AI should make sub-optimal choices

One thing that I've noticed is that the AI seems to always pick the same specialization options in the tech tree. That makes things a bit bland and boring and makes it less worthwhile to attempt to horse-trade with everyone so that you have all (3) specializations.

A characteristic of lesser skilled players is that they make sub-optimal choices. They don't understand what things are worth and will undervalue / overvalue various choices. There is a lot of experimentation as a result and a lot of random choices.

When it comes to the specialization picks, lower-skilled AIs should be more likely to pick the 2nd or 3rd best choice for their circumstances. So after it values the three specializations, it might randomly fudge those valuations by +/- 50% before picking.

Locked Post

Advanced Search

Watch this post

Do not email me updates for this post Email me updates for this post

Successfully updated karma reason!

I noticed that also in the recent patches. AI always picks manufacturing specialization and supportive population. Every AI I talked to have these two specializations.

In the old versions the AI seems to randomly pick specializations.

Successfully updated karma reason!

Supportive Population needs to be hit with the Nerf bat. +4 morale means another 4B population at high approval. That is significantly better than a paltry gross income bonus or trade route.

Successfully updated karma reason!

I think +2 for supportive population is more reasonable. The +10 gross income is a bit too low, I'm thinking +15.

Income and trade income overall should give 100% more.

Successfully updated karma reason!

admiralWillyWilber

Join Date 05/2009

+99

Since we are not talking about normal ai. we are talking about easier ai's then yes lowering picking the less preferable on lower levels than normal seems reasonable. I agree then for flavor they should pick different specializations if they can play them. Since the computer has not problem making decisions based on what choice it makes, or comparing values. There should be no reason that it have a different strategy based on choices. This takes more work writing more strategies. I think they should fix the specializations where none is optimal.

Successfully updated karma reason!

Yes, but isn't it almost inevitable in gaming that there will ultimately be One Best Way - ie with "FTL: Faster Than Light", missiles might be cool and everything but ultimately anyone who wanted to do well stuck with Lasers. How can developers create 3 pathways without - in some way - favouring one? (which, as I've posted previously, reflects the way they would play the game) Nerfing is probably the most likely option the developers would take, but making the huge advantage of Supportive Population a small advantage so the AI will still pick that...In other words, keeping the One Best Way, but trying to hide it so it's as unobvious as possible.

Successfully updated karma reason!

I've noticed that the AI's will start diversing their specializations when it comes to the later technologies, at least in my game, and they definately aren't only taking the best option every time. I've seen AI's choose the Sensor cost reduction branch, whilst some others opted for the mass reduction. I've also noticed it specifically with the precursor relic specilizations (which currently have no effect or am I mistaken?), e.g. some of them get the Cultural, whilst others get the Economic etc.

Successfully updated karma reason!

I have seen a bunch of what I consider non-optimal choices, or at least choices different than I would have made. I also watch for when they take optimal choices before I get to the branch. This lets me research the second best specialty and then trade for the better one. That makes me feel clever.

Stardock has spoken out against the concept that the AI behaves differently at different levels. They follow the paradigm that difficulty level adjustments should only be standard production/research/wealth/military bonuses. They have even taken out the FOW that had been implemented for higher difficulties. They seem to have a rather purist approach to AI. I appreciate that. If I were building an AI for this game, I would definitely be following the techniques you recommend, for both the lower levels and the upper levels. I would make the upper levels have more and more aggressive tendencies. It upsets the peaceful players such as myself, but if the nasty aliens aren't constantly pushing at you both economically and militarily, it is just not the same challenge. Also, it is still possible to have a mostly peaceful game. I don't think I would dumb down the lower level AIs, but I would let various faction flavor parameters affect the choices to a greater degree. On upper levels, they would be laser focused on which choice was best. And there would be the inevitable discussion on the forums disagreeing with those laser focused decisions.

On the other hand, I think the specialization techs are better balanced than people give credit for. Different games can call for different things. This last game I had some serious cash flow issues, mostly due to my choices, but I needed a solution. I went for a few key lower-the-maintenance techs and it really helped. I traded for a couple others and it built up in effect. So some things are situational and subtle and some things that are obvious to one person are obviously different to another. There may be some tweaking and polish that can be done based on these discussions, but so far, it seems to be working okay. At least, it seems that way to me.

Successfully updated karma reason!