Intelligent AI still fails at this » Forum Post by onomastikon

Thanks Frogboy!

Some other suggestions:

(I am playing at intelligent level)

- Make it very expensive to go over your logistics limit for starbases - I find it depressingly easy to afford to build them over limit now and spam them everywhere. Even early in the game the cost for going over limit is so low that I have no trouble affording it.

- Another help may be to increase logistics costs somehow for starbases that are far away from any of your planets. The father away you build them, the more they cost in per turn maintainence.

- Only make the first level of offensive techs available to non-military starbases. After all, regular starbases are not dedicated military bases. Perhaps also only allow the first level of defensive techs for these non-military type bases too - but allow them to form a fleet with small hulled or tiny hulled ships to increase defense instead.

Successfully updated karma reason!

So what I've done is make the starbase militaration techs cost more to get to. The first one is still relativley cheap to get to, it's defensive in nature and we still want people to be able to use them to give their home units an edge (or even offensive units an edge). But the subsequent ones cost on par with what the higher level military techs would cost.

Secondly, the attack assist modules will cost $300 (base) to add on. Now, experienced players later in the game will consider $300 chump change and that's fine. But what it does prevent is a player from being able to go right after these techs and crank out a game-ending strategy in the first couple of years. It will give the AI time to put outs strategy.

Making techs and modules more expensive probably won't hurt and might be a step in the right direction, but I doubt that it'll be the solution to the somewhat uber-strong military starbases. One problem is that modules come for free for evil civilizations anyway (I never quite understood this advantage, by the way). And, the problem with monetary costs in general is that they are not scaled with regards to how much income (galaxy size, number of habitable planets, etc.) you actually have. 300bc is expensive in an early game in a small galaxy, it's not even worth mentioning in midgame in a not-so-small galaxy. (The remark regarding lack of scaling could be made for tech costs and some other things as well, by the way.) And, additional costs would cause a number of new popups for the player to deal with as well.

I have two questions regarding military starbases:
- how are their effects considered for the military power graphs?
- does the AI, both in strategical evaluation of a potential war and in tactical decisions, factor those effects in?

Successfully updated karma reason!

I like the tweak to the starbase techs to make them more expensive meaning that for the most part they will have to wait until an empire has built up the research buildings to get it in a reasonable period of time. However the 300bc cost added is not a deterent in any way if you research Xeno Ethics and go evil which negates all starbase module costs. This can be done really early in the game when that 300bc price tag would actually mean something.

I'd say if you want to slow down starbase construction make starbases more expensive and/or require a few weeks to build after requesting the construction begin. Maybe something like how ship upgrade works where you have to wait a number of weeks determined by how far from your planets the ship is. Also you could modify the cost based on if the starbase is within your sphere of influence or not. Ones outside should be very very expensive (high cost of security) and take several weeks to complete meaning I've now got to defend it and it gives the AI time to react to take it out before it becomes operational. Same goes for upgrade modules added.

Successfully updated karma reason!

What difficulty are you all playing at? The ai has no problem defeating my starbases. His fleets are at usually 30 + attack, with shields and with 100s of hitpoints. I dont know whats going on in your games. Either the difficulty is too low, you are very good at the game or the enemy you are fighting is poor economically. In which case you are bound to beat them.

I have seen the ai lose to my starbases in one game but he was poor militarily. In my current game he is ignoring them and attacking planets and fleets. Destroying them when he feels the need and has a decent sized fleet.

This is not the beta game, but the other patch.

Successfully updated karma reason!

Frogboy

Join Date 03/2001

+2419

That's an interesting suggestion and I'd like to hear what other thigns:

Attack and defense Assist only do the assisting if you have that particular type of weapon.

I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

Successfully updated karma reason!

Maybe it should also be a percentage increase instead.

Successfully updated karma reason!

That's an interesting suggestion and I'd like to hear what other thigns:

Attack and defense Assist only do the assisting if you have that particular type of weapon.

I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

That is just what I suggested in another thread. I would really like to see it implemented, since it makes strategic choices more meaningful. Less meaningful choices = more no-brainers = less fun, while the opposite also holds. Less no-brainers!!

However, I would also like to see
Armor defense boost = adds only to ships with armor
Shields defense boost = adds only to ships with shields
Point defense boost = adds only to ships with point defense

Finally: I think all of these changes would be helpful and good. But at some point, you will need to work on optimizing the algorithms which determine how a ship or a fleet picks a fight. The AI must desist from attacking opponents against which it has less than X chance, where X is a very small number based on that target's defense related to its attack rating.

Successfully updated karma reason!

