Aggressive AI

[Gazebo]

Fair enough, but this is exactly why this topic has been (and continues to be) so popular: it would simply be amazing if the AI was capable of better judging the skill of the player by observing how he conducts himself in wars such that it would overestimate an average player and underestimate an above average player both to a lesser degree, instead of being more static and expecting an "average player".
I get that this might be too much to ask for, though, and in the end it is you who has the deepest insight in the actual code and what is feasible but a more dynamic appreciation of player combat efficacy and efficiency is always gonna be a good thing; not that there haven't been huge strides in improving the AI warring ability already.

And what I keep saying, over and over and over and over again, is that the solutions presented here wouldn’t do anything meaningful. Handicaps exist to make an AI built around medians and averages more difficult for players that are above average.

G

 

[CppMaster]

@Gazebo Is it feasible if AI could estimate player's skill by his posts on this forum? You can associate a user with local machine by comparing IP addresses. Then you just run user's posts with a NLP model to assess his skill. Of course liked posts would improve user's score.
If not, then I would be ok with just local deep network.

 

[Jdoug312]

And what I keep saying, over and over and over and over again, is that the solutions presented here wouldn’t do anything meaningful. Handicaps exist to make an AI built around medians and averages more difficult for players that are above average.

G

I get that you probably don't want to hear it, but as I said in my last post, your retorts are ambiguous on a topic that people obviously have a sustained interest in.

Your response reads a bit ambiguously to me even accounting for your bolding the font. My first inclination is to interpret your reply as you stressing that there isn't a discrepancy between the AI's calculations relative to the military score that the eui shows, and that just raising the military score itself wouldn't change the way the AI views the situation. If so, then my question is modified:

In a good 95% of my games, I don't neglect my military. I hover right around my cap, and I'm the type to prioritize the military techs over the others more often than not. So in situations where I'm the tech leader or close to it, that means that I almost always have the more advanced units as well. Why then are there scenarios where I have ~300 less military score than a civ with a similar number of cities and a more outdated army? (I know this score isn't important but it's relatively accurate to AI calculations right?)

-------

Perhaps, however your statement was intended to stress the huge part, not the entire clause. In that scenario, I'd interpret it as you meaning that the score and the calculations are still relative, but that the numbers don't translate to a massive shift in AI decision-making.

So in a case where I have 300 military score and an AI has 600, they aren't thinking "oh I'm twice as strong as them, let me take advantage of this" without other factors outside of those calculations first coming into play.

-------

Im sure there's other ways to interpret your statement as well and it's possible I misinterpreted your ultimate point in this discussion. Also I'm sure specifics examples/numbers from specific games would be more helpful than throwing an arbitrary 300 military score difference out there; with the next newest version, I will be able to provide specifics.

 

[Gazebo]

I get that you probably don't want to hear it, but as I said in my last post, your retorts are ambiguous on a topic that people obviously have a sustained interest in.

You're getting caught up in semantics. Let me be clear:

1.) The AI cares about relative military power, and uses a function that compares the two parties (strength, city strength, unit power, etc.) to weigh relative power. Handicap does not factor into this model at all.
2.) The AI does not care about the 'score' number that appears in game.
3.) Adjusting military power based on performance would have a negligible impact on AI decision-making.

G

 

[Aristos]

3.) Adjusting military power based on performance would have a negligible impact on AI decision-making.

G

Not only that, but I can foresee a very clever player "simulating dumb moves" in order to trick such AI into underestimating him...

(and then the complaints will come about how easy it is to fool the AI)

 

[ElliotS]

3.) Adjusting military power based on performance would have a negligible impact on AI decision-making.

Why is military strength such a small factor? I feel like in a game with 8 real players military strength would be just behind logistics (Can I physically attack them?) in my decision making when it comes to war?

Does putting more focus on it actually hurt weaker civs? Why is it like this?

Not only that, but I can foresee a very clever player "simulating dumb moves" in order to trick such AI into underestimating him...

(and then the complaints will come about how easy it is to fool the AI)

Purposefully losing a war to trick the AI into losing a war? Sounds like a bad trade to me.

