A maxed out fleet logistically is about as concentrated as you get. 

Anything that fights it on an even technological footing will take SIGNIFICANT losses.

The reason is because of the way the fleet system works.  You might have TEN maxed out fleets against ONE maxed out fleet, but only ONE of your TEN can battle it at a single time.  

The AI will still take higher losses than you.  But this is the most cost-efficient method the AI can currently battle a human.  

My main point is the ALTERNATIVES are worse.  Exactly how do you program an AI to have the situational awareness necessary to combat human tactical ingenuity? 

Imagine if you were a extremely productive machine race battling against nimbler and more ingenious foes.  The numbers are on your side, but the tactical flexibility is on the other side.  What sort of strategy would you choose? 

Also, I mentioned in 7a and 7b the conditions that the doom fleet will react to an enemy.  Please read them again.  

Some AI should not organize all of their ships into fleets, they should swarm and attack in many places at once, especially against an enemy who is spread out the same way. You only need one tiny armed scout to eliminate any non-armed ship, a few well armed or one very well armed ship to take out a shipyard, and a task force to take out a starbase. The whole full stacked fleet vs full stacked fleet isn't the only situation or be all end all either.

Just my opinion but it would be wide for the AI to have more than one strategy to be able to utilize.

Not really. If I have 2 maxed out fleets at a planet, my forces are more concentrated there than an enemy who only has 1 maxed-out fleet present. Sure, I can only attack him with 1 fleet at a time, but I have twice as many independent mobile units in the area. I may take horrendous losses with (or even lose) the first fleet that fights... but the other one will definitely win the following fight and I am left with space superiority, where upon I can do what I like. If you actually kept playing a game long enough to have a war where both sides have fleets, you might have a better grip on how war works when both sides have a couple of thousand ships

Your AI suggestion has no means to counter this. That's a glaring error amongst a long list of other glaring errors. Your AI will never, ever be in a situation where it has two fleets in one place to face my one fleet except by pure chance. It would be better to at least have it calculate a random weighted array for the target it sends fleets to, so that one planet tends to get hit the most. You could also, just fer'instance, have the AI look at the comparative position of it's planets, and move any ships produced in a shipyard further than 50 hexs from the enemy to a nearer shipyard so that it builds up fleets in the combat zone faster. You could avoid it actively gimping it's own combat fleets by giving it two different fleet patterns rather than rigidly sticking to one. You could have it actually garrison some planets before just hurling everything at the player scattered and piecemeal.

Your scheme literally makes sending two fleets to attack one planet into an exploit, because the AI can't come up with any strategy to defeat that. It makes filling up a fleet with all combat ships into an exploit, because the AI will break and never, ever fight it. It makes the AI demand the most powerful fleet of ships possible to fight with, and then simultaneously forbids the AI from ever constructing such a fleet. Your AI will send huge numbers of ships and population at random targets that it has little chance of taking and absolutely no chance of holding. The only element of unpredictability you've included is at the worst possible point in the AI's subroutine, since it causes it to scatter its forces, and then has them travel in a completely predictable straight line for half a dozen turns immediately afterwards so that any element of surprise from the random choice is lost.

And this is just the stuff I spotted from one reading of your post, at midnight, in bed, on my mobile phone. Just imagine how many problems an actual AI designer could find in it if he spend some time working through it. 

It is, in short, a truly terrible blueprint for an AI. It is indefensibly bad. As I said before, it is like the placeholder that gets put in before the AI programmer begins working on the project. It's not even really an AI - it actually has no intelligence in it, since the one and only time you have offered it a choice of responses to an external stimulus (in fact, the only time that an external stimulus is ever checked)is so utterly predictable that the player always make the decision for it (and so exploitable that players will exploit it without even meaning to, since just by building the best fleet they can they'll trigger it). If this was the game's AI, you'd have posted a thousand threads by now going on about how it's too easy and exploitable.

Well, one thing is for certain, if you specialize an/the AI to pick a certain strategies, a human can build a tactic around it and exploit it sooner or later. It usually just a matter of playing a few games a realizing that specific patterns always show up.

For example, you've proposed tat the AI should flee from unbeatable forces and give planets up in order to regroup, which, consequently means also that the AI should engage hostiles if the chances are great he is going to destroy his target. Well fine, the exploit against this mechanism would simply to crowd one large attackfleet together to shy away all planetary defenders or nearlike fleets away from the AI territory, fastly followed up by Transports to take his planets without any loss in spacebattles. If the AI manages to crowd a strong vengeance force you simply lure this force away by the use of fast but weak decoys. As long as he chases these you can effectively take all planets from him without having to fight with him....

This is why I agree when someone else said that a good AI is kept a simply as possible. It might not make the best situational choices but it's also the most likely to be exploited.

AI fleets will retreat to their worlds upon meeting a superior foe.

