civfan_999
Chieftain
- Joined
- Jun 23, 2020
- Messages
- 99
I like the other two punch list threads going and thought I'd start a new one for AI improvements.
1. Prioritize builders – It is insane how few tiles AI cities will have developed. They should attempt to develop all or some high percentage (80%?) of worked tiles and prioritize builders to that goal. They should particularly prioritize strategic and luxury resources.
2. Upgrade units – There will be AIs that are ahead of me on the tech tree that are still using units from eras before my units. They should always look to pay to upgrade units as soon as possible, starting with units on the front lines.
3. Value their great works – It costs thousands or tens of thousands of dollars to buy a great person, yet the AI is willing to sell a book or painting for usually around 150-300 gold. That does not make sense at all. Especially civs geared towards the culture victory should value their works much more highly. Maybe AIs that are not designed for a culture victory only have half or a third as much value as the culture victory civs, but still a lot more than 150$ for a book. AIs will not be willing to accept anything for a single coal or aluminum or oil. Why would they not have a similar regard towards great works. I think they should be willing to trade great works one for one (art for a sculpture, artifact for art, artifact for artifact) but they should not be willing to sell great works for such a pittance.
4. Aid requests – The AI spends WAY too much production on aid requests. I suggest that those civs designed for a diplomatic victory should do everything they can to win the 2 points, but other civs (designed to war, science, faith, or culture) should really be weighing the costs vs. benefits, i.e. not spend much production or even make the call to not participate at all. You could even have those civs vote down the aid emergency so that the diplomatic civ needs to burn more favor to have the opportunity to compete.
5. Emergencies – the AI shouldn’t vote for an emergency that they don’t intend to participate in. When an AI is neighboring a captured city state and then does nothing to try to retake it, then all they have accomplished is to make the aggressor more powerful. They should make the decision to either vote it down or vote it up but then prioritize victory.
6. Build anti-air and jet bombers – Required for defense against planes. Should be especially present on the front lines and should move back to other cities when a certain city is about to fall.
7. Value luxury resources in accordance with need – I think the AI is getting better at this. In same games, I will be able to sell something for 10GPT and then a second resource to the same civ with go for 4GTP, then a third may only get 1 gold total. But it doesn’t necessarily work the other way. Often, I can one to one a luxury resource and they will still give me 3-8 GPT on top of it. I think the AI should consider one luxury resource for one luxury resource to be a fair trade and not be willing to pay extra. That is, unless they are really struggling, i.e. cities with negative amenities, and they are willing to pay anything to fix that crisis.
8. Build less encampments? – the AI seems to build encampments in almost every city. This comes at the cost of other options, like science, culture, money, and production. It makes sense for military AIs to have lots of encampments and it certainly makes sense to build strategic encampments on the front lines, but the AI needs to better assess when encampments are valuable and when other districts would be more important.
-------- edited to add -----------
9. Difficulty Settings – I'd like to see them change the way they think about difficulty settings. I understand the struggle of writing better AI code to make them not suck at combat, but there can be improvements even to the current structure where basically some difficulties buff and others nerf the AI. Buff them a little at a time instead of a ton of buffs from the beginning of the game. The current difficulty settings give the AI a lot of perks from day one so that you spend the first few eras playing catch up. This makes it difficult to secure early game wonders and still gets you to that cross over point where once you have passed them, you can leave them in your dust. I would recommend giving the AI extra things each era rather than a lot of extras all at once at the beginning. For example, instead of starting them off with 5 civs and 5 techs (I might have those numbers a bit off), they could start the game with 1-2 civs/techs and get 1-2 additional civs/techs each time a new world era begins. That way, instead of us being behind early game and pushing to catch up, the AI wouldn’t be quite as strong in the ancient era, but would be better able to keep pace with an experienced human player throughout the game.
1. Prioritize builders – It is insane how few tiles AI cities will have developed. They should attempt to develop all or some high percentage (80%?) of worked tiles and prioritize builders to that goal. They should particularly prioritize strategic and luxury resources.
2. Upgrade units – There will be AIs that are ahead of me on the tech tree that are still using units from eras before my units. They should always look to pay to upgrade units as soon as possible, starting with units on the front lines.
3. Value their great works – It costs thousands or tens of thousands of dollars to buy a great person, yet the AI is willing to sell a book or painting for usually around 150-300 gold. That does not make sense at all. Especially civs geared towards the culture victory should value their works much more highly. Maybe AIs that are not designed for a culture victory only have half or a third as much value as the culture victory civs, but still a lot more than 150$ for a book. AIs will not be willing to accept anything for a single coal or aluminum or oil. Why would they not have a similar regard towards great works. I think they should be willing to trade great works one for one (art for a sculpture, artifact for art, artifact for artifact) but they should not be willing to sell great works for such a pittance.
4. Aid requests – The AI spends WAY too much production on aid requests. I suggest that those civs designed for a diplomatic victory should do everything they can to win the 2 points, but other civs (designed to war, science, faith, or culture) should really be weighing the costs vs. benefits, i.e. not spend much production or even make the call to not participate at all. You could even have those civs vote down the aid emergency so that the diplomatic civ needs to burn more favor to have the opportunity to compete.
5. Emergencies – the AI shouldn’t vote for an emergency that they don’t intend to participate in. When an AI is neighboring a captured city state and then does nothing to try to retake it, then all they have accomplished is to make the aggressor more powerful. They should make the decision to either vote it down or vote it up but then prioritize victory.
6. Build anti-air and jet bombers – Required for defense against planes. Should be especially present on the front lines and should move back to other cities when a certain city is about to fall.
7. Value luxury resources in accordance with need – I think the AI is getting better at this. In same games, I will be able to sell something for 10GPT and then a second resource to the same civ with go for 4GTP, then a third may only get 1 gold total. But it doesn’t necessarily work the other way. Often, I can one to one a luxury resource and they will still give me 3-8 GPT on top of it. I think the AI should consider one luxury resource for one luxury resource to be a fair trade and not be willing to pay extra. That is, unless they are really struggling, i.e. cities with negative amenities, and they are willing to pay anything to fix that crisis.
8. Build less encampments? – the AI seems to build encampments in almost every city. This comes at the cost of other options, like science, culture, money, and production. It makes sense for military AIs to have lots of encampments and it certainly makes sense to build strategic encampments on the front lines, but the AI needs to better assess when encampments are valuable and when other districts would be more important.
-------- edited to add -----------
9. Difficulty Settings – I'd like to see them change the way they think about difficulty settings. I understand the struggle of writing better AI code to make them not suck at combat, but there can be improvements even to the current structure where basically some difficulties buff and others nerf the AI. Buff them a little at a time instead of a ton of buffs from the beginning of the game. The current difficulty settings give the AI a lot of perks from day one so that you spend the first few eras playing catch up. This makes it difficult to secure early game wonders and still gets you to that cross over point where once you have passed them, you can leave them in your dust. I would recommend giving the AI extra things each era rather than a lot of extras all at once at the beginning. For example, instead of starting them off with 5 civs and 5 techs (I might have those numbers a bit off), they could start the game with 1-2 civs/techs and get 1-2 additional civs/techs each time a new world era begins. That way, instead of us being behind early game and pushing to catch up, the AI wouldn’t be quite as strong in the ancient era, but would be better able to keep pace with an experienced human player throughout the game.
Last edited:
