What Can We Do by Getting Opponents to Fight?

[Spoonwood]

Suede made an interesting video on using debut to observe AI behavior (there's probably some old thread which talks about someone who did this before... I'll have to look). I hypothesized that the correlation between AI move order would be from top to bottom according to the spaceship launch screen. I was wrong. Fortunately though after for playing for a few more turns, it appears that there exists an exact correlation between being on the bottom of the spaceship launch screen and moving first in the game. That's the most generally applicable thing I have to say.

I next hypothesized that the spaceship launch screen and the order of opponents when you preselect them would be in the exact same order. I confirmed this. Thus, one can effectively preselect the order in which your opponents move. The human player always moves first (thus Regent wouldn't be exactly even if the AIs were as intelligent as the human player... the human player would still have a slight advantage).

There's an old succession game that I think Sirian was in that I recall reading where the team made it so that one of the AIs would win. So, some of the knowledge here again, might, lie somewhere around here.

But, I still have some questions and thoughts. Let's suppose that unique units don't exist. All things else being equal, if two AIs who fight, if attacking is strong enough, the AI at the bottom of the list will win. But, if defense is stronger (and often it is), the defender might win the war by defending against the other AIs units and counter-attacking their attackers more effectively. So, is this correct, if you want an AI to win a war, in general and all things else equal (they aren't due to golden ages, unique units, resource availability, and production capability) you want them at the top of the preselection list.

More interesting to me though comes, can things get influenced by preselection of opponents so that the AIs can stalemate in a war and just end up losing units?

Also, how can preselection of opponents get used to make it easier to fight a war with help from one AI against another? With unique units in mind, would we ideally want India on the bottom to get the most help in weakening some AI we're at war with when out opponent has middle age units?

How powerful can preselection of opponents be in dividing up AIs?

 

[need my speed]

But, I still have some questions and thoughts. Let's suppose that unique units don't exist. All things else being equal, if two AIs who fight, if attacking is strong enough, the AI at the bottom of the list will win. But, if defense is stronger (and often it is), the defender might win the war by defending against the other AIs units and counter-attacking their attackers more effectively. So, is this correct, if you want an AI to win a war, in general and all things else equal (they aren't due to golden ages, unique units, resource availability, and production capability) you want them at the top of the preselection list.

Why? I don't understand this. The combat odds and the terrain of the cities (influencing production influencing the amount and quality - technologies - of units), are all irrelevant to the turn order of AIs, right?

 

[Spoonwood]

Why? I don't understand this. The combat odds and the terrain of the cities (influencing production influencing the amount and quality - technologies - of units), are all irrelevant to the turn order of AIs, right?

In terms of how they got programmed, I would guess them irrelevant. In terms of game effects, I think there exists an interaction with respect to who does well in that war.

Let's say that you have civ A attacking civ B. Civ A has strong attackers and move first. Civ A could do well against civ B, if they fought an aggressive war well. Most AIs don't usually, but if civ A has cavalry and the other doesn't have saltpeter or iron, and it's before the industrial age, civ A will likely crush civ B and probably do even better against civ B if they move first.

Also, let's say that civ C attacks civ B. But, in this example, civ B moves first. Now, civ C will more likely end up chasing after civ B's weak units. A redlined infantry unit? Civ c will try to chase that thing down usually instead of attacking a closer city. So, civ C's attack ends up weaker in terms of attacking cities since they chase down units more. But, civ C, moving second, probably destroys more of civ B's units earlier than civ A will, given them as equivalent.

At least I think the following: If the AI fights wars with each other, the civ that moves second will tend to do better than the first, given all other things as equal, because of the combat odds and the terrain of cities make defenders more powerful than attackers give that equivalent numbers of them fight (hence when you fight an AI bringing overwhelming numbers ends up the time-honored and optimal strategy). But, I don't have data to back this up, so I could have missed something. Things like golden ages and resource differences can also make it not exactly straightforward to figure out, I think. And also, agricultural AI will tend to have more units early than non-agricultural AIs.

 

[tjs282]

I next hypothesized that the spaceship launch screen and the order of opponents when you preselect them would be in the exact same order. I confirmed this. Thus, one can effectively preselect the order in which your opponents move.

I usually select All-Random opponents (without Culturally-linked starts) on the 'Choose-Your-Civ' startup screen, so I've never really examined if/how the selection-order of AI-opponents on that screen affects the turn-order in-game (although I presume that it does).