That's an interesting suggestion and I'd like to hear what other thigns:
Attack and defense Assist only do the assisting if you have that particular type of weapon.
I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

This kind of tweak make sense and will reduce the imbalance brought by military starbase when assisting ships

However, I would also like to see
Armor defense boost = adds only to ships with armor
Shields defense boost = adds only to ships with shields
Point defense boost = adds only to ships with point defense

Well, that is exactly what Frogboy is suggesting

I should add that all those tweaks on military starbase, as well as those made on sensors range will make the Siege campaign mission much harder: I have beat it during the gamma thanks to awacs sensor ship with at least 22 sensor range and fleet of tiny fighters with one weapon boosted by military starbase. With that settings, I was able to shoot down all incoming Dreadlord ships, thanks to the weapon boost in the 3 kind of weapons. This strategy will no longer work, and I don't think it is a really bad thing.

Successfully updated karma reason!

The remark regarding lack of scaling could be made for tech costs

??? Not sure to understand correctly this sentence: the tech cost depends on the size of the galaxy. They aren't the same in a tiny galaxy or in a gigantic one.

Successfully updated karma reason!

That's an interesting suggestion and I'd like to hear what other thigns:

Attack and defense Assist only do the assisting if you have that particular type of weapon.

I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

To me that's almost a must.

I expected it to work this way the first time I used military starbases and was surprised that it didn't.
(Starbases boosts didn't affect ships with zero values in GC1 IIRC).

This solution could make the idea of extra module costs obsolete, and I never felt comfortable with it anyway.

Successfully updated karma reason!

??? Not sure to understand correctly this sentence: the tech cost depends on the size of the galaxy. They aren't the same in a tiny galaxy or in a gigantic one.

Are you sure? I was under the impression that they are.

Successfully updated karma reason!

Are you sure? I was under the impression that they are.

There is an easy way to check this: at the start of a new game, change sliders to produce exactly 1 RP and see how weeks you need. Then do the same thing on another galaxy size.

Unless something is broken, tech cost depends of the galaxy size. The cost for the 2 bigger galaxy size are supposed to be the same.

Successfully updated karma reason!

Okay, I did the test and you're right: There seems to be a 75-80% increase of tech costs between tiny and gigantic galaxies. I compared the research time for xeno engineering, using a custom race without research bonus or starting techs in a tiny and a gigantic galaxy (with everything set to occasional).
I don't think this is enough though - if you compare the number of habitable planets in both cases.

The best way might be to base tech costs on the total number of habitable planets or even on the sum of all planet classes of the galaxy you're playing in. I'm not sure if people are interesting in having that though...

Successfully updated karma reason!

Sorry, yes, you are right, I misread Brad to have indicated attack assist helping that type only, not attack and defense. Sorry.

All good stuff here. Still, I am afraid an "evaluation algorithm" for determining what fights are tactically winnable is in order (call it: Algorithm T), and a strategic algorithm upon which the tactical one superviens (call it: Algorithm S), that is: Sometimes, some fights which, from a tactical perspective (algorithm T), yield a very poor chance of success, should certainly not be undertaken (Algorithm S decides that), while in rare cases, even a very low chance of success computer by Algorithm T might indicate that a try is very important (Algorithm S decides that, for example, if a big enemy fleet with troop transports is just in front of an undefended home planet, but it needs just one more turn to get some defenses up, that attacking that big fleet with only a 2% chance of success should be undertaken).
Otherwise, Algorithm S should be more conservative (depending on how accurate Algorithm T is).

Successfully updated karma reason!

The descriptions for the bases will all have to be changed then, as with this new system it no longer makes sense.

Some of the descriptions are that the base sends out new fighters, which is why you get lazer boost ect. Like small drones. But they turn around and go home if your ships dont already have lazers?

Successfully updated karma reason!

That's an interesting suggestion and I'd like to hear what other thigns:

Attack and defense Assist only do the assisting if you have that particular type of weapon.

I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

I would also like to see this. Not that I ever military starbases anyway but I think it makes sense.

Successfully updated karma reason!

Hi!

i.e. it won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

I vote for.
BR, Iztok

Successfully updated karma reason!

Just want to clarify my previous post here: In other words, I think that the AI has a fairly decent strategic AI, but a very suboptimal tactical one at the moment. After updating the strategic algorithm S I tried to mention above, the tactical AI would need to be updated slightly too to reflect the situations it would still need to integrate to implement suicide runs, too, but I think that it needs a general overhaul in any case.

Successfully updated karma reason!

What difficulty are you all playing at? The ai has no problem defeating my starbases. His fleets are at usually 30 + attack, with shields and with 100s of hitpoints. I dont know whats going on in your games. Either the difficulty is too low, you are very good at the game or the enemy you are fighting is poor economically. In which case you are bound to beat them.

