What now????

[Talenn]

Kentavr:

Well, for a game this complex I'm quite sure the AI will need a boost to beat or compete with competent human players. The ability to produce an AI which can make long term plans but take into account reactions to actions taken by others is more or less beyond the ability of programmers and current machines. Oh sure, given a LONG time, I'm sure someone could up with something close, but realistically, that aint gonna happen for SE4. So in the absence of an 'uber-AI', I think it will have to have a 'bonus' of some sort (even if just informational to help in it's decision making).

Concerning diplomacy in MOO: Well, I agree that it was pretty weak in MOO2. My favorite for atmosphere and general play was the original MOO. I think the flavour of the music and facial expressions of the enemy leaders was priceless! It added so much to the game. Sure, there werent a million options, but the options that WERE there meant something. How many of all of the options in SE4 have you actually seen accomplish anything? I've tried here and there to 'ask for help' or 'stop hostilities vs a friend' etc in SE4 and usually receive a generic blow-off or else they agree, but then dont follow through. Having tons of options that dont do much doesnt count for too much in my book. And the fact that all the AIs currently 'act' the same makes them feel bland and stale after a while. At least the races in MOO (not as much in MOO2 though) all felt unique and had to be dealt with in different ways.

And FWIW, MOO/MOO2 were remarkably stable for me and the folks I played with (including by modem in MOO2).

Would I rather play SE3 than MOO2? Yes, I sure would. SE3 was a fantastic leap ahead and the ability to customize the enemy races and have them react more or less 'in character' was one its greatest strengths IMO.

And for me, I play with 5000 racial points, but only use 1k or 2k of the points. This gives the AI a decided advantage in that department and I think its sorely needed.

Hmmm, no IMO, nuking the AI before I even meet it is the LEAST of my problems in this game. In fact, whenever I encounter an opposing race I'm usually BEGGING for them to be hostile right off the bat. Its far more challenging than when they come groveling for treaties every other turn. Sure, I can (and do) turn them down, but IMO, they should have some fight in them on their own, especially the ones like 'violent berserkers' etc.

Yes, the combat AI in the SE series is pretty decent as long as there are no complex situations. MOO2 got around this by having a butt-simple combat system in which only the bigger and better direct fire weapons were worth much. This made their combat AI work OK. There wasnt much else to do except move forward and mass firepower, something it did a fair job of. SE's combat is far more diverse and I'm far more willing to cut it some slack in that department. Again, there are just too many permutations for a standard AI programmer to have it perform optimally. No problems there IMO, just make sure it has enough ships in the right places and its off to a good start.

Anyways, there is nothing shockingly new here. Its a simple restatement of the way people are currently perceiving the game. I am still enjoying it immensely due the ability to mod and tweak. If I couldnt or didnt have the time, I probably would have ceased playing as well. The stock AI in the stock game just doesnt hold the interest for too terribly long.

I dont mean for this post to be negative or 'bashing' the game. Overall, I think its great. With some more work, I'm sure it will be even better. Luckily, MM and SG seem commited to continuing that work.

Finally, dont be afraid to pick on us. I think most of us are big boys and girls and can handle ourselves just fine. Its all a matter of opinions in here and everyone is entitled to their own.

Talenn

[Jochen Schmidt]

Hi all,

after reading with surprise how many folks here, think in the same way like me, i haver say something to this discussion:

I never did an Ai coding - this must be very hard to do - yes - but, it can done!!

Well, my baseline is MoO2 - the diplomacy, the characters of the enemies/allies, the strategic AI - really well done!! O.k., now in SE4, there is much more implementet, things a always wanted to have - but, the essential thing, the AI is to weak, much to weak. Btw, i like to play against the computer, cause in a such long-time game like SE4, i don´t know how it should work to play by E-mail. It is even now excuse, to say: "If you don´t like the Ai, play against a human". Also doing my own AI could be a funny thing, but as i said, i don´t know how, and when i should do this - this is the job of the programmers.

