AI in v2.6 beta0

Can't beat the AI? Is it too tame? Discuss the best strategies!
Post Reply
bard
Veteran
Posts: 121
Joined: Fri Jun 14, 2013 2:00 pm

AI in v2.6 beta0

Post by bard »

I have played a couple of games against the AI with v2.6 beta0 (civ2civ3 rules, and all AI traits to 50), and I have noticed nice improvements compared to previous versions:

1) The AIs expand pretty fast. They research triremes quickly and use to colonize void continents faster than me. They do not protect settlers or newly created cities very well, but I think it is worth if this way they achieve such fast expansion. I just suggest that they try to send at least one defensive unit to new continents, else it is easy to use cavalry units to destroy all their newly created cities before they can build defenders. When fighting for a new continent, just one city with a defender that survives these rush attacks can change everything.

2) I like how isolated cities protect themselves before they reach size 2, so it is not easy to capture them.

3) The AIs sometimes perform naval invasions near your weakest cities, so you can no longer leave your cities completely undefended.
However, they still do not take full advantage of marine units to attack weak cities directly from boats.

4) They build wonders in an effective way. At least for me, there is a fun race now to try to build the wonders first. There was a time where I was able to build them all, now they use to build several wonders first.

5) While at peace, the AIs are able to develop their cities with most improvements. They seem to build now harbours, granaries and even supermarkets.
However, while at war, they keep building only military improvements, and wonders. I think they should think a bit more about the future, even when losing a war, as they do with wonders. Imo, if the cost to survive is to stop building improvements and stop researching, then it is not worth because you will have no chances to win the game anyway.

6) AIs research faster than previous versions. It seems they do not forget about science rates as much as they used to do.


The main weaknesses that I still see:

1) It is still too easy to stop most AI attacks by placing a line of defensive units that block the pass due to ZoC rules. In my latests game, the AI was able to cross such lines a couple of times by following the path of his own diplomats/explorers, but I think it was by luck.
If they were able to send IgZOC units on purpose to cross ZOC tiles, they would be more dangerous.

2) AI does not seem to take into account the danger of Kill Stack rule. They create large stacks of units that can be killed by one single attack.

3) Their attempts to attack unprotected workers are too predictable and mostly useless, since the attacking unit (usually more expensive) use to die next turn too.
I think the AI should stop trying to kill unprotected units in enemy territory, and focus on trying to conquer cities. I mean, the only reason to enter enemy territory should be to try to conquer a city, and then, once there, it is ok to kill unprotected units if they want. But there is no reason to send full stacks of units to enemy territory just because there is one single unprotected worker, as current AI use to do. It is not effective against veteran players, nor fun for new players.

Knowing all these weaknesses, it is easy to make a trap where there is only one path from AI territory to a tile where you place an unprotected worker, and you can easely kill full stacks of incoming units.

Anyway, I just wanted to say thank you for your work improving the AI, I'm enjoying my games against the AI more with every new freeciv version.
User avatar
dunnoob
Elite
Posts: 327
Joined: Mon Dec 23, 2013 3:13 am
Location: Hamburg
Contact:

Re: AI in v2.6 beta0

Post by dunnoob »

bard wrote:2) AI does not seem to take into account the danger of Kill Stack rule. They create large stacks of units that can be killed by one single attack.
+1, but certainly no new 2.6.0-beta0 issue, also not better in beta1.
3) Their attempts to attack unprotected workers are too predictable and mostly useless, since the attacking unit (usually more expensive) use to die next turn too.
Depends, suicide attacks of paratroopers against engineers, i.e., a dead engineer and a dead paratrooper in the next turn, force me to protect engineers with a mech. inf. (or armor), or at least a marine (but that might be not good enough). I haven't seen that in beta0 + beta1, maybe because my ruleset is tuned against silly AIs (= it's their fault, they should be smarter. Actually I think a "real" AI should be distributed as a stripped down client and run as a separate process, like a human, but no graphics and no music. :lol: )
Ignatus
Elite
Posts: 644
Joined: Mon Nov 06, 2017 12:05 pm
Location: St.Petersburg, Russia
Contact:

Re: AI in v2.6 beta0

Post by Ignatus »

Really, the AIs were good in building wonders in beta0? In beta2, when playing civ2civ3, they hardly can finish building one, even large civilizations with many cities far from any enemy just start and stop in few turns. Maybe they don't on maximal difficulcy?
bard
Veteran
Posts: 121
Joined: Fri Jun 14, 2013 2:00 pm

Re: AI in v2.6 beta0

Post by bard »

I use to play on cheating difficulty where AI gets heavy bonuses to production, and I place all AIs in the same team so they never fight each other, only against me.
There was a time where AI was not challenging for me, even with those big advantages. The comments in my first post were improvements I noticed compared to games in previous versions. However, if you play normal difficulty or you allow AIs to fight each other, you might not notice much improvements.
What I meant was that AI in v2.6 seems to take better advantage of the bonuses from cheating AI level, at least compared to previous versions, and now it is possible for me to mod these bonuses until it is somehow challenging to defeat the AI. But I guess one single AI in normal level will always be far from challenging.
Post Reply