SAP2 AI V's other mods AI

[Colwyn]

Ive played most of the other mods and still found the AI so weak.

But the new SAP 2 AI is brillant and for the first time makes CTP2 a very good game, I would love to see this new AI in MM as I like the new units etc.

What are other people's experiences with the AI in SAP2

[Maquiladora]

Have you tried Cradle? I thought the AI was tougher on that. Maybe more in WAW.

[Immortal Wombat]

SAP uses a modified version of Cradle, as does WAW, so I think they'll be pretty much the same. Except that the tech tree, amount of units, buildings, decisions in SAP is less than in Cradle, so the AI is probably able to cope better overall.

I wonder if it is possible to set up a game with 8 AI players, 2 each using MM2, SAP, WAW and Cradle AIs... See who wins!

[Maquiladora]

My vote goes to the ones using WAW, with all that modern weaponry and big cities

[HuangShang]

Originally posted by Maquiladora
My vote goes to the ones using WAW, with all that modern weaponry and big cities

Me too!

and i think Cradle has the hardest AI

[Maquiladora]

I think he just meant the AI, so yeah id go for cradle too.

[centrifuge]

The AI in SAP is tough, but from my point of view, the problem with it as well as the AI from WAW, (I assume that this goes for cradle as well, but I haven't played enough of it), is that in a war, if you can withstand the initial bombardment of the masses of troups that are sent your way, the AI's cities are then easy pickin's. In my last SAP tournament game, the AI defended its cities fairly good at some points but most of them were defended only with a single unit...against my stack of 12.

[Maquiladora]

And more often than not, a large amount of their units are old ones, hoplites archers etc when the standard is infantrymen and cannons. I think its because they dont go to war with anyone else so they dont need to replace them, so they think a city defended by 12 units is well defended. Yet they dont realise not with older units. i dunno.

[HuangShang]

WaW is easier for me b/c of the quick start

cradle doesn't start w/ too many types of units thanx to the age... and the AI can stack up quick

[TheArsenal]

Originally posted by centrifuge
The AI in SAP is tough, but from my point of view, the problem with it as well as the AI from WAW, (I assume that this goes for cradle as well, but I haven't played enough of it), is that in a war, if you can withstand the initial bombardment of the masses of troups that are sent your way, the AI's cities are then easy pickin's. In my last SAP tournament game, the AI defended its cities fairly good at some points but most of them were defended only with a single unit...against my stack of 12.

Agreed. I play a lot on this mod (SAP). And for me it has a challenging fun AI. But it is definitely attack minded at the expense of defense. But those initial onslaughts are generally fierce enough that I'm in reactive mode for much of the start of the game anyway, and if I'm careless I lose cities. So in one respect its really a case of the best defense is a good offense because I'm too pinned in to go after the AI's cities. But as you said, once I do mount a counter-attack, I am able to go slicing through AI civs. And as a rule, I'm finding the larger AI civs (especially if they've been at war a while with other civs) with lots of cities will only have a few well defended ones, but the more compact civs are better defended.

[Dale]

Since the issue of AI defense has come up again, I'll put my two cents on as well. What I've noticed is the same as TheArsenal, where smaller AI civs are well defended, where large AI civs have trouble putting up a good defense across the nation.

But if you think of how IRL history is, isn't this true of our world? When nations got too big, they found it hard to maintain a strong defense across the board. Consequently, smaller nations were able to take little bites out of the large nation bringing it back to size. Our whole history is rife with "The Rise and Fall of Nations".

The only thing I see happening is for the player, we overcompensate for this by building more units. Maybe we need something SLIC'ed in, that as the player gets larger, military costs escalate exponentially, as opposed to sequentually like it does now. That sounds reasonable to me. The more units you have, the sheer logistics and cost of keeping track and co-ordinating everything exponentually grows.

What's other's ideas on this? My thoughts are that this idea could cause the "Rise and Fall of The Player".