Don´t get me wrong (again), i know, that the AI is never as good as a human, but if the Ai here would have only a little more "intelligence", the game would still be my favourite. THIS, i mean the tweaking of the AI, should be the first job to do at the moment. Give it the possibility to use all the compenents, wich are implented in the game, give it the right behavior in strategy... and, this game would rock!!!!

(One of the most famous things for me, was the decision of BTS, to code an (nearly perfect) AI for their Combat Mission game - this is the best AI i´ve ever seen - so, their success as an "independent label" was born by doing it so - not 3D-Graphics, not warsimulation - the AI is the secret of their success.)

I hope MM will work on this. I won´t - ´cause i´m "only" a customer with all good wishes for them and their/my game.

Greetings to all

Jochen

[Courageous]

What will often help an A.I. to defeat human
players (in just about any game) is giving
it the ability to process information in such
a way that gives it an advantage, but at the
same time that information processing would
be far too tedious for a human to engage in.
AI's in such situations excel.

This is one of the reasons Imperialism II was
such a bastard to beat. I'm in favor of
allowing a little "AI cheating," as long as
the cheating is mostly related to the AI
having information that a human player might
not have. There's good rationale for this:
human players have intuition and can make
good guesses. Computers can't.

Not yet, anyway.

C//

[Psitticine]

I think the AI's new intolerence for pushy human players (as mentioned in the Up To Date Beta thread) will help make the game more difficult. One of the problems now is that it is too easy for the human player to pick his (or her) battles. I'm rarely forced into a multi-front battle until the Mega Evil Empire setting kicks in. If the AI had more of a tendancy to start fights when they aren't convenient for the human players, it'd be a more difficult opponent.

[Kentavr]

An AI needn't cheat to win.

Humans are not perfect: An AI can be. This constant idea that AIs can't be smart is just... Well, junk.

Given enough data, an AI can soundly beat a human. Can you ever beat an AI at chess? C'mon.

the only difference here is that the AI doesn't quite know what everything does. It's not even programmed yet to make decisions based upon the universe, but just random chance! That's not AI, that's just... Random.

The computer plays a pretty straight game in tactical combat... If I don't have some sort of advantage in the combat, I'll lose. I'm fairly sure none of you are any better than that. You're going in with hog-wild ships and then saying the AI is dumb, when you obviously outclass it. Try going up against the AI ship that has missiles when you've just got guns. You'll get trounced if you don't have initiative. And I noticed the AI seems to run it's own ships better than a player's... What's with that? :P

But really, until the AI is making decisions based upon the planets within range, what they can do, and making raids rather than just shoving pieces around the board randomly, of course the AI will suck... Because it certainly is A, but it ain't I.

Humans play their games based upon sound rules. It wouldn't be that hard to write them into the game. ...Getting the to work, and have the data for the AI to be able to manage it, that's another thing entirely.

And PS: MOO2 sucked. Small universes, repetitive animations (What's the difference between the same animation twenty times over and a bitmap? The bitmaps is faster, cheaper, and easier to ignore), races that really didn't do anything 'different' than each other, at least when it came to declaring war.

C'mon, really, did you want SE4 next year, after a year of AI tweaking, or did you want it now, when you can play against each other, and learn to make mods? I've been waiting for SE4 all year... And I'm perfectly happy with it.

Crissa

------------------
Crissa Kentavr
Lost Girls

[Taqwus]

I've seen quite a few battles where, in tactical combat, I'll do FAR better than an AI with the same forces.

I once ran a test involving something like 2 heavy carriers and a few DNs for escort, attacking several identical DNs. Under my control, the carriers won hands down: complete obliteration of enemy force. With the AI controlling both sides, the defenders win hands-down, obliterating the carrier force with no losses. Conclusion: Micro-manage fighter combat. It has no understanding of holding back waves, or 'recycling' fired rocket-fighters to soak up PD so that unused RFs can fire unharmed, or using ships as decoys so big guns impact on ship shields, not fighter swarms.