I generally play at painful in the recent beta (for testing), but with the starbase strategy I was able to defeat suicidal in 1.0x without much problem. It is not that difficult to get your starbases to get to 80+ of all defenses in the late game if you have a fully mined military resource (the defense bonus helps starbase defense, but the offense bonus doesn't). This essentially means the starbase has an effective defense of 99. I would take more than 4 blackhole erruptors (the best weapon available) to get through this setup, but I've never seen an AI (beside the dreadlords) use this. Basically, you can throw out all tiny, small, and probably medium hulls, because they will be useless against this defense (just not enough space, exept for maybe mediums with full miniturization). Large and huge ships fully loaded with weapons would be able to scratch it, but most of the time, this is never the case (they tend to carry defenses). So if you spend some time researching starbase fortification and defensive technologies, the AI just doesn't have a chance against your starbases.

Attack and defense Assist only do the assisting if you have that particular type of weapon.
I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

I vote for the idea, although I will admit that this change would just make me switch to using mediums with 1 beam, 2 missiles (sparrow is too big), 1 gun, 1 shield, 1 PD, and 1 armor instead of tiny 1/1 off/def setups. However, this would raise the effective cost of the basic defender by several folds, and cut the number of ships in half (because of logistics), so I believe it will be an improvement. But like many, I don't think it will be enough until the AI is able to acknowledge battles that it has no chance in (it wouldn't be able to scratch a ship because of the other's defenses - if it can hurt part of the fleet, I have no problem with it trying).

On a side note, I firmly believe that this kind of change would make tiny and small ship completely useless and might need to be looked at again in the future... after testing.

Some of the descriptions are that the base sends out new fighters, which is why you get lazer boost ect. Like small drones. But they turn around and go home if your ships dont already have lazers?

This is a minor problem that can be easily corrected. However, the modder in me wants to take this a bit further and ask if it's possible to set up two separate ways for the AI to support ships. One that "boosts" the existing weapon and defenses like what is being suggested, and one that provides overall benefits to all within it's range like what had existed before but even further by not requiring any type of defense or weapon to be present (so that it can support even unarmed and defenseless cargo ships). Even though it might not be used by the main game, or perhaps only offer TINY benefits with a TON of research, it might be useful for modders later on.

Basically what I mean is that the more powerful solutions from starbase militarization like missile bloom, beam multipliers, etc... that offers 2, 4, and 5 attack become support only if that ship has the apropriate module, while basic things like fighter drones which offers a measly 1 laser bonus can be left as support anything in the area, regardless of requirement. This would make them somewhat useful, even though much weaker. Just a thought...

Successfully updated karma reason!

according to my other post, I have still the question, that I've never see the AI build capital ships.
The AI is running suicide attacks on to my starbase only with flets with thin or small ships, but I never saw a battleship or something like that.
So, does anyone ever seen the AI build capital ships?

Successfully updated karma reason!

So, does anyone ever seen the AI build capital ships?

Yes, the Altarians and Humans like battleships a lot more than most. The others seems to like mediums in my games. I think you need to play on larger maps so that the game lasts a little longer and give them some chance to research the appropriate tech and roll those out.

Successfully updated karma reason!

That's an interesting suggestion and I'd like to hear what other thigns:

Attack and defense Assist only do the assisting if you have that particular type of weapon.

I.e. Iit won't add missiles to a sihp that doesn't have missile. It won't add shields to a ship that doesn't have shields.

I think that would be a good idea - and it makes a lot more 'sense' as well.

To be honest, I'd go one step further - the bonus should only apply to Attack/Defense types the ship already has (ie, only get a Shield bonus if the ship has Shields) and the bonus should be a percentage.

Successfully updated karma reason!

To be honest, I'd go one step further - the bonus should only apply to Attack/Defense types the ship already has (ie, only get a Shield bonus if the ship has Shields) and the bonus should be a percentage.

A percentage would be a very bad idea. With a very low percentage, military starbases would be completly useless in the beginning and extremly powerful in the end. With a little higher percentage, they will be more useful in the beginning but they would also get even more extremly overpowered near the end.

Successfully updated karma reason!

A percentage would be a very bad idea. With a very low percentage, military starbases would be completly useless in the beginning and extremly powerful in the end. With a little higher percentage, they will be more useful in the beginning but they would also get even more extremly overpowered near the end.

Actually, I believe the percentage is to stop people like me who uses all 1/1 ships beginning, middle, or end...

But you are right, making it a percentage would make it WAY overpowered if I use blackhole erruptors instead of Stinger IVs

;;;

Successfully updated karma reason!

All interesting, i wonder if its the size of the maps which causes a problem. I also play on the larger maps, and the ai builds big ships, has good weapons. Often builds fast ships (though is not that quick at adding speed to its ships) and can avoid starbases if it has other targets to destroy.

Wonder if the game is more balanced on larger maps. As i say i have not seen these major problems.

Successfully updated karma reason!