In other words, the AI fleets will STAY in their worlds until they can gain local superiority.  Or until they max out and can engage in offensive operations.

That's what they do in Civ series.  The AI DOES NOT launch sorties out of their cities generally if they're facing an overwhelming force.  

Um, no.  Because the AI has massive production bonuses.  It can AFFORD to take 2 to 1 losses.

What it can't afford to do is to take 10 to 1 losses.  

My point isn't for the AI to fight at parity with the human.  My goal is to make the AI fight CLOSER to parity with the human so eventually their massive production bonuses kick in AND WEAR YOU DOWN.  After you take out one fleet, 3-4 more maxed out fleets will quickly follow.  

It doesn't matter if the AI loses its fleet as long as it can do significant damage to yours.

In other words: war of attrition.  What the Russians did to the Germans in WWIII.  The Russians took horrendous losses, but in the end it didn't matter.

Also, I might point out that if the AI masses its fleets BEFORE war starts (and smacks you with a surprise) you won't even have local fleet superiority.  

Finally, if you do have local fleet superiority, the AI should be rallying its ships from the shipyard to its local homeworlds instead of sending them off piecemeal. 

The goal isn't to make the AI non-exploitable.  The goal is to MAKE THE AI LESS EXPLOITABLE THAN THE AI CURRENTLY.  Just think of all the dumb things the AI does now in war.  That's the standard you have to compare against.  

Their massive production bonuses amount to nothing whatsoever, because they cannot attack my fleet. I simply move my maxed-out fleets toward their shipyards, watch the AI run all its fleets away to hide, destroy all the shipyards and then patiently build up my own invasion fleets, destroying any new shipyards that pop up with my completely unmolested fleets.

Say X is the maximum number of combat ships you can fit in the fleet. Now, your fleets already max out at X-1, due to the transport. But the moment it reaches X-1, it rushes the player. So the most ships that will ever actually be defending a planet is X-2. X beats X-2 every time, so your defensive fleets will never engage. Therefore the shipyards are completely defenseless, and thus the AI cannot channel its production (however enormous) into ships.

Moreover, if Y is the number of enemy planets, I just need Y+1 maxed fleets and a bunch of transports. I send 1 fleet to every planet, destroy the shipyards, and blockade their <=X-2 fleets in orbit. Then I move up my 1 spare fleet. This destroys the enemy fleet defending the planet, probably dying in the process. My transports move in and take the planet. I now send the fleet that was blockading this world on to attack the next one, and repeat the process.

Your rules do not work. They do not make the AI fight CLOSER to parity with the human, because they deny it the ability to build up parity on individual fleets, and deny it the ability to co-ordinate fleets as a group.

Exactly, the AI is only a part of the game. (And BTW, I didn't explicitly state this, but I was only referring to "strategic AI".) In 4x games, contrary to intuitive approach, the AI is not what makes up the challenge. The biggest part of the challenge of 4x game is an "economic puzzle". The "AI" is there only as an imitation of variable opposition. In reality your opposition is still more or less just a "plank" that you either jump over or don't jump over. The efficiency of AI's strategy and "brains" is nothing more than a part of that plank. Thus when making the AI it is more important to make it look smart than to actually make it be smart. It's a mater of perception not a matter of reality. And despite of what we all claim about needing a deep and adaptable opponent we are all easily fooled by a "prop".

That includes MP btw, many players claim that they need an opponent that can "react to their actions" but then go and play games where "everybody and his uncle" follows one simple "ultimate" strategy per "class", and where everyone knows exactly how the battle between two skilled players of two given classes will play out. Apparently that's what reacting and adapting means nowadays... yeah, right...

AI's exploitability indeed makes it look silly and 101% of the time ruins the ruse. However fixing it while sacrificing the ruse itself is NOT a correct approach.

Teaching it to "do the stupid things really well" breaks both it's appearance and it's actual "smarts".

The only thing worse than resorting to spamming the strongest unit is spamming a unit that is not the strongest.

Bottom line - trust me on this, while it may all look great "on paper", in actual game, seeing AI always resort to the same strategy is even worse than seeing it miserably fail at strategy. AT BEST playing against such an AI will feel like ...

XOX

XXO

OXO

... a deja vu.

FWIW, this isn't necessarily true. Yes, if two fleets meet in the open field, both consisting of the same units, then losses will be about equal between the two (or at least it should). However, that rarely, if ever, happens. Given the fact that you can customize your ships, by choosing the correct ships you can minimize-- maybe even negate-- your losses. Even in Civ4, if you promoted your units rights, you could easily kill a stack equal (or even larger) in size, even if you were behind technologically.

The stacks of doom in Civ 4 sucked composition wise.  Hopefully Gal Civ III can do better with their fleet compositions.

Point is: you tell the AI which sort of units it should build, and which it shouldn't.  Consequently, if it only builds "good units" (like carriers) and maxes out logistically a fleet of them, there's little the human player can do to stop that except to take heavy losses.  