[HuangShang]

i'd agree w/ Dale too

after i grow enough, i leave some cities totally unguarded on perpose... and only heavily(6-9) on border cities.

but at least no puny nation will take a peice of this!

[Dale]

I had a thought at lunch.....

The AI has no problems making PW and gold, and TONS of it! (PW cheat, free unit upgrades, tile imps, etc)

CTP2 only simulates wages when it comes to military support. It doesn't take into account:
- Barracks maintenance
- Infrastructure maintenance (the more tanks on the road, the more damage they are doing to the roads)
- Wages of auxillarys (cleaners, cooks, etc)

So here's my proposal:

- Unit's 1 --> 25 cost: 1 gold + 1 PW per turn.
- Unit's 25 --> 50 cost: 1 gold + 2 PW per turn.
- Unit's 50 --> 100 cost: 1 gold + 3 PW per turn.
- Unit's 100 --> 200 cost: 1 gold + 4 PW per turn.
- Unit's 200+ cost: 1 gold + 5 PW per turn.

So, to compare how much extra support you pay:
- 1 unit: 1 gold + 1 PW (no big deal)
- 20 units: 20 gold + 20 PW (still no real big deal)
- 40 units: 40 gold + 55 PW [25 + 15*2 PW] (starting to take a dent out of economy)
- 70 units: 70 gold + 135 PW [25 + 25*2 + 20*3 PW] (this is starting to hurt production!)
- 150 units: 150 gold + 425 PW [25 + 50 + 150 + 50*4 PW] (OUCH!)

So basically, the larger an army you have the more hurtfull to the economy that army is going to be. This in my mind should cause the "Rise and Fall of Nations". Of course, my figures are only a guide as to how it'll work, and is open to discussion.

Any thoughts on the above?

[Maquiladora]

I like it. Really brings public works into its own, makes it far more important. And stops that insane build up of PW at the end when all tiles have decent improvements.
But there seems to be a general want to make the game harder for the human rather than improving the AI thinking when it comes to its armies. You can add this to make the human build units slower but the AI will still have the same problems with units defending cities, and what type of units. In the end the human *will* catch up in terms of units and will still find the AI woefully defending its cities.
I dont claim to be in any way an proficient in SLIC or this AI, but like production bonuses, making the game harder for the human rather than concentrating on improving the AIs efficiency seems a false way of making the game more difficult.

[child of Thor]

Yeah, it seems to make sense - and i suppose if it doesn't penalise the AI too much(with all the bonus's it gets), then it would make fielding huge armies much more of a logistical problem for the player.Gold from trade would be very important, PW from improvements,treaties and special attacks(franchise?), would all be vital to try to maximise. I like it very much.

[Immortal Wombat]

What happens when you run out of PW?

[J Bytheway]

Originally posted by Immortal Wombat
What happens when you run out of PW?

You lose units - presumably those farthest from cities first. If you want to be nice make each disband give you a PW boost so that you don't lose too many units in any one turn.

[Immortal Wombat]

I think it won't be easy to balance where it hinders the human but not the AI.
I'm not convinced it will even the odds either, but it depends on the AI bonuses.

Also, it will probably cripple early-game warfare, and the barbarians will have a field day. I use almost all of my early-game PW in farms.

It'll certainly be interesting to see how it affects play.

[Hermann the Lombard]

Originally posted by Immortal Wombat


I think it won't be easy to balance where it hinders the human but not the AI.
I'm not convinced it will even the odds either, but it depends on the AI bonuses.

Well, if it's too crippling we'll have to moderate the costs, keeping the early game barb threat in mind. Hmm...seems like ancient and classical units might need proportionately less PW support (more self-sufficient, in other words). A legion can repair its own weapons and forage, an armored division is less able to do so.

"We" -- Easy for me to say, since I'm not a SLIC coder!

[Peter Triggs]

Also, the AI doesn't use infrastructure. So this might actually give the Human an advantage later in the game.

