AI Issue #61c

Sorry, but I wanted this to bounce back to the top so someone would see it and answer before it is buried. Excuse me for self promotion~~~ But it is for the community I push an inqury eh?.....

If you're expecting a dev reply, you won't get one for a few days. It's the weekend and the internet is down at SD to boot.

Hate to tell you this buddy, but if you're still running a beta you don't have the latest version. You should update - the AI is lots better in the more recent versions.

A couple of thoughts...

1) I agree that the AI can be improved at the level of positioning for fleet combat.

2) In the situation you cited, the AI may be preferentially defending their home planets against an invasion "raid" (transports just outside of their sensor range that would take advantage of invading an undefended world. Probably not the best strategy since it leaves enemy ships to be diced up individually by your fleet, but it is a strategy. Further, by keeping ships in orbit, they get an additional attack and defense power point added to existing attack and defense making individual ships a bit stronger when left in orbit.

3) Ships in orbit around worlds by design must defend individually until a specific improvement is made on that planet (orbital something-or-another), and then they can defend in fleets. The idea is that your ships initially cannot be coordinated into fleet action while in orbit until the researchers solve some of the technical issues. Doesn't fully make sense to me, but that the way it is...for both the human and AI players.

I have the version from June 16...1.2. Yeah I just noticed that the "Beta" is stripped but because I have to leave the damn computer ON 24/7 just to keep from having to load a "game save", I didn't notice the change. Hehehe (flame me, go ahead). Just very happy to hear that we can disable Metaverse in the near future as I would like to play the game a few times to completion (on Gigantic) before I submit a score. Correct me if I'm wrong, but most players need a few games under their belt to get up to snuff and know the "score".

Speaking of "knowing the score"...how does one know when a planetary improvement impacts only the specific planet, or the entire culture? (all planets?)...Things we need to know~~~

It's not at all clear in most cases. Maybe someone has already posted a list of "Planetary vs. Entire Culture" improvements? If so can anyone point me in the right direction to said list?

Thanks Gents!

Thank you Tert...and yes I know that planetary coordination "something or other" that you mention, but it just kills me when I approach a system and know I can decimate 2 or 3 times my # piecemeal 90% of the time. Even in this age of cell phones...if I want to throw a party I can blanket SMS everyone I know, to show up at 9pm. Gawd Forbid in the age of Space Exploration that the same partiers wouldn't show up at the same Bat Time, same Bat Channel, to thwart an advance into the System that by most outside observers would have ships populated by the very System planets they were produced within. Therefore said Defenders would have a pretty hectic stake in the dilemma of the enemy approach and would probably want to "gather up".

If Earth were to recieve warning about an impending Asteroid collision, much less Alien Invasion (California aside) , I doubt very much that China, the EU and the US would be developing separate Counter-methods instead of pooling our resources to combat the issue (France aside).

Correct me where I'm wrong, and thanks for the rant. >steps off Soap Box<

I'm gonna try and make an attempt at answering this;

Imagine for a minute, your at war with another race and your getting your butt handed to you. Your fleets are subpar and planet after planet are being invaded and conquered. SO what do you do? Probably after an hour of this, you restart the game. AN option the AI doesnt have.

As an alternative, the AI has another strategy. Tie up your military resourses attacking and invading they out-lying colonies while enmassing they're fleets in the centre of thier space. I think the key word is tieing up. If it takes two or three movement points for you to attack a planet, thats two or three movement points you have lost for attacking another planet.

I always found the closer one gets to the race's homeworld the more likely the planets will have orbital defense platforms (or whatever they're called).

If that's the strategy, it should be building cheaper ships. I'd love to see the AI build faster ships that run away from a killer fleet, and try to get a bead on my transports.

Sadly the AI doesn't build ships fast enough, or have enough sensor range to really do this. If it at least did this it would sometimes take out some transports, or at the very least force me to protect my transports.

I'd love to see the AI also occasionally build raider type ships that would try to get into my backfield and suprise me. (like attacking my unprotected eco starbases)

I imagine it is very hard to program an AI to wage an effective war.

First, the AI has to answer these questions:
1) What is my objective for this war?
2) What general strategy will I follow to meet this objective?
3) How do I implement the details of this strategy (building ships, fleet areas of concentrations, etc.)

But these answers need to be able to change and adapt based on what the AI's enemy is doing and what their strategy is. So what balance is required for offense vs defense? How can I avoid the strength of my enemy while exploiting his weakness?