The problem is the AI is doing too much: it's building bad units along with good units.  And it's fleet compositions are equivalent to the Civ 4 army compositions: not very good.  Both of those are relatively easy to fix.  

Quoting Reianor3, reply 33

Quoting sweatyboatman,

reply 23

Quoting Reianor3,

reply 15

The point - it's not AI's job to BE smart. It's not the AI's job to be EFFICIENT. It is not even his job to force YOU to be efficient and smart to beat him. His only jobs are to pretend to be smarter than he is and to make YOU feel smart when you beat him.

This is silly.

The AI is part of the game. Beating the AI is not about proving how smart you are, anymore than doing well at Tetris is about proving how well you can line up blocks. You can design challenging games without an AI (Tetris). You can design easy games with impressive AI (the Sims/Cities Skylines). It depends on what you want to do and how you design the game.

GalCiv3 is not supposed to be a sandbox, the game mechanics aren't interesting on their own (though the UI could be thought of as a puzzle-game ).

The design of this game is premised on the AI not only providing a challenge for newbies and casuals, but also for hardcore players. Without a challenging AI the game's depth disappears. Developing a challenging AI is a difficult task, but it's the task that Stardock took on when they designed this game.

Marigolran's contribution to the AI discussion mostly revolves around the player exploiting the ideology tree and rushing on a map with tons of available planets. It's pretty clear that Brad and Stardock didn't really design the AI (or the game) around early rushes, so it's not surprising that the AI doesn't handle that contingency well. But then, I wouldn't handle that contingency well.

I think they'll get this stuff worked out, and the AI will get better at things like martialing its fleets and recognizing threats. But it wont make much of a difference unless they also are open to adapting the game to match the improved AI.

Exactly, the AI is only a part of the game. (And BTW, I didn't explicitly state this, but I was only referring to "strategic AI".) In 4x games, contrary to intuitive approach, the AI is not what makes up the challenge. The biggest part of the challenge of 4x game is an "economic puzzle". The "AI" is there only as an imitation of variable opposition. In reality your opposition is still more or less just a "plank" that you either jump over or don't jump over. The efficiency of AI's strategy and "brains" is nothing more than a part of that plank. Thus when making the AI it is more important to make it look smart than to actually make it be smart. It's a mater of perception not a matter of reality. And despite of what we all claim about needing a deep and adaptable opponent we are all easily fooled by a "prop".

That includes MP btw, many players claim that they need an opponent that can "react to their actions" but then go and play games where "everybody and his uncle" follows one simple "ultimate" strategy per "class", and where everyone knows exactly how the battle between two skilled players of two given classes will play out. Apparently that's what reacting and adapting means nowadays... yeah, right...

AI's exploitability indeed makes it look silly and 101% of the time ruins the ruse. However fixing it while sacrificing the ruse itself is NOT a correct approach.

Teaching it to "do the stupid things really well" breaks both it's appearance and it's actual "smarts".

The only thing worse than resorting to spamming the strongest unit is spamming a unit that is not the strongest.

Bottom line - trust me on this, while it may all look great "on paper", in actual game, seeing AI always resort to the same strategy is even worse than seeing it miserably fail at strategy. AT BEST playing against such an AI will feel like ...

XOX

XXO

OXO

... a deja vu.

Just tell the AI to spam the strongest unit.  

Besides, this is the perfect programming approach for synthetic races like the Yor.  If you were a synthetic race with the tactical intelligence of a doorknob, but an incredible ability to mass produce yourself, how would you fight? 

But that's exactly what it shouldn't be doing.

Aside from the whole big deal of unit balance and the idea that spamming one kind of unit is bad in itself (not to mention the matter exposing that lack of balance), this automatically stops the whole ruse dead in it's tracks. It's supposed to look and feel alive. Spamming one unit it just looks stupid.

If I were synthetic I wouldn't listen to organics who make baseless assumption about the level of intelligence of synthetics. 

I've seen no lore evidence that suggests that Yor are primitive. And a number of evidences that suggests that they are cunning.

Yor aside - don't try to mix the AI as it is in real life, where humans have repeatedly failed to get it anywhere near to being intelligent (and let's face it - whatever it is that we CAN make is not going to match us. We aren't "god" we can't create something that in itself is capable of creativity.), and the fictional AI that is free to have the intelligence level that, in real life, would contradict common sense, logic, historical precedents (or lack there of) and other similar "grounding" factors.

Fair enough.  I am being racist against the Yor, though I still believe the Yor should not desire snuggler colonies.  But is the ruse in Gal Civ III better than the ruse in Civ IV? If not, then Brad should adopt more tricks from the Civ IV people.  

What if they never gain superiority?

And more importantly, how could they gain superiority when nothing is telling the AI to reinforce them?