[Pedrunn]

Peter, Player1 wrote this.
What oyu think?

[code]

////////////////////////////////////////////////////////////////////////////
//Infrastructure SLIC v1.01
//
//This code fixes AI bug in which AI players, when trying
//to use Infrastructure or Capitalization, leave build queue empty.
//It detects cities with empty build queues and gives them
//15 per city pop PW and 10 per city pop gold bonus,
//lowered by PW tax percent value.
//
//Example:
// City size 28, PW tax 20%
// gives 28*15*(1-0.2)=336 PW and 28*10*(1-0.2)=224 gold bonus
//
// by player1
////////////////////////////////////////////////////////////////////////////

HandleEvent(AIFinishBeginTurn) 'Infras' post {
if(!IsHumanPlayer(player[0])) {	// not human
	int_t i;
	int_t my_tax;		//PW tax
	int_t my_pw;		//toloal PW
	int_t my_capit;		//gold from cities
	int_t my_infras;	//PW from cities
	city_t my_city;
	player[0] = g.player;
	my_tax = player[0].publicworkstax;
	for(i = 0; i < player[0].cities; i = i + 1) {
		GetCityByIndex(player[0], i, my_city);
		if (!my_city.buildqueuelength && (HasAdvance(player[0], ID_ADVANCE_MASS_PRODUCTION) || HasAdvance(player[0], ID_ADVANCE_GLOBAL_ECONOMICS))) {
			my_infras = my_city.population * 15 * (100 - my_tax) / 100;
			my_pw = player[0].publicworkslevel;
			my_pw = my_pw + my_infras;
			SetPW(player[0], my_pw);	//add pw
			my_capit = my_city.population * (100 - my_tax) / 10;
			AddGold(player[0], my_capit);	//add gold
		}
	}
}
}

[TheArsenal]

Personally, I didn't mean for this to sound like its a problem on my part, because it is not. I've always looked at it as an AI civ's "strategy" more than a game flaw because every civ doesn't play that way. As Dale alluded to, it makes sense that the civs that have been at war (for a few thousand years sometimes) and sent huge stacks of troops all over the land mass pay the price in the end with poorly defended cities. And it makes sense that civs that are mostly peaceful are, as a rule, smaller but well defended. To me this is balanced. Plus I get a great deal of satisfaction from punishing aggressive civs in the end that tried to wipe me off the map at the start.

I'm not sure if a larger Army should be economically penalized. Large armies, especially at war, are actually good for the overall economy (the so called "War Time Economy"). And even in peace time where large armies directly drain government budget, the actual military personnel themselves put a great deal back into the country's coffers because of their unusual amounts of disposable income.

[HuangShang]

Originally posted by TheArsenal
Personally, I didn't mean for this to sound like its a problem on my part, because it is not. I've always looked at it as an AI civ's "strategy" more than a game flaw because every civ doesn't play that way. As Dale alluded to, it makes sense that the civs that have been at war (for a few thousand years sometimes) and sent huge stacks of troops all over the land mass pay the price in the end with poorly defended cities. And it makes sense that civs that are mostly peaceful are, as a rule, smaller but well defended. To me this is balanced. Plus I get a great deal of satisfaction from punishing aggressive civs in the end that tried to wipe me off the map at the start.

I'm not sure if a larger Army should be economically penalized. Large armies, especially at war, are actually good for the overall economy (the so called "War Time Economy"). And even in peace time where large armies directly drain government budget, the actual military personnel themselves put a great deal back into the country's coffers because of their unusual amounts of disposable income.