BUT

I am 99.9% certain that the listing-order of the AI-Civs (which is initially generated by the pRNG) in CivAssist II, on e.g. the 'Potential Trades' screen, does correspond exactly to the AI's play-order over the IBT, i.e. if the human is Player 1, then the AI-Civ at the top of CAII's list is Player 2, the AI-Civ in second place is Player 3, etc. So (if you're still using CAII yourself?) it should be easy to test how the 'Choose Your Opponents' boxes affect the order of play in-game, by selecting, say, a Large or Huge World, inputting a specific arrangement of the 11-15 Civs into the 2-3 columns of boxes, and then comparing how that specific row/column order gets imported into CAII, i.e. which play-order it corresponds to.

So if I have understood your postulate correctly(?), that playing earlier in the IBT confers a general-if-slight advantage -- which seems reasonable* -- then it should certainly be possible to order the 'stronger' and 'weaker' AI-Civs as you prefer, for any effect you might want to achieve

Spoiler * :

Since AI-Civs tend to first-strike DoW only when they are already stronger than their target -- or at least, 'think' they are -- it seems likely that the Civ which takes the initiative and strikes first will tend to win more often -- so long as their target isn't e.g. Sumer, Greece, Carthage, Netherlands or France...

e.g. if you were going for Space, and wanted to give some/all of the SCI-Civs a better chance of survival into the late game -- however marginal -- to be able to harvest their free-techs, then you could assign them to the slots at/near the the top of the play-order.

 

[lohaller]

Interssting read, but i am thinking of one aspact that might have been overlooked. But i am new to civ 3 so correct me if i am wrong, playing at emperor diff atm.

Most units have higher attack value then defensive value and the ai does not sit back defending cities. Take the medieval age with the medieval foot guy 4 attack 2 defense. Even in hills the attack is fauvored. Archers in stack and the counter attacking possibilites might even out the odds but how does the defender have a clear advantage when the attack value is usually higher?
Cav also seems to favour an attacking civ due to the withdraw and mobilitiy.

Cheers

 

[tjs282]

Archers in stack and the counter attacking possibilites might even out the odds but how does the defender have a clear advantage when the attack value is usually higher?

Attackers don't get an 'attacking-bonus' (unless they're in a Army) on their A-values, i.e. WYSIWYG.

Conversely, a defender's base D-value can be boosted by multiple, additive defensive-bonuses (IIRC, the augmented D-value is then rounded down to the next lowest integer). According to my cheat-sheet*:

— From being fortified (+25%)
— From the terrain:

Mountains = +100%
Volcanoes = +80%
Hills = +50%
Forests, Jungles = +25%
Marsh = +20%
Even open, flat terrain and water tiles still give a +10% bonus, though obviously this doesn't make any difference until you have units with D>=10)​

— If attacked across a river (+25%)
— From Town-Walls (+50%), or later Civil Defence (+50%)
— From a City (+50%) or Metro (+100%)
— From a Fortress (+50%) or Barricade (+100%); also gives Zone-of-Control ability to units that don't usually have it, IIRC

So the only time you're going to see anything close to a matchup between the attacker's A-value and the defender's displayed (base) D-value, is when the defender is unfortifed on flat terrain, and/or during the early game, when unit D-values are still relatively low.

But it should also be noted that the AI's pathfinding-routine will generally prefer to end a unit's move on the 'high ground' whenever available.

Spoiler * :

For the correct bonus-values, consult your conquests.biq file — or @justanick!

 

[justanick]

Hills give +50% on defence, being fortified gives +25%. So that is +75% on defence 3 of pikemen. That is defence 5.25 for 30 shields or 0.175 defence per shield. That is 75% higher than attack 4 per 40 shields. It is the defender that is favoured unless the attacker has some sort of significant advantage.

+10% bonus, though obviously this doesn't make any difference until you have units with D>=10)

No, no such rounding does occur.

Calculating the values of armies is a different matter.

 

[SuedecivIII]

I think this would be mostly useful at the highest of high level games. Not especially useful, but at that level, you take what you can get.

If I'm understanding correctly, let's say you're playing a sid game on a huge pangea map. You choose all the AI civs you want to play against. Naturally, you choose the easiest opponents (The portuguese have very weak traits on Sid for this map type, for example). But you'll run out of weak civs to choose, and some of the mid tier civs will have to be chosen as AI opponents. Let's use the Indians as an example of a mid tier civ.

If we wanted to min max our chances of victory, we'd order the civs such that the portuguese would have first move over the Indians. We could then trigger a war between the two civs during an era where we believe the attackers has the advantage.