@Esper

It's a slightly better approach overall.

Imagine if you have an AI being 230/30 (so 200 points over the admincap for 80% penalty) while making 180 research. This means that the AI is teching up about as much as a 100-research empire would if it was below admincap.

Maybe your reaction is "Wow, that's horrible, we better build an admin building to fix it". But if you look at the situation closer if an AI builds an admincap building it will cause it to be 180 points over admincap (decreasing the penalty from 80% to 72%). This means that now we are teching up as fast as the 180/1.72 = 104.7 research empire would.

But instead starnet would simply build an extra lab. Scientists make 12 research base, but realistically since research can be boosted by stability and techs (and AI bonuses if you play above ensign) the scientists will be making not less than 15 research. Thus building a lab increases the research output from 180/1.8 = 100 to 210/1.8 = 116.7.

In other words in terms of science output even in the situation which looked disastrous (200 over admincap, yikes!) building a lab increases research more than 3 times better. It's only later in the game when the empire is making much more science per month when building an admin building becomes worthwhile.

In this math example, the "consumer goods per tech" is completely ignored. This has severe impact on how effective this "200 over the cap" truly is. The fact that you need so much more consumer goods to get a tech simply means you also spend more building slots and pops on that.

You can play with 0 consumer goods as a normal empire. The -50% research and unity hurts, but you can get above the normal baseline with techs and stability bonus. 20% happiness is also painful, but you can alleviate it with enough amenities.

Originally posted by Teal:

You can play with 0 consumer goods as a normal empire. The -50% research and unity hurts, but you can get above the normal baseline with techs and stability bonus. 20% happiness is also painful, but you can alleviate it with enough amenities.

This is not the point in question here. We're not asking whether it's possible or not. We're trying to achieve close to optimal gameplay, in this case, for an AI. Your statement doesn't answer either OP's question or add insight to my observation.

I'm not saying that you're wrong or that my observation proves that "200 over the cap" is not the correct way to go; I haven't done the math on my observation to come to a strict conclusion.

But yes, I am taking the "200 over the cap" into severe question because salvor's equation doesn't account for community goods.


That said, if you figured out that playing with 0 consumer goods can be (close to) optimal gameplay in certain circumstances then that would definitely be in the interest of perfecting salvor's AI, but I think we need more details on the steps required for that.

the math example also ignores the increase in tradition costs that being over admin cap brings so leading to slower tradition/ ascension perks

The AI going over Admin cap makes the AI as a whole far less efficient, due to diminishing returns. In observer and multiplayer games, it is not uncommon for the AI to have 200/30 all the way through 500/30 on admin cap. We've seen all the way to +303% in tech costs for some of the AI.

And as Agent Orange touched on, it means the AI has to build more consumer goods facilities to maintain the increased demand on research and unity structures.

500/30 equates to an approx 188% increase in tech costs. This means labs are reduced to an effectiveness close to 3-4 research a month (from 12), compared to base tech/research rates.

alright...I deside to set basic admin cap to 9999 which may resolve the probelm that ai get weak and no acsension in late game
after all its a good mod

I for my part set up a admin cap bonus of 100 for any AI. So they will come to the mid game without problems, after that they can "stabilazie" with that method.

Originally posted by salvor:

@Esper

It's a slightly better approach overall.

Imagine if you have an AI being 230/30 (so 200 points over the admincap for 80% penalty) while making 180 research. This means that the AI is teching up about as much as a 100-research empire would if it was below admincap.

Maybe your reaction is "Wow, that's horrible, we better build an admin building to fix it". But if you look at the situation closer if an AI builds an admincap building it will cause it to be 180 points over admincap (decreasing the penalty from 80% to 72%). This means that now we are teching up as fast as the 180/1.72 = 104.7 research empire would.

But instead starnet would simply build an extra lab. Scientists make 12 research base, but realistically since research can be boosted by stability and techs (and AI bonuses if you play above ensign) the scientists will be making not less than 15 research. Thus building a lab increases the research output from 180/1.8 = 100 to 210/1.8 = 116.7.

In other words in terms of science output even in the situation which looked disastrous (200 over admincap, yikes!) building a lab increases research more than 3 times better. It's only later in the game when the empire is making much more science per month when building an admin building becomes worthwhile.

The AI hurts his economy fairly seriously now by not building for admin cap. I only see them building it after they have way too many scientists already. The actual math is rather simple since 25 bureaucrats double your research once you have 25 scientists bureaucrats are strictly superior. (There are some extreme exceptions but they can be ignored.) This is the worst case scenario ignoring unity and without any bureaucrat bonus or other sources of research such as stations, realistically you want to build them earlier. The bonuses scientists get don't matter. (Except for the difference between each and because of other sources but this is irrelevant.) So going for close to admin cap at least after 25 scientists would be a solid idea for any situation for the AI. Also while culture workers are weak unity itself is not useless.
Unrelated but AI also shouldn't build strongholds (They don't get AI bonus on this right?). They also build too many shipyards yet they go over naval cap so anchorages would be much more beneficial. Worst thing is that they don't deathball their fleet but that can't be modded I guess.

playing my second game with this AI mod i was frustrated how the AI seemed to fall way too far behind in both Technology and traditions...DIscovered the same thing being the problem with admin cap and the AI trying to combat that wit building more Research..thats good. BUT its economy cant handle that. All the AIs in the galaxy are costantly at 0 consumer good and half their planets keep revolting...at around the 2300 mark they have COMPLETLY collapsed and they are just there to be steamrolled...this is NOT fun or challenging...

I've seen this brought up multiple times in discussions, and I can tell you that going over admin cap massively in early game is not the reason why the AI falls behind in tech. I used the same strategy myself early on and not build any admin buildings until my research is around 500, in a 2x tech/ traditions Grand Admiral scaling game, and I still tech and econ snowball after mid game and sped ahead of the AI. The strategy does work and doesn't kill your tech progress long term. The reason the AI still falls behind must be a different reason.