Faulty AI ships/All AI Enermies ships are useless

Try out the balance mod, ill dig up the link later if someone else doesn't beat me to it. Stock AI sucks. Balance mod AI vs stock AI is no comparison. They actually build weapon platforms and fighters on their planets and can overwhelm you if you are too lazy early on. I am regularily 4th or 5th in the comparisons even though I am colonizing planets left and right and building lots of ships and units.

Both The Balance Mod and the IRM fix the AI in this regard and fighting the AI’s in both especially at higher difficulties is getting harder. I have not lost a war so far but I am a good tactician and strategist with lots of practice but I have lost plenty of battles and especially in the IRM won a few which could be described a Pyrrhic.

Exactly, there is definitely something wrong with AI...

on 1.25 play alone vs all comps - half of galaxy is conquered by 200th turn (despite my conquest was lazy - I did not hurry to wipe out anyone - I wanted enemies to evolve... ehhh), the greatest battle which had to occur had to be comenced between my last generation cruiser vs 89 of enemy light cruisers, I hoped to test new design, but as it became obviuos later - the 89 size feet of enemies in their own system was out of supplies... So the battle became a simple genocide of sitting ducks.
And in general while my ships pack techs 100, their are still flying with techs around 10... it is not big fun since one cruiser can wipe out 3-4 entire star systems with all enemy defending ships/structures before the need of resupply... Yes, exactly, resupply, as repairs is not needed at all

In general game and tech tree is GREAT fun, but AI which for some reason does not use mass seeker weapons, plus usually does not try to corner retreating ships - makes game REALLY easy...

Where is it possible to get that famous Balance mod? cuz last 1.25 mod is way too easy.

BM 1.04 is avalible through a link on this forum look in the Mods Sub forum. IRM 9a will be avalible shortly probably just after 1.3x main patch comes out as will BM 1.05 both on this Forum look in the Mods Sub forum

hope that helps

IRM 9a will have many AI upgrades to it.

Indeed it will and as I am using the skills gained playing past and current versions of the IRM currently from 9th place trashing an Major AI empire in BM 1.04 (sithak) in 4th place over all I have his home world, indeed his home system, without loss of more than half a dozen troops I look forward to the challenge.

"Great Korru the sky is full of stars"

Lead gunner Kzin battle carrier Swift strike thirty seconds after engaging the lead elements of the Badger fleet (photo analysis reveals that the stars where exploding badger warships twenty seven of which where destroyed without loss)

In balance mod the AI does everything it can to be a good opponeant, try that and you will see decent designs and strategies.

I can ALWAYS out research the AI though, thats in any of the mods.

IRM 0.9a AI has a few surprises in store.

ya i have to admit, the stock AI- i end up having 1 ship busting up everything... they can send in their fighters, cruisers, carriers, whatnot against my lone battlecruiser and i wipe them all out...

and i run into the same thing, ships that are out of supply...

ground battles, bunch of troops that have no weapons!!! just pure outright slaughter...

you'd think they'd choose some sort of weapon and cut back on movement as opposed to a bunch of units for target practice-

they don't research anything so that they can be more accurate either.

I've noticed in one of the files, the AI divide its research points evenly over 3 main research groups. Example, it will spend 25% on applied science, 25% on theoretical science and I forget the other. Setting this value to 100% in all areas mean the AI will stop dividing its points up. Would this help the AI research technology better? Each empire starts the game out researching items that are in its setup file. Some of these are small weapons, missile weapons, intelligence services etc,. These are hardly useful. A remake of this file and the main research file might make for a better AI.

I was just thinking for a minute like a player and then trying to figure out how to design an AI that works like a player-so much that he would appear human. Then it occured to me that we have the finest minds and players of the entire space empires universe all on one forum. The AI can't design, the designs are absolute garbage. I've seen a descently designed frigate take on multiple dreadnoughts and came through without a dent in the armour.

So then I poked around with the design file, it's pretty straightforward, but it lacks the common sense of a human player, and the specific needs. Most of us agreed on the other thread that a combined arms effort makes the best force, not a single ship. So then I got to thinking that instead of giving the AI predesigned ships (which is sort of tough thinking about all the different tech levels and resource availabilities!) we could vary up what it does use and implement sound fleet tactics.