In smaller galaxies before CSMs were boosted in speed, I routinely used an escort with one CSM I to destroy an escort with one CSM I by the obvious run-while-firing-out-yer-rear tactic.

I've used *1* DN, with no weapon of any range above 7, to defeat a fleet of *7* DNs and a CA, with range-16 seeker parasites and range-8 beams, partly because the AI doesn't understand when missiles are useless.

Also, it still doesn't take full advantage move-fire-move, which is a must when taking on a planet protected with 30 weapon platforms with range-8 beams and range-16 speed-5 missiles... something that can be done with zero risk w/ a speed-6 ship w/ range-7 beams, but only in tactical.

It's really not that good, tactically, as of the first patch...

------------------
-- The thing that goes bump in the night

[Courageous]

It's funny; you conclude that you should
micro manage fighters, where, in a game
against the A.I., I conclude that all games
should be strategic. Why would someone want
an ADDITIONAL advantage against the A.I.?

C//

[Seawolf]

Taqwus,

Since MM is working on the game I'll defend them.

Personally I think you are asking too much from the AI at this stage of the game. Also considering that this game runs on home PC rather than a Cray or Deep Blue a human player will allways have an advantage in tactical combat. How could a AI micromanage a large fleet of fighters like it seems you do? An AI improves as it plays others and learns from experience. This AI is under the constraints. MM wanted to get the game out and running, so while the Ai may not be much of a challenge for you try playing other human players. Have you uped the AI bonuses in the game you are playing? That does make a difference too.

Sorry this game doens't met your expectations as of yet but give it time.



------------------
Seawolf on the prowl

[Baron Munchausen]

The AI is getting better at tactical combat but still flubs now & then. It's programmed to fire missiles & then run away, for example, but occasionally runs TOWARDS the enemy it just fired at instead of away. Duh! I've slaughtered more than a few Jraenar ships because of that goof. The latest beta, as the info file posted in another thread here states, is supposed to have improved tactical combat. I've got to get into contact with some major rivals in my new game and see how it goes now. I did have one serious contest with a neutral that I stumbled across shortly after starting the game. They were in the adjacent system, and even though they only had 20 turns or so start they were suprisingly tough.

[Taqwus]

One improvement would be if the fighters didn't go tearing off towards the enemy until all were launched, or other ships on the same side do. Going piecemeal is a good way to get diced... especially against a PD-heavy fleet, because PD now targets a finite number with high precision (an alternative: PD fires w/ fairly low probability to hit, but for EVERY incoming seeker/fighter). But for now, fighters are much more potent massed, and with a keen eye on enemy PD count.

If it could hide its attitude, it'd also gain a bit methinks; right now, once your score hits 500k+ (possible to do quickly even w/ a single armed ship, if you go econ-happy), you know that you still haven't tripped 'evil empire' status until they go Murderous -- and once they do, you still have a few turns warning methinks. Add an Intel op to discover their "true" impression if need be; you may *think* they like you, but... I'm also tempted to think that the racial traits maybe should be hidden, which would slightly help AIs since people are much better at deducing a possible playing style from the traits...

I suspect the AI may not analyze ships as often as it should; saw a fleet running around w/ Org-6 weapons, and a ship that I'd captured and gave to them, that IIRC had Org-10 weapons. Hyper-plasma V, anyway, versus the much worse Lightning Ray III. Something to check for whenever it receives a ship that's a gift, or it captures one. A bit odd, since analyzing for a new tech is usually a plus. If it doesn't have any tech you can't learn (e.g. you've gotten a ship w/ WMGs and a Talisman, and you're not religious... might actually want to keep it, then), then it's almost certainly a plus, unless you're absolutely, critically low on resources and you'd prefer to scrap it.

------------------
-- The thing that goes bump in the night