Pacifist AI game settings

[Eagle Pursuit]

I suspect that the nature of the AIs' pacifism may be related to the game settings, particularly map size and game speed.

So, for those of you who have experienced this, what are the particulars of the games where the AI is inordinately peaceful? Difficulty shouldn't matter because you are saying it's too peaceful compared to G&K's games of the same difficulty.

For me, It is Huge, Continents, and Marathon speed.

 

[Buccaneer]

Marathon = -2 difficulty if you want a direct comparison

The game has always been most balanced at standard Continents and the like, if I recall.

 

[Eagle Pursuit]

Marathon = -2 difficulty if you want a direct comparison

The game has always been most balanced at standard Continents and the like, if I recall.

For me that is irrelevant , because those are the settings I've played on in vanilla and G&K and now the AI is kittens compared to before. They don't even attack each other.

 

[Ulthwithian]

Marathon = -2 difficulty if you want a direct comparison

The game has always been most balanced at standard Continents and the like, if I recall.

A proper analysis will give those factor coefficients without having to do it by hand.

I am VERY interested in this information. I would be more than willing to do regression or other analysis to try to get the values here.

 

[OmniPotent42]

A proper analysis will give those factor coefficients without having to do it by hand.

I am VERY interested in this information. I would be more than willing to do regression or other analysis to try to get the values here.

The reason that Marathon is -2 difficulty is because military conquest doesn't scale down with speed, so warfare is much easier and smaller forces can take out cities before they can make reinforcements.

 

[Ulthwithian]

The reason that Marathon is -2 difficulty is because military conquest doesn't scale down with speed, so warfare is much easier and smaller forces can take out cities before they can make reinforcements.

I'm not saying it isn't Omni. My point is that if you give as inputs the data Eagle asked about, and you output a probability of winning the game, a proper regression analysis will reveal the different coefficients.

Let's say you have a regression equation that takes as inputs the map parameters and game speed and even which AIs you're fighting against. The output is the probability of winning the game, or, if you like, the 'normalized difficulty level' of that game. If what you say is true, the variable that denotes Marathon speed would have a -2 next to it, if Standard were 0.

You could also do this with ANNs.

 

[Buccaneer]

Yes and a comparison to a marathon emperor game would be a standard prince game, as well as a marathon emperor game.

 

[Ulthwithian]

Well, I would rather have that as an output of the analysis, rather than as an input.

E.g., I would expect an interaction effect between map size and speed. Correct me if I'm wrong in that the differences between difficulties is also impacted by map size. That is, the difference between a Marathon-Emperor-Duel game's difficulty and a Standard-Emperor-Duel game's difficulty is much less than a Marathon-Emperor-Huge game's difficulty and a Standard-Emperor-Huge game's difficulty.

 

[Buccaneer]

I would think so but not all standard Small Continents (for example) are created equal, so that can't be a constant. Also placements on said map can make or break a civ. We do know about scaling of units, though.

 

[Ulthwithian]

That's where there's going to be noise in the data, Buccaneer, which again the regression can handle.

The issue will be getting a dataset in a large enough size to do a useful analysis.

 

[Svest]

The level of challenge to the player based on the different map settings is completely irrelevant to what the OP is asking. The question being asked is not what settings make the game harder or easier. It is if certain settings make the AI less likely to declare war.

To the OP, I've played 2 games, both on standard, continents, immortal. The first one seemed fairly normal, I didn't get involved in many wars, but the AI fought against each other quite a bit. In the second game I am in the Industrial era and I don't think there has been a single war in the entire world yet.

 

[Buccaneer]

Just making the point about comparisons as the OP asked. Marathon can invalidate any comparisons except to other marathon games and standard games -2 levels below X.

 

[Ulthwithian]

Just making the point about comparisons as the OP asked. Marathon can invalidate any comparisons except to other marathon games and standard games -2 levels below X.

*sigh*

Again, with the proper analytical techniques, this is not true. And I for one am perfectly willing to do that analysis.