[Peter Triggs]

Pedrunn,

I think that must be an early version. Here's the version from SAP:

//
// pw_cheat.slc
// by player1
// ver 1.01
//
// gives 5 gold per global pop evry turn,
// gives 0-20 PW per global pop evry turn (depending of PW tax)
// for AI players only
//

HandleEvent(BeginTurn) 'PW_Cheat' pre {
	int_t my_tax;
	int_t my_pop;
	int_t my_pw_level;
	int_t my_pw_points;
	int_t my_gold;
	if(!(IsHumanPlayer(player[0])) && (player[0] != 0)) { // not human
		my_tax = player[0].publicworkstax;
		my_pop = player[0].totalpopulation;
		my_pw_level = player[0].publicworkslevel;
		my_pw_points = my_pop * my_tax / 5;
		my_pw_level = my_pw_level + my_pw_points;
		SetPW(player[0], my_pw_level);		//add pw
		my_gold = my_pop * 5;
		AddGold(player[0], my_gold);		//add gold
	}
}

This just gives each AI player a gold and PW bonus based on their population. Either player1 or Dale must have simplified it.

I'm a bit dubious about the 'Infras' handler: I guess it works but I don't like the line "player[0] = g.player;" There's already a player[0] here, the AI player who's finishing his BeginTurn. I don't even know what that line does (if anything) in this context. I think you'd only want to do something like that in the very special circumstances where you're triggering on the UI: g.player will be the obviously human player who pressed the button and then when player[0] gets a value you can access the other members of the player array (like, player[0].militaryunits).

Another thing is that it doesn't actually solve the problem: there's a BuildListElement Infrastructure which you can find in some of the BuildListSequences but it doesn't seem to work. I say this because i) in Cheat mode I've never seen the AI civs using infrastructure, ii) if I put my cities under mayor control where they use those BuildListSequences, I've never noticed them go to Infrastructure, and iii) I can't construct a custom buildlist that contains Infrastructure.

So, when I said the AI doesn't use infrastructure, I should have said "yet". IMO, here's a better solution. Create a dummy building with a huge production cost and a BuildingBuildlist that contains only it:

IMPROVE_QUEUE_INFRASTRUCTURE {
DefaultIcon ICON_IMPROVE_INFRASTRUCTURE
Description DESCRIPTION_IMPROVE_INFRASTRUCTURE
EnableAdvance ADVANCE_SLAVE_LABOR

ProductionCost 100000
Upkeep 0
}

BUILDING_BUILD_LIST_QUEUE_INFRASTRUCTURE {
Building IMPROVE_QUEUE_INFRASTRUCTURE
}

Disappointingly, there's no way to access a city's production directly via SLIC, but:

HandleEvent(CityBeginTurn)'InfrastructureWorkaround'pre {
int_t cityPW;
int_t playerPW;

    if (IsBuildingAtHead(city[0])== BuildingDB(IMPROVE_QUEUE_INFRASTRUCTURE)) {      
	    cityPW=100000/city[0].buildingtime;	
	    player[0]=city[0].owner;
	    if (IsHumanPlayer(player[0])){
	         cityPW=(500*cityPW)/1000;
	    }   
         playerPW=player[0].publicworkslevel;
	   SetPW(player[0], playerPW+cityPW);
         Event:ZeroProduction(city[0]);
	   if(AtWarCount(player[0]) && !IsHumanPlayer(player[0]) ){ // if at war, stop AI queueing	       	     
	         ClearBuildQueue(city[0]);
         }
    }
}

That is, if you divide the dummy building's production cost by the number of turns it will take the city to build it, you get how much the city is producing this turn. (This was Wombat's idea, neat eh!)

And now you can use the above BuildingBuildlist in any of the BuildListSequences that you want. It allows a considerable simplification in them. For example, here's the Science BuildListSequence I'm currently using:

