Arkatakor
King
As it has been decided that AND AI needed to be made "leaner and meaner" I would like to make some suggestions that stem from 100's of RoM + AND games for how to improve the AI:
Combat:
BetterAI makes strange combat decisions as opposed to Vanilla AI. I will illustrate this with two examples:
1) Sacrificing overwhelming units to win a minor battle:
A typical example is attacking a stack of human units which are in tiles with significant defensive bonuses resulting in heavy unit losses for the AI. I imagine the justification behind such behavior is that BetterAI calculates that with enough units sacrificed, it will win the battle and destroy the human stack of units, albeit with great losses. This is faulty reasoning because winning wars is more important than winning battles, and if the AI loses too many units, it ends up losing the war. Vanilla AI never did this.
2) Missing Opporitunties to Capture Cities:
I have been in quite a lot of situations where the AI declares war on me, and within 2 or 3 turns has a huge stack right outside a poorly defended city. Its forces are so superior both in strength and in number that if it decides to invade the city, its a sure win. Yet it decides not to do so. Instead it chooses to either wait or walk around my territory with its huge stack, giving me ample time to prepare my defenses in the cities that it marches towards. Again Vanilla AI never did this. If Vanilla AI knew it had a decent chance of capturing a city it would do so without fail. This is by far the most pressing issue and needs to be fixed at all costs.
Diplomacy
1) AI not understanding that its diplomatic status / decisions makes its citizens unhappy
This is the only bug / exploit that extends to BOTH Vannila AI and BetterAI. If an AI is at war with one or more civ's an easy exploit to cripple it is to send in a few spies to alter its civics to democracy / liberal at a VERY cheap espionage cost with high success rate. This effectively cripples the AI, because as a result of war weariness, its citizens become WILDLY unhappy.
I will illustrate the above with a recent example. An aggressive AI (Shaka) was in the lead, exceeding the 2nd ranked player (myself) by a good 20% in points. He had some 17 cities while I had 9. It was the industrial age, and my Zulu friend had declared war on multiple nations. His army was by far the largest and most powerful as he was crushing the minor civs. Shaka was also techning like mad (3 turns for biology which would take me 7 turns). It was take him down NOW or lose. I tried preparing an all out assault with many canons, a good stack of machine guns and lots of riflemen. Just before declaring war I saved. I fought vehemently for 15 turns and it was clear that altough I captured a couple of cities I was outmatched. My troops were falling apart from the overwhelming opposition. All the while my Zulu friend was out techning me like mad, pacing ahead. It was a lost cause, so I was about to give up that game.
Then after a few hours break, an idea came into my head - going all out on espionage. And so I did - I reloaded my save from before the war, built and sent a stack of 10 spies into one of his cities. A few turns later I went ahead and changed his civics to Democracy / Liberal while his declared war with the other 2 civs was ongoing. As I had racked up so many espionage points against him, I was able to see into his cities. Basically more than 50% of his citizens were unhappy due to war weariness. Shaka's cities were shrinking like mad, his ability to produce units / buildings was out the window, and his tech research slowed down to near stagnation (13 turns for biology as opposed to 3). All the while Shaka REFUSED to make peace with the 2 nations he had declared war on (he wante me to bribe him to do so). He even declared war on ME a few turns later while still at war with the other AI's. This time round I was able to fend off his troops easily as he was not able to replace them due to slow unit production pace. Needless to say his citizens war weariness as a result of a declaration of war against 3 civs crippled him completely. Even when all the wars ended, Shaka never recovered after that. His cities had lost more than half their size and by the time they grew back to normal size again, I was way ahead of him and stayed so indefinitely.
Thus, I would propose that the AI makes a CHECK to see if its citizens are unhappy because of its diplomatic (war) status and does another check to see if it has the option to declare peace and do so accordingly, should the option be available. I will also make a separate proposal in another thread so as to keep this conversation focused, about racking up costs for changing civics to a civ during war time. In the meanwhile modifying this behavioral change would greatly benefit the game in my opinion.
2) Not understanding that its losing the game:
This would be by far the most difficult to implement properly and is thus just a proposal. I would suggest that it only apply with Ruthless AI enabled if implemented:
Lets take an example where one Civ is way ahead of the other players, again by a good 30% or more. The leading civ has been ahead for maaaany turns and is leading in everything by far. What I would propose is that all the AI ranked 2nd or below recognize whats going on, and say to themselves, "hey we are losing". They do a feasibilty check to see if the troop power of all Civs combined would be enough to take down the top player. If thats the case, they decide to ALL declare war on the top player. And by declaring war, I mean a coordinated effort - each and every AI marching troops up to the borders of the top player in question and THEN declaring war such that the invasion is overwhelming for the top player.
For this to be implemented an algorthim would have to be created:
1) When a civ is ahead of all other civs by 20% or more then it stats to do a check for coordinating war efforts.
2) After, say 50-70 turns on normal speed (the amount of turns can be randomized), if the top player is still ahead of all other civs by more than 20%, they do a feasibility check to see if the troop power of all Civs combined would be enough to take down the top player.
3) If they can take out the top player, the AI then has the option to choose this war path, but will not necessarily do so. Having a predictable AI is the last thing we want. Maybe other players have suggestions on how this could be implemented. Again this, would only apply with the "Ruthless AI" option enabled.
Will be looking forward to seeing feedback on this thread
Feedback And Further Suggestions:
1) Suggestion from Aline: Avoid having AI declare super early wars that it is unequipped for.
2) Suggestion from IPEX-731BA5DD06: The AI MUST use Great Generals and field Commanders offensively and properly.
3) Suggestion from IPEX-731BA5DD06: Limit the number of civic changes the AI makes. It keeps switching back and forth and never tries to capitalize on golden ages in order to make civic switches. Having the AI going thru prolonged periods of anarchy effectively makes the AI fall behind in virtually category.
