommy Lee video, and leave the room! If I want to play, *I* want to play, and it won't make the decisions I will make.

As far as Windows goes, I have one computer with 768 meg of RAM. and disable the swap file - everything, not just SMAC runs with rock solid stability as a standalone app - so if you can tweak the OS, or not multitask and the problem goes away, the problem ain't in the app.

What you need for the level of AI is about like this: A geopolitical module that will look at military force, resources, relative balance of power, and make long term strategic decisions.

A military module that will send (code) outputs to the other modules, "requesting" a specific mix of forces to take the assignments handed it by the geopolitical module, and deploying those forces effectively.

An internal development module that builds and forms everything, subject to the deployment and production demands of the (master) geopolitical module.

All of these need to be simply adjustable for different factions and play levels by toggling a few flags, not by creating new AI modules for each faction.

The number and complexity of different algorithms to do all this is huge, and HAS NEVER BEEN DONE anywhere in the gaming industry. My prediction is that it won't be done for a long time, if ever, in the commercial world.

Maybe, if enough of us convinced the industry that we really do have screaming Pentium II's and serious hardware/ram, AND we'd all shell out $150.00, not $50.00, for a game with kick-ass serious NO BS real AI, then maybe someone would do it - but with IP and PBEM games, and the number of hard-core games who would shell out that kind of bucks, I still doubt they'd be ALLOWED to do it. The money people make the final decisions, not the designers.

I consult as a project manager on development of industrial process apps which use intelligent features, and the scope is not a lot different, but not as extensive as AI-ing a game like SMAC.

posted 05-30-99 06:20 AM ET         

Like Jimmytrick I love SMAC, but, IT"S JUST NOT A FINISHED PRODUCT.

The more you play it, the more it gives the appearance of a good concept that was rushed to completion. The early game is addictive, but from there on it's all downhill, Tricked out and dressed up.... boring...

posted 06-02-99 01:09 PM ET            

It helps if you are nice to the AI at first and trade tech. This makes the midgame much more challenging.

posted 06-02-99 02:08 PM ET            

Seriously though, why should they bother upgrading their AnonI engine as play results come in? They don't CARE, because they are off developing a NEW product. That is the future fuel source for the corporate engine, and much more important than making 50+% of their customers HAPPY with how the game plays or is balanced, etc, etc, etc. After all, unless you are SILLY enough to actually put your name on the game, the customers won't know its another one of the same AnonI engine clones with all the same features and cheats. But then, Sid and Brian have been throwing their names on things... so we are catching on.

Dilithium Dad - Since you haven't played the Civs (I and II) that is the reason you don't realize how OLD those things are. (advancing under peace/truce and sneak attacking, yadda yadda yadda). Aside from the customizing to fit SMAC, the truly new elements of the game AnonI tacts are: Massing a few units before attacking, and understanding when it violates a land border. I am sure there are a couple more, but those are the only ones I can RECALL right now.

You are correct in that its a better Opponent Engine than in previous games by Brian and Sid, and that its better than many other Opponent Engines in other TBS games. But... the Opponent Engine is not advancing anywhere NEAR to the rate that the TBS genre has. And the further it falls behind, the easier it is to kick its butt. Which just puts the cost effective money to make it a brain dead monkey who spastically hits random tactic buttons. Serious players would only play Multi, anyways, or so seems to be conventential thinking. Which makes a self-fulfilling prophesies if they do dis the Opponent Engine.

What we really need instead of Brian and gang slowly improving their Opponent Engine, is to have an Open Opponent Engine Interface or API or something so that we (the players and customers with the want) can tinker with customizing the Opponent Engine to recognize or favor certain tactics over others. This would allow other fellow players to download a copy of our Custom Opponent Engine (be it scripts, dlls, whatnot) and try them out. And modify them. Firaxis would then be able to do an expansion with new scenarios/maps/and alternate Custom Opponent Engines. This effectively allows us, the strategy community of the game, to "teach" the game how to better defend itself and take us on. And and the cost of opening the Opponent Engine... at which point, Brian and the gang can continue to use the technique that they have worked out for their subsequent games. Happier customers, and it cost them just a little more to set up and develop. Its worked successfully in other Strategy games, so the industry is not set against such a thing.

On the matter of Deep Blue... its not a very GOOD Grand Master player, from what I have read. And if it is matched against GMs that don't get flustered at their own mistakes so easily, or that calm down and actually play chess, it gets stomped. That only makes sense though, as we humans have MUCH more pattern and multitiered processing power than any linear processor system. Even when True AIs start playing and learning complex games, I suspect most humans will STILL beat them until the AI is optimized for that single environment. Why? Our ability to apply what we learned in NON-RELATED areas to other areas. The AI won't have the opportunity to see a Sci-Fi movie in which 4 Suicide Troopers units are transported in huge armored transports that sport a light sky craft and get the idea of trying a new combined arms set-up. It will only have its Up Time to draw upon, making it dumb as a box of rocks unless shared on the internet (or some other means of mass communication) so that it has a LOT of uptime to draw upon. And of course, it can only be as GOOD as the players that it is exposed to. If its only exposed to poor players, or players that don't understand the game, it won't get a chance to see things done well, and thereby learn to do similar things.

posted 06-02-99 05:50 PM ET            

Yes I'd rather have Firaxis spend the gold on AI development than nice videos but that's my opinion and I'm certain that a good deal of you will disagree .
(They didn't spend the credits on the game graphics that's for sure, those awfull goddamn ugly copters is driving me nuts! )

SMAC2 will be better