use unfair ai?

His problem isn't with the difficulty persay, but rather that the AI will throw all its eggs in one basket so if it loses one big fleet battle the game is effectively over.

The AI doesn't use many high level abilities, and often won't use them well even if it does.  It doesn't retreat units and lets its high level capital ships die.  It will get huge fleets destroyed by throwing them around recklessly and be unable to replace them because its upkeep is too high.  Speaking of which, it will usually progress very fast through fleet tech and very slowly through other tech types, leading to a generally weak economy and fleet.

The AI's resource cheating just gives the player a long-term experience advantage because it builds copious amounts of low tech level units.  It doesn't know how to use multiple fleets effectively and will just move around with one fleet of death.  It is not proactive, and will run face first into very stupid scenarios, like jumping into a heavily fortified system, then turning around and losing half its fleet because there's a PJI.  It will do this over and over again, getting its fleet blasted repeatedly because it makes the same mistake.

There are countless other ridiculously stupid AI mistakes, but the bottom line is that like most AI's it leans on brute force for its victory.  This means that the moment a player has a distinct economic advantage the game is over.  The AI cannot win because this was its only real leg up.  This is a fundamental problem in most RTS games, and it's clearly not an easy one to solve because I haven't seen a good AI that is proactive.  Every AI I've ever seen is reactive, which is the hallmark of an inexperienced player.

Hi!

It doesn't... against other AIs. Yesternight when I've been attacking the second strongest (hard Vasari) AI, I've fought very strong SBs with 1 and therafter two armor upgrades and at least level-1 weapon upgrade. I had to retreat TWICE before I managed to conquer that planet. Well, the first time I had only 2 cap's, 30 fregates and a few koshikos, and at second time the AI sent its main fleet with LOTS of strikecrafts to defend that planet. 

But when I've been watching that game, I saw that on the other side of it's empire it had just basic SBs: 3k armor and no weapon upgrades, so it barely chased away a fleet half the size I had at the first attack. What put me in a bad mood was the fact that the other AI was constantly sending its ships just to that planet. So why not upgrade there?

That's why I stopped MB it either. Gentlemen's agreement. I let most abilities on autocast, and virtually never direct them by myself, so we fight on approximately even terms. I also stopped using mines. Too much hassle. I'm also considering not fully upgrading my SBs, if not completely abandoning them. Should make games more interesting.

Yep. I set up a SB at chooke point (asteroid field with two mines near sun), where two AIs were trying to send fleets into "my" space. The most hilarious thing was they fought between themself much more than they were fighting me. I just set my ships to hold ground around my SB, and my two cap's there were level 6-8 in no time at all.   That "death train" hadn't stopped until I "circled" he sun and started attacking the most aggressive AI on its flank. And for what they fought? Two metal mines, and behind them in the next grav well a small asteroid. They had MUCH juicier targets significantly closer, but no, they decided my SB on the other side of the solar system is their main target.

AI really needs better routines to asses if it can win the fight. They never managed to make big fleets, because they trashed most of thier newly-built ships on that "ship-grinder" of mine.

BR,  Iztok

You might also try playing against more than one opponent. This assures that you'll have at least a couple full fleets to deal with.

Doesn't help.  I play 10-way FFAs against Hard AIs, or 6-way FFA against Unfair AI, and win constantly. They produce huge numbers of fleets (so that their upkeep is clearly up in the higher techs, which makes sense since those are CHEAP techs), but only make small numbers of capital ships; I've never seen an AI player build more than 4-5 capships.  So, there's really only one part of the game that's actually risky, and that's during the early "land grab" phase when you've only got a few capships and the AI's been mass-producing light frigates.  If you can claim a choke point system and build even an unupgraded starbase there, there's no risk at all; by the time the AI has a decent-sized fleet of heavier ships, you'll have a massive economy and fully-upgraded starbases at each chokepoint.  Then, knock out even ONE AI player and develop his planets, and you'll have such an insurmountable economic edge that the second half of the game just becomes "mopping up"; when I have four fleets that each match the best an AI opponent can throw at me, it's not really much of a challenge.

Really, I just wonder if the problem is in how cheap the fleet-size techs are.  400 credits and 25 crystal is the list cost of the first tech like this, but its ACTUAL cost is the 9% upkeep.  Same for the later ones; if they had a higher up-front cost, would the AI be so quick to run up the tech chain like that?  And if the AI took fewer +fleet techs, would it use the resources it saves to build up its economy and starbases better?

Artificial Idiocy

I think you're referring to RA2.  The issue there was more that the AI had a significant resource cheat (that puts the unfair of Sins to shame...) and then spent it all on spamming base defense.  This made its base virtually impregnable.  As well, its resource cheat was so high that it couldn't actually spend all its money.  As a result, it built just as much stuff on low resource maps as it did on high resource maps (effectively changing the difficulty setting based on the map).  This meant that keeping superweapons active on a low resource map was often impossible to win unless you did a hard rush.  There was barely enough money on the map to actually build a superweapon in the first place, but the AI is rushing you while building massive defense while teching to a superweapon.  It wasn't that the superweapon was good, it was just that the thing had such a crazy resource cheat.  If you played on a higher resource map, you could match the AI's tech pace and it wasn't a big deal at all.  In fact, on medium-high resource maps I only ever got nuked if there were more than 5 enemy AI's, it was that easy to stop.  The fact that I could still win after taking three or four direct nukes should tell you that they were hardly game enders.

RA2 superweapons were actually quite mediocre, contrary to popular opinion.  It sucked to take a nuke to the face, but in reality it often cost less to repair all the damage than the enemy spent on the nuke silo to begin with!  They cost $5000 to build and had a 10 minute delay between uses (including the first use).  Because you typically needed to fire off two nukes before it actually did more damage than the silo costed you (three if you had the expansion where enemies could activate a force shield), they just weren't a very viable investment compared with getting more units.  They were good weapons to break a fortified turtle, but that was the only reason to ever get them instead of more units.

Warcraft III wasn't bad.  If players sat in base and built up, then it would often end in a steamroller, just whoever had the right army to counter the other, but if you were fighting all the way through the game WC3's end game was superbly balanced.  Of course, the AI couldn't handle a decent player that constantly pressured them, and could be easily wiped out in the first few minutes, but I think we've already established that AI's typically aren't anywhere near even an intermediate human's abilities.