BUILD_LIST_SEQUENCE_SCIENCE {//scientist: Top 0.3 ProductionCities, others Top 0.1 ProductionCities
BuildListElement { BuildingBuildList BUILDING_BUILD_LIST_SCIENCE }
BuildListElement { BuildingBuildList BUILDING_BUILD_LIST_PRODUCTION }
BuildListElement { BuildingBuildList BUILDING_BUILD_LIST_GOLD }
BuildListElement { WonderBuildList WONDER_BUILD_LIST_SCIENCE }
BuildListElement { GarrisonUnitBuildList }
BuildListElement { AllUnitBuildList }
BuildListElement { Freight }
BuildListElement { BuildingBuildList BUILDING_BUILD_LIST_GROWTH }
BuildListElement { BuildingBuildList BUILDING_BUILD_LIST_HAPPINESS }
BuildListElement { WonderBuildList WONDER_BUILD_LIST_GOLD}
BuildListElement { WonderBuildList WONDER_BUILD_LIST_HAPPINESS }
BuildListElement { BuildingBuildList BUILDING_BUILD_LIST_QUEUE_INFRASTRUCTURE }

}

You'll get a chance to see more of this in my forthcoming mod, "Call to Bloodlust", which uses SAP as a base.

As far as large armies go, I agree with The Arsenal, although not for the reasons he put forward. IMO, the AI needs LOTS of military units before it will do anything really interesting.

[Martin Gühmann]

Originally posted by Pedrunn
Peter, Player1 wrote this.
What oyu think?

I fixed this in a different way. I just removed the Infrastructure and Capitilization from the AI builds list so they have now to build wonders, buildings and units. The mean goal was more units so that the AI send themto the slaughter at the front.

-Martin

[Immortal Wombat]

Code:

cityPW=(500*cityPW)/1000;

I'd have used

cityPW=cityPW/2;

, but whatever
Peter, any clues on when CtB will be out?

[Martin Gühmann]

Originally posted by Peter Triggs
Pedrunn,

I think that must be an early version. Here's the version from SAP:

This just gives each AI player a gold and PW bonus based on their population. Either player1 or Dale must have simplified it.

Actual these are two different scripts the one Pedrunn posted deals with the problem of empty build queues. This one should give the AI additional Gold and PW each turn per city if the AI has the according advance and the city has an empty build queue per city. The one you posted is the AI PW cheat, that gives the AI additional PW and Gold each turn. I don't use the infrastructure script, because I removed the infrastructure from the AI build list, so no need for me for this.

Originally posted by Peter Triggs
I'm a bit dubious about the 'Infras' handler: I guess it works but I don't like the line "player[0] = g.player;" There's already a player[0] here, the AI player who's finishing his BeginTurn. I don't even know what that line does (if anything) in this context. I think you'd only want to do something like that in the very special circumstances where you're triggering on the UI: g.player will be the obviously human player who pressed the button and then when player[0] gets a value you can access the other members of the player array (like, player[0].militaryunits).

The player array, it was a pain in Dale's airunit script, with DebugSlic=Yes I got each time an airunit moved an array index 0 out of bounce error. I replaced the 0 by a 1 in that array and the error index out of bounce errors were reduced, in the end I was able to replace the player[] array entirely, and the problem was gone, just an occasional error in the DIP_EstabEmb is left, actual I would like to switch to your new DiploMod Peter, so is it now in a stage ready for use.

Originally posted by Peter Triggs
Another thing is that it doesn't actually solve the problem: there's a BuildListElement Infrastructure which you can find in some of the BuildListSequences but it doesn't seem to work. I say this because i) in Cheat mode I've never seen the AI civs using infrastructure, ii) if I put my cities under mayor control where they use those BuildListSequences, I've never noticed them go to Infrastructure, and iii) I can't construct a custom buildlist that contains Infrastructure.

Yes Player1 actual gave back only the wasted work.

Originally posted by Peter Triggs
You'll get a chance to see more of this in my forthcoming mod, "Call to Bloodlust", which uses SAP as a base.

I hope you took a look on my GoodMod, so when it will be finished.

Originally posted by Peter Triggs
As far as large armies go, I agree with The Arsenal, although not for the reasons he put forward. IMO, the AI needs LOTS of military units before it will do anything really interesting.