It may be the only way to get enough information to do a proper analysis.

 

[omnishakira]

The reason that Marathon is -2 difficulty is because military conquest doesn't scale down with speed, so warfare is much easier and smaller forces can take out cities before they can make reinforcements.

actually its the opposite.

warfare is much harder and unforgiving on marathon which is why the AI is doing bad on marathon. the faster the game, the easier it is and that's why the AI do better on faster games

and that X-2 thing is only true for wars, the rest is the same which means that overall, you probably need to substract like X-0,5 or something

 

[Svest]

Just making the point about comparisons as the OP asked. Marathon can invalidate any comparisons except to other marathon games and standard games -2 levels below X.

No, actually comparing AI behavior of a marathon game to a standard game -2 levels would be a terrible idea. Again, just because a marathon game is similar in challenge to a standard game -2 levels doesn't mean that the AI will behave remotely the same way in the two. The challenge posed to the player is completely irrelevant in this case.

If this were about what strategies to use to win the game you might have a point.

 

[Buccaneer]

I don't disagree with that. There are base interactions with the AI at marathon (warfare being one) that makes any comparisons skewed, unless identically set up games. It's like any strategy discussions at marathon are pretty much ignored. But I'll agree that comparing standard game (even at -2) to marathon would not be fruitful. I just wish the OP did not start by using a marathon example.

 

[seasnake]

Playing Huge Map, Continents Plus with Epic Speed, 2 additional city-states and 1 additional civ. Crowds things up a little, let's civs like Austria and Venice flex some UA muscle while leaving enough for the rest of us. King difficulty.

Every time, every civ, a perfect game. Current game I'm in a war where four of use are tackling Napoleon in the Renaissance, only I'm trying to do it solely to Recall to Life Willem so I can own his vote (I'm the head of the World Congress) and get most favored trade nation status.

12 Civs on a Huge Map and you have civs that can settle 30 cities without offending or encroaching. You need the extra civ, definitely.

 

[Svest]

I don't disagree with that. There are base interactions with the AI at marathon (warfare being one) that makes any comparisons skewed, unless identically set up games. It's like any strategy discussions at marathon are pretty much ignored. But I'll agree that comparing standard game (even at -2) to marathon would not be fruitful. I just wish the OP did not start by using a marathon example.

It doesn't matter what the OP started with. It's just a single datapoint. All he is asking is for people to list the game settings they have played so far and say if the AI was passive or not. With a big enough sample size we can determine if specific settings or combinations of settings increase the odds of the AI being passive.

 

[Eagle Pursuit]

I don't disagree with that. There are base interactions with the AI at marathon (warfare being one) that makes any comparisons skewed, unless identically set up games. It's like any strategy discussions at marathon are pretty much ignored. But I'll agree that comparing standard game (even at -2) to marathon would not be fruitful. I just wish the OP did not start by using a marathon example.

I was and am comparing Marathon BNW games to Marathon GNK games. Compared to Marathon games in GNK, the AI in Marathon games in BNW is kittens.

I want all submitters to tell their settings as comparing their BNW to their GNK games with the same settings.

Until you FUBARed the thread by focusing on marathon.

 

[Svest]

Playing Huge Map, Continents Plus with Epic Speed, 2 additional city-states and 1 additional civ. Crowds things up a little, let's civs like Austria and Venice flex some UA muscle while leaving enough for the rest of us. King difficulty.

Every time, every civ, a perfect game. Current game I'm in a war where four of use are tackling Napoleon in the Renaissance, only I'm trying to do it solely to Recall to Life Willem so I can own his vote (I'm the head of the World Congress) and get most favored trade nation status.

12 Civs on a Huge Map and you have civs that can settle 30 cities without offending or encroaching. You need the extra civ, definitely.

I suspect this is getting to it. Crowd the game up a bit and the AI is more willing to fight early. My guess is because this will give them somewhere else to send their trade routes. I really think the AI are putting too much value on trade routes early in the game. They won't attack a neighbor unless they can send trade routes somewhere else.