Combat:
BetterAI makes strange combat decisions as opposed to Vanilla AI. I will illustrate this with two examples:
1) Sacrificing overwhelming units to win a minor battle:
A typical example is attacking a stack of human units which are in tiles with significant defensive bonuses resulting in heavy unit losses for the AI. I imagine the justification behind such behavior is that BetterAI calculates that with enough units sacrificed, it will win the battle and destroy the human stack of units, albeit with great losses. This is faulty reasoning because winning wars is more important than winning battles, and if the AI loses too many units, it ends up losing the war. Vanilla AI never did this.
2) Missing Opporitunties to Capture Cities:
I have been in quite a lot of situations where the AI declares war on me, and within 2 or 3 turns has a huge stack right outside a poorly defended city. Its forces are so superior both in strength and in number that if it decides to invade the city, its a sure win. Yet it decides not to do so. Instead it chooses to either wait or walk around my territory with its huge stack, giving me ample time to prepare my defenses in the cities that it marches towards. Again Vanilla AI never did this. If Vanilla AI knew it had a decent chance of capturing a city it would do so without fail. This is by far the most pressing issue and needs to be fixed at all costs.
Diplomacy
1) AI not understanding that its diplomatic status / decisions makes its citizens unhappy
This is the only bug / exploit that extends to BOTH Vannila AI and BetterAI. If an AI is at war with one or more civ's an easy exploit to cripple it is to send in a few spies to alter its civics to democracy / liberal at a VERY cheap espionage cost with high success rate. This effectively cripples the AI, because as a result of war weariness, its citizens become WILDLY unhappy.
I will illustrate the above with a recent example. An aggressive AI (Shaka) was in the lead, exceeding the 2nd ranked player (myself) by a good 20% in points. He had some 17 cities while I had 9. It was the industrial age, and my Zulu friend had declared war on multiple nations. His army was by far the largest and most powerful as he was crushing the minor civs. Shaka was also techning like mad (3 turns for biology which would take me 7 turns). It was take him down NOW or lose. I tried preparing an all out assault with many canons, a good stack of machine guns and lots of riflemen. Just before declaring war I saved. I fought vehemently for 15 turns and it was clear that altough I captured a couple of cities I was outmatched. My troops were falling apart from the overwhelming opposition. All the while my Zulu friend was out techning me like mad, pacing ahead. It was a lost cause, so I was about to give up that game.
Then after a few hours break, an idea came into my head - going all out on espionage. And so I did - I reloaded my save from before the war, built and sent a stack of 10 spies into one of his cities. A few turns later I went ahead and changed his civics to Democracy / Liberal while his declared war with the other 2 civs was ongoing. As I had racked up so many espionage points against him, I was able to see into his cities. Basically more than 50% of his citizens were unhappy due to war weariness. Shaka's cities were shrinking like mad, his ability to produce units / buildings was out the window, and his tech research slowed down to near stagnation (13 turns for biology as opposed to 3). All the while Shaka REFUSED to make peace with the 2 nations he had declared war on (he wante me to bribe him to do so). He even declared war on ME a few turns later while still at war with the other AI's. This time round I was able to fend off his troops easily as he was not able to replace them due to slow unit production pace. Needless to say his citizens war weariness as a result of a declaration of war against 3 civs crippled him completely. Even when all the wars ended, Shaka never recovered after that. His cities had lost more than half their size and by the time they grew back to normal size again, I was way ahead of him and stayed so indefinitely.
Thus, I would propose that the AI makes a CHECK to see if its citizens are unhappy because of its diplomatic (war) status and does another check to see if it has the option to declare peace and do so accordingly, should the option be available. I will also make a separate proposal in another thread so as to keep this conversation focused, about racking up costs for changing civics to a civ during war time. In the meanwhile modifying this behavioral change would greatly benefit the game in my opinion.
2) Not understanding that its losing the game:
This would be by far the most difficult to implement properly and is thus just a proposal. I would suggest that it only apply with Ruthless AI enabled if implemented:
Lets take an example where one Civ is way ahead of the other players, again by a good 30% or more. The leading civ has been ahead for maaaany turns and is leading in everything by far. What I would propose is that all the AI ranked 2nd or below recognize whats going on, and say to themselves, "hey we are losing". They do a feasibilty check to see if the troop power of all Civs combined would be enough to take down the top player. If thats the case, they decide to ALL declare war on the top player. And by declaring war, I mean a coordinated effort - each and every AI marching troops up to the borders of the top player in question and THEN declaring war such that the invasion is overwhelming for the top player.
For this to be implemented an algorthim would have to be created:
1) When a civ is ahead of all other civs by 20% or more then it stats to do a check for coordinating war efforts.
2) After, say 50-70 turns on normal speed (the amount of turns can be randomized), if the top player is still ahead of all other civs by more than 20%, they do a feasibility check to see if the troop power of all Civs combined would be enough to take down the top player.
3) If they can take out the top player, the AI then has the option to choose this war path, but will not necessarily do so. Having a predictable AI is the last thing we want. Maybe other players have suggestions on how this could be implemented. Again this, would only apply with the "Ruthless AI" option enabled.
Will be looking forward to seeing feedback on this thread
Feedback And Further Suggestions:
1) Suggestion from Aline: Avoid having AI declare super early wars that it is unequipped for.
2) Suggestion from IPEX-731BA5DD06: The AI MUST use Great Generals and field Commanders offensively and properly.
3) Suggestion from IPEX-731BA5DD06: Limit the number of civic changes the AI makes. It keeps switching back and forth and never tries to capitalize on golden ages in order to make civic switches. Having the AI going thru prolonged periods of anarchy effectively makes the AI fall behind in virtually category.