I agree to this, if the AI doN#t attack you, you can build up an huge army in all the silence and suddenly attack and slaughter the enemy. That's the reason why I made this better AI slic file that groups the AI stacks and moves them into transports, you can see the results in the tournerment game, unfortunatly only in the late game, when the capacity of the transports increases, but I saaw a lot of AI cities that changed hands countiously. Unfortunatly the current public version of this slic file isn't totally bug free, only the version on my hard drive has no known problems with DebugSlic=Yes.

-Martin

[Peter Triggs]

Originally posted by Martin:

I hope you took a look on my GoodMod

Natually, I've included GM1_Goods.slc: no mod can be without it. You know Martin, maybe you should consider renaming GoodMod because when people see 'GoodMod' they might be thinking 'I've got that already in Cradle (or MedMod)' and not realize that there's more to it than that.

I would like to switch to your new DiploMod Peter, so is it now in a stage ready for use.

I haven't done any work on that side of diplomacy for some time. I'm not particularly happy with it: I want to tie the responses more into the AI player's current diplomatic state. But this means first sorting out what diplomatic states I want them to be in under which circumstances, and that's part of what I'm working on now.

That's the reason why I made this better AI slic file that groups the AI stacks and moves them into transports, you can see the results in the tournerment game, unfortunatly only in the late game, when the capacity of the transports increases

That's interesting, I've tried increasing the capacity of transports right from the beginning but still the AI won't use them. It looks like there's no way out except by doing it with SLIC.

Originally posted by IW:

Peter, any clues on when CtB will be out?

That's a good question. I spent literally all day yesterday (till the wee small hours of the morning) tracking down a fault in it: more than half the time when I made contact with a nearby civ they would immediately go to war with me. I mean we want belligerency in the AI, but on turn 17? So I made some simple adjustments, and now they're a bit passive. I'd really like to get it working in such a way that a human player isn't forced into a single strategy and if you make the AI too aggressive, that's what seems to happen.

[Dale]

That's interesting, I've tried increasing the capacity of transports right from the beginning but still the AI won't use them. It looks like there's no way out except by doing it with SLIC.

I dunno. I've found the AI makes decent use of transports in WAW. The amount of times I've been naval invaded in WAW has surprised me. And I don't mean invaded by one unit. There was this one time at band camp...... Just kidding.

But seriously, one game I had an AI naval invade me using transports, escorts and landed infantry and tanks. FOUR TRANSPORTS FULL!

[Martin Gühmann]

Originally posted by Dale
I dunno. I've found the AI makes decent use of transports in WAW. The amount of times I've been naval invaded in WAW has surprised me. And I don't mean invaded by one unit. There was this one time at band camp...... Just kidding.

But seriously, one game I had an AI naval invade me using transports, escorts and landed infantry and tanks. FOUR TRANSPORTS FULL!

The question here is how big is the transport capacity of your transports, probably five units. The other question is what did you?

So far I thought it is possible to improve the AI just by modifying the text files. So I used slic to stack the AI units that hand an effect, it looked that the problem was that the AI didn't stack it's units. So I did this the next problem was that I found settlers in these stacks, so next stage sort the armies. Ok this was done then I discovered the AI would restack the settlers, so then they had to leave the stack. Another thing was that you shouldn't do it in a city. So I reduced the stack size in cities to allow the AI to produce new units. Then I didscovered that all the transports were empty, because the stack size was too big, so a ship loader was neccesary. So when ever now a ship comes to the cost and there is a huge stack then I will move all the units into the transport, but only if I can move into the transport more than four units or I can fill the transporter and the stack is on homeland. you can see the results in the tournerment thread were I posted a screenshot with an invasion, the AI used a stack of transporters no escorts, because it needed a deep water ships for the escorts and these one weren't available at that point.

For the better AI slic I recommend the version I posted in the tournerment thread. That is a debugged version with DebugSlic=Yes.

In the last tournerment with ApolytonPack I had more then 1000 years war with the Romans and all what I saw was some greek fires pillaging my cost, no invasion, even with one unit nothing.

-Martin