AI Annihilates two of its teammates with moon in GW

Not sure how this just happened. I am playing a GW game on Annihilator difficulty in PTE, this was the final faction leader fight against the Reds on Xianyao, the system with 4 moons in it.

I spawned on the smallest moon thinking that it would be the best way to try to prevent an enemy from rushing Halleys.

Of course once an AI or any player really gets around to building 3x Halleys, 3 more isn't so much as an afterthought!

Anyway, like it did the last time I played this, they plopped 2 other red AI on my 3-halley moon, and I had no subcommander friends, so I was quite preoccupied and decided against spending resources on orbital. It's conceivable that I could have held on because the AI only sent raiding parties at me, but anyway the point is that I simply had no orbital.

Then "Celestial activity detected" and fast forward to the end of the game, in the chronocam the AI on Xianyao Gamma destroyed his strongest teammate on Xianyao Prime with 4 halleys, killing both the teammate on his former planet (which was the actual guy that had the huge base on my moon) and on the target moon and then sent his lone commander to my base, which was obviously really dumb, and I took him out quickly.

This is quite possibly the worst possible strategic decision outside of leaving your comm on the planet you are about to smash. This was a complete annihilation of his two teammates as well as himself (by throwing away all eco). It seems like the AI's decision making process about planet smashing targets needs some work. The AI that smashed the planets had both an orbital launcher and an orbital radar. There was ample evidence that one of my teammates had completely taken over Xianyao Beta and I just don't understand why this moon was not chosen as the target instead.

I thought maybe they just pick the largest planet automatically (... what would be the reasoning for that though?) but the target planet was a 3-halley moon while the planet used to smash into it was a 4 halley moon.

actually, let me reason this out. Well, the game is rigged atm, to only allow larger bodies to be targeted so, smallest planet in the system, you weren't a valid target. Secondary target, if it even decided it was possible you were on another planet, he smashes it instead.

I did same thing, built and sent a Tele, that tricked it into smashing the world I invaded, he killed his ally before I could get to him. I tracked him down and killed in orbit.

not using galleys should be added as an option for Ai logic. They use them even if bad idea.

ya... they really need to make a more comprehensive check.
if planet has no/little friendly and all/most enemy forces, enemy should not smash (like, 0/100 to 25/75 unit ratio between friendly and enemy)
if planet has the reverse, mostly to nothing but allied forces, then this should be first target for smash. (75/25 to 100/0)
if there is none of the above, then it should look for planets that are equally paced (25/75 to 75/25) then it should probably hold off, or send at where commander is.

I don't think that is the case. The 3-halley planet Xianyao Prime was targeted by a 4-halley planet Xianyao Gamma. I started on Delta, and Beta was taken over by two of my subcommanders.

I am not certain but I do believe these named GW solarsystems are all the same so you can check for yourself.

Anyways, since this was PTE maybe it's not a noteworthy bug.

The new units are pretty cool, though! The changes with bots and air are pretty neat. Now if only naval can get some more love.

It also looks like the enemy AI tends to prioritize targeting my T2 bots, which is both annoying and pretty cool. All I know is, those new missile bots kick some ***.

Fixing so where a large planet can smash a small planet is a start.

Making the AI smash the commander if at all possible of knowing where the commander is, is second most logical.

Making the AI refrain from smashing in presence of allied commanders or minor threat level, is also a necesary fix.

Anyway, I know for certain that is the case. Make a game with 5 planets of various sizes and halley number. Make a 600 with 4 halleys, a 400 with 3 halleys, a 500 with 2 halleys, and a 200 with 1 halley. Set AI to sandbox. Try to build halleys on all and try targetting each planet.

The 2 halley wouldn't target the 3 halley, because of actual size. None would target the 1 halley, it is practically invunerable simply by game locks.

I had a similar experience in GW. Was against a faction commander in the 4 moon system with moons that needed 1 halley each to move, the moon I spawned on was just me and a friendly AI so I went and colonised one of the others that was empty and set about building a Halley.

The enemy had full control over 1 of the other moons with the remaining one being pretty evenly matched. When my Halley was complete I found that I wasn't able to crash it in to anything (from reading what other people have said I think it was because the other moons were all smaller - not sure though as at the time I was having a slight UI issue)

Anyway as I was clicking about trying to get my damn Halley to fire and hit the totally controlled AI planet the alarm starts going off and I assumed that somehow I'd got it to work. I checked back on the ChronoCam and it turned out that the AI had built its own Halley on the main planet it controlled and ploughed it into my Halley planet. Was really an awesome thing, but tactically a dreadful decision as the AI had annihilated its stronghold and about 70% of its overall resources.

Didn't affect my gameplay as that planet was my target anyway so things went to plan, albeit that my weaponised hunk of rock just had to sit there and get hit by the target instead of the other way around!

All in all, the AI really sucks at multi-planet combat. They can't even use teleporters yet!

To be fair, multi dimensional thinking isn't as straightforward for AI's as it is for us squishies. And I imagine programming that sorta stuff is a nightmare.

That being said, the AI is a lot less braindead than it was not too long ago.

But yea, *enter russian accent* needs some 'vork'.

AI that knows how to use halleys and teleporters properly.... guys stop asking Sorian to make it better... we need some exploits!

Sorian I'm sure there's some more awesome work you can do on the UI and so on... no need to keep worrying about the AI now

This is actually confirmed in playtesting now. So the AI can actually do this.

Nooooo.... What have you done.... Brian has doomed us all.

I have been smashed by overwhelming armies by the AI. Then, I have been nuked by the AI. Then, I have been planet-smashed by the AI. Now, I look forward to being wiped off a planet with a surprise teleporter surge.

THE AI IS THE SPANISH INQUISITION!

*runs around screaming*