Human players have a better time of making some of these intuitive leaps, whereas computers have a much harder time.

Consider how long it took for a computer to play a respectible game of chess...and that is on an 8x8 grid with each player limited to 16 pieces. Galactic warfare happens on a much grander scale.

Most game AI's I've played against do poorly in warfare...their strategy is simply to "rush" their opponent and overwhelm them with forces. Attrition at its best. And they are usually beat by human players with appreciation of the centers of gravity of the AI and who do surgical strikes to take out the AI main capability.

I always keep "sleeper" ships, with long-range and speed, into everybody's deep territory (even peacetime, just in case..), just to mess things around with freighters, constructors and transports

Yes I like "Wolfpacks" in enemy territory. Usually I upgrade my otherwise unused Survey ships to become high movement, low cost harrassement when I've already mapped out the current galaxy.

90% of the time I attack planet #1 in said System, and if there are 3-4 ships, I get to attack them one at a time. WTF? A flaw like this is really inexcusable IMHO.

... Huh? Having to attack individual ships seriously kills off the momentum of an invasion. I INTENTIONALLY leave my ships out of fleet until such time as I can muster a counter-attack. The enemies are actually easier to beat in fleets.

I don't think disagreements with one's personal opinion count as bugs.

Ok Dread... Point well taken. However I must say that I really dislike so many attacks turning into some Pearl Harbor, where even the Air Craft Carriers are anchored as sitting ducks (IRL they were already out to sea). How about the same scenario that I'm living out now "in game"....I have 3 fleets in proximity to the Capitol world of my enemy...and they (the enemy) still don't gel into a cohesive group to combat me. My invasion therefore, is not remotely slowed down "momentum" wise.

I would think that the logical thing for the AI to do would be for it to "assemble" at the highest Q planet with all forces in the system. And if you find fleet battles easier to win than against single ships broken off from a fleet...well then that should raise an eyebrow too eh?

Huh?

Furthermore....adding a comment about what seems to me to be a Major Undersight shouldn't be judged as a "bug report". It is rather submitted as stated....a Major Undersight on the plates of the Dev's.

Remember...I like this game in general so far too~~

Huh?

They don't even form fleets in each of their systems! WTF?

Each planet has to have an Orbital Fleet Manager for the ships to defend in a fleet. Most of the time by mid-late game almost all AI planets have these. I never build them, it is a waste of a tile and without one, it takes more time to break through a planet's defenses. So I understand the AI's point of view (wait.. it doesn't have one...)

When you launch ships and "create fleet" at a planet, don't they then defend that system as a group if the planet is attacked from any tile around the same planet? If that's not the case then yikes~~~

When you launch ships and "create fleet" at a planet, don't they then defend that system as a group if the planet is attacked from any tile around the same planet? If that's not the case then yikes~~~

Nope. Fleets only defend the tile that they're on, no Civ3-style opportunity fire or bombardment on adjacent squares. The only way to defend a planet with a fleet is to have the ships in orbit (actually on the planet's square) and for the planet to have an orbital fleet manager built (a much better idea now that 1.2 has removed the "first strike" that attacking fleets got in earlier versions).

So, the AI doesn't fight the in the smartest way, big deal. They'll still kick your butt. Try playing a game on Masochistic or above, you'll be glad they don't fight smart.

At higher levels the AI fights better.
Sensor range is a AI limiting factor.
The AI doesn't quit like a flesh and blood player would, delay tactics are sometimes its only option.
The AI tends to min def every planet. Stops rush invasions with fast transports.
The game is a simulation, some empires are just militarily inept. Think France in 1940.
Coordinating fleets is tough for AI developement. Should get better in 1.3 and beyond.
Play at higher AI levels, increased production means the more toys for the AI to play at which compensates for its lack of a warm blooded intelligence.

"No plan of battle survives first contact with the enemy"

Short of loading someone's brain into a computer system, expecting an AI to think of a near infinite number of stategies in a reasonable amount of time is a lost cause.

Examine the chess comparison above: Even the world's preeminent chess computer, Deep Blue, requires a significant amount of time to determine its next move. And just *one* of the sectors in GCII is nearly 4 times that size (15 x 15 vs. 8 x 8), then magnify that by the number of sectors on any given map (minimum is 4 sectors, max is what 256?). Toss in the fact that all ship directions are unlimited and the range and distance is variable *during play*, not to mention piece strength.