 

[Aristos]

Purposefully losing a war to trick the AI into losing a war? Sounds like a bad trade to me.

Who said anything about losing the war?

 

[John.B.C]

Not only that, but I can foresee a very clever player "simulating dumb moves" in order to trick such AI into underestimating him...

(and then the complaints will come about how easy it is to fool the AI)

I often try to lure the AI into traps during war, putting workers in positions up for grab. More often than not, he doesn't go for the bait. But still often enough for me to keep doing it.

 

[BiteInTheMark]

Who said anything about losing the war?

So, pls inform us, how do you would "trick" the AI with bad moves? I dont think the AI even knows what a "bad" move is (if your definition is, moving units into bad position). AI only cares about killed and lost units, cities, warmongering and warscore. If you lose units on purpose or lose cities on purpose, this sounds like a very bad strategy to "fool" the AI, cause you do excatly, what AI wants.

 

[Gazebo]

Why is military strength such a small factor? I feel like in a game with 8 real players military strength would be just behind logistics (Can I physically attack them?) in my decision making when it comes to war?

Does putting more focus on it actually hurt weaker civs? Why is it like this?



Purposefully losing a war to trick the AI into losing a war? Sounds like a bad trade to me.

Because it is a small part of AI approach. We’ve discussed this. Round and round.

G

 

[Aristos]

So, pls inform us, how do you would "trick" the AI with bad moves? I dont think the AI even knows what a "bad" move is (if your definition is, moving units into bad position). AI only cares about killed and lost units, cities, warmongering and warscore. If you lose units on purpose or lose cities on purpose, this sounds like a very bad strategy to "fool" the AI, cause you do excatly, what AI wants.

Please take the time to read the post to which I responded, and any relevant post that lead to it, in order to be able to respond within context.

 

[ElliotS]

Because it is a small part of AI approach. We’ve discussed this. Round and round.

G

I understand that. Not arguing. I'm asking why the AI cares so little about relative military score in that case. In my head I would think everything is based off of that, so I'm wondering why it's different.

 

[Gazebo]

I understand that. Not arguing. I'm asking why the AI cares so little about relative military score in that case. In my head I would think everything is based off of that, so I'm wondering why it's different.

It's not that they care 'so little', but rather it's a large cog alongside other large cogs.

G

 

[tu_79]

It's not that they care 'so little', but rather it's a large cog alongside other large cogs.

G

But sure there are some ways to make some leaders to behave more aggressively. Some leaders are more aggressive, aren't they?

Ideally, AI should be able to decide when it is better to be more or less aggressive against the human player, but it would do if we can force AI to chill when setting up a new game.

This way we won't have players with balanced skills complaining that AI is too peaceful.

 

[Gazebo]

But sure there are some ways to make some leaders to behave more aggressively. Some leaders are more aggressive, aren't they?

Ideally, AI should be able to decide when it is better to be more or less aggressive against the human player, but it would do if we can force AI to chill when setting up a new game.

This way we won't have players with balanced skills complaining that AI is too peaceful.

Already works that way.

G

 

[tu_79]

Already works that way.

G

??

We can already tell AI to chill out? How?

 

[Gazebo]

??

We can already tell AI to chill out? How?

Not the latter part as you say. But if you turn off domination victory the AI is less aggressive.

G

 

[tu_79]

Not the latter part as you say. But if you turn off domination victory the AI is less aggressive.

G

Interesting. It's good to have this as option for those who complain.

 

[Recursive]

This is what I’m trying to say: any of the changes mentioned would be like adding wind to an already-billowing sail. It just won’t have much impact.

That DOES NOT mean I’m not looking at making some changes - rather, it just means that I’m not approaching it this way.

G

I'm surprised no one's asked this yet, but uh...in what way are you approaching it?

 

[Gazebo]

I'm surprised no one's asked this yet, but uh...in what way are you approaching it?

Tracking DOWs backwards and tracing the behavior to see if it meshes with the civ's actual reality.

G