Using the number crunching ability of the AI, a real enemy would appear. What we need to do as the bright minds of the SE universe is to decide what are the usual ship types. Here is a list of design archetypes that I thought up quickly:

Light:
Scout Ship
Remote Miner
Sensor Ship
Cloaked Raider
Point Defense
Missile
Interceptor
Troop Transport

Medium:
Repair and Construction
Heavy Scout
Missile
Gun
PDW
Offensive (Light armor, faster, more damaging weapons)
Defensive (Heavy armor, slow, lighter but more weapons)
Light Carrier
Sattellite/Drone
Stellar Manipulation
Troop Transport

Heavy:
Repair and Construction
Stellar Manipulation
Gun
Missile
Hybrid
Fighter Killer (emissive armor with lots of PDW)
Offensive
Defensive
Heavy Carrier
Heavy Ship/Planet Capture
Tank (very very heavy protection, less weapons)

Is it possible to somehow teach the AI to use a mix of the above ships, and design them with the aid of our experience?

"Light travels faster than sound, that's why people seem bright until you hear them."

The problem isn't that it divide it's research in 3, but what it research. My AI in IRM divide it's research more, but try to research cheap stuff first before going to techs that take long to get. And it doesn't research more than 1-2 levels at once (they can do more if the amount invested is more than it takes for 1-2 levels). It's better for the AI to upgrade things by small increments, less expensive and less likely a ship will become un-upgradable...

Yes and No.

As an example, there are four technologies with the same price. And both sides have the same research capacity.

The AI researches all at the same time. It gets all technologies on turn four.
If I research them one after another, I will get one on turn one, one on turn two, ….
So, I will have an advantage.

I find it much better to have 4 after 4 turns than one each turns. The main reason is that if they are ship components, they will be availiable for the new design all at the same time... The AI is never in such an hurry to get new tech. And the way it is in stock it will research 5-7 levels of the same thing in a row instead of dividing it's research into several items...

Researching one after another you can still build your ship on round four with all four components. So, there is no negative effect. But maybe you can trade tech one for tech four, so you can build your new ship one round earlier. Faster, concentrated research brings more opportunities for trading.
And if the AI researches things like better facilities parallel, it is simply nonsense.

P.S.: Tech trading is another very big problem for the AI. They often research the same things, even if they have a technology exchange agreement.

The AI will attempt to upgrade designs regardless of what is comming 1-3 turns later. The AI don't plan ahead like a human does. Same for facilities, it will attempt to upgrade them even if a new level comes every turns. It just doesn't know...

The AI don't know what it's trading partner is researching. They follow the same pathern. The AI of IRM is a bit more clever, it follow different tech list depending on states (ie attack or defense bring more combat tools while the other states focus on different things).

One problem with the way research is handled is it's reset each turns. The AI don't remember what it was doing, it just clear it's research orders and redo the script. That's why we can't make it plan ahead unless we make a really complicated script that take into account a lot of variables that a human player take for granted...

There should be some calculation like:

Sum of Component Tech Levels on new design – Sum of Component Tech Levels on old Design = x

If x > y then retrofit ship.

Y could be calculated like:

y = Number of Component Tech Levels researched in the last 10 turns.

Even a fixed number would be better than retrofitting on each new Component Tech Level.

And for facilities there needs to be a check like:

If same facility tech is in research and research time is lower then x, don’t upgrade.

X could be calculated on upgrade cost and benefits of the upgrade.

But, these are crutches. If the AI can’t remember what has happened in the past (such as Warp Point blockades) ore make plans for the future (such as research queues for a new planned ship type), it will never be a good opponent.

To circumvent that, it would be better for the AI to exchange tech percentage instead of full techs. One AI researches 20 percent and the other 30 percent, and in the next round both will have 50 percent research success. But that is also a thing only Aaron can fix.

In my mind, the AI should never give away researched techs unless you are in an alliance and in very good terms. There should not be any freebee before that.