Now, I don't know how SD has programmed their AI strategies, I am sure its no where near as many as there are in any standard Chess text. But as things progress, and certain strategies are discovered, maybe the AI can be made sensitive to them and be a more efficient opponent. froggie said it well, when the AI is losing, it has fewer options in reaction.

Maybe some AI code needs to be added to give it stronger survival instincts when their demise is looming. Right now, its best reaction is to surrender to another AI race. but maybe some more proactive retreating is required.

All in all, as I increase the difficulty levels, the AI seems to improve.

Cheers,
Reaver

The orbital fleet manager will force engagements to occur on the fleet level when planetary battles are fought. If 10 ships orbiting a planet with no orbital fleet manager are attacked, they will battle 1 ship at a time against the attacker. If there is an orbital fleet manager, as many as that civ has logistic points for will assemble and come out. Not building orbital fleet managers is a cheesy way of making things take longer, imo. No navy would send its starships out piecemeal like that inthe face of an organized attacker.

I still don't see how "create fleet" in a planet or star system would hurt. And that can't be all that hard to implement. Small steps first of course. But the current actions are a shame. On ANY level of play, they (the AI's ships) should gel together into a fleet at the approach of an enemy of any size, at the planet of greatest PQ. IMHO~~~

Even at my level of complexity (as I mentioned I don't remember which I picked but certainly a bump or 3 above "normal") they really need not to babysit their planets 1 by 1 with the approach of an enemy fleet. And REALLY not be able to be attacked 1 ship at a time at their planet when there are several there. I don't want to call it pathetic....but really it is.

We can discuss the Deep Blue thang till our eyes drool out of their heads from reading posts...but this is a rather simple issue I bring to the table I think. Convergence. Read any non-fictional history....hell watch Lord of the Rings again...and we all know....combine your stregnth....or converge it at some decisive point. The AI probably can't handle a "grand convergence" on an Empire level, but surely...at least on a planetary/system level.

I'll save the topic of "Priority Checking" for another thread.

Perhaps instead of Orbital Fleet Manager, there ought to be a tech tree thread that deals with incrementally improving fleet coordination over a planet. This changes things from a binary yes/no for fleet combat in orbit to more of a continuum. Just a thought.

there is also the omega defense system which double the hitpoints of ships orbiting the planet its built on.

there is the hyperion fleet defense which assembles all orbiting ships to fight, regardless of logistical limitations.

there is a spin control center which makes your fleets look much stronger than they actually are.

I think military starbases also help, but I'd have to double check to make sure they actually do assist in such battles.

With all of the above, an individual could put 10 huge hulls with lots of firepower and a decent bit of defense around his homeworld. When attacked, they would not only fight as a fleet regardless of logistics, but they'd be a LOT tougher than if they were out in deep space.

I hate to be a "stick in the mud"...but ANY ships in a planetary system would be able naturally (at this point in time) to coalese into a coherent fleet of ...what is it ...@ 10 midsized ships at the highest level of logistics?.... and fight an oncoming enemy. To my way of thinking...if they are orbiting a friendly planet, their logistics should be almost unlimited anyway. It is the oncoming enemy fleet that is light years away from a friendly base that needs to worry about the logistics. Correct me where you find me wrong here eh?

Think WW2 and the US needing to ship supplies to troops in the Phillipines or Siapan, while Japan, if her merchant fleet could have, would have been able to deliver supplies in 1/4 the amount of time.

I'm sure the SD Devs will deliver on this type of issue one day, and I know it's an issue not easily tackled -- nevertheless, it needs to be if this game is to find it's way into legend (which I think most of us in these forums hopes it does).

Dump those tile wasting imlements like "Orbital Fleet Manager"(or give us more tiles in general) and let a group of 4-10 ships in a friendly system assume (gawd forbid) that the locals would be hunting and gathering for the sake of the local ship crews, "pretty pronto" at the approach of an enemy fleet. Logistics IN a radius of a Planet System should be done away with altogether. Probably for a Star System as well.

The Orbital Fleet Manager concept could then be used to AUGUMENT an already naturally occuring and assembled fleet to combat an approaching enemy. Taking enemy planets should be much harder therefore. More expensisive in life and $$$. No "single transport" planet rushes possible. Much less un-escorted transports. Any other way just isn't "cricket".

Fleet Convergence in a Planet/Star System. Not a huge concept, and I think rather easily implemented in this case.