Anaconda Plan

[Dom Pedro II]

One of the problems with the AI is that it lacks a grand strategy. Rather than manage the military by finding units to do particular tasks, it has each unit look for something to do.

When a player goes to war with another player, they look to not only capture an enemy's cities but also to deprive the enemy the ability to resist conquest. The best way to do this is to remove their access to resources needed to train units. The AI will surely take advantage to pillage or air bomb sources of Iron, Oil, etc. if they happen to be in range, but there's no systematic effort by the AI to destroy the enemy's ability to field more units.

I would say that the AI should try to do several things in a systematic way: 1) bomb or pillage plots with important strategic resources, 2) capture cities that these resources are located around, 3) blockade ports to cut off overseas sources, and 4) rigorously pursue diplomacy to get players providing the enemy with resources to stop.

In addition to waging economic warfare, I think that we should consider adding the ability for the AI to use deception. In addition to a real target city, it would make sense to me for the AI to have a decoy city. When the enemy ships their defensive forces to the decoy, the AI will attack the real target. Of course, you don't want them to ALWAYS do this since then you'd know that the first city they move against is never the real target.

I also think that the AI doesn't know when to quit either. The AI doesn't really measure its capacity to continue fighting when weighing the value of peace. I've had games where the AI has run wild through my territory winning many early victories until I eventually started pushing back and took out their entire supply of Oil which meant it was going to be all downhill from there for them. But because they had all that early success, they were unwilling to throw in the towel and end the war (I was only looking for it to be over). Now, granted, real people in real wars sometimes have that problem too, but the AI doesn't even weigh the trouble its just gotten into when evaluating a peace deal.

 

[jdog5000]

I agree whole-heartedly on the need for some higher-level strategic planning in the AI when at war ... and in several other parts of the game too. There are some things that can be done (and a few that I have already) to try to coax higher-level thinking out of the unit AI, but it's a pain in the behind.

The real solution is to rework the entire unit AI setup to, as you mentioned, match units to tasks as opposed to have each unit/stack look for it's own task individually ... but that's a bit too much work!

 

[r_rolo1]

I would be more than happy if the AI evaluated who is winning in a better way.... even using the soldiers ratio would be a improvement to the count-heads method that they use now.

 

[Dom Pedro II]

I agree whole-heartedly on the need for some higher-level strategic planning in the AI when at war ... and in several other parts of the game too. There are some things that can be done (and a few that I have already) to try to coax higher-level thinking out of the unit AI, but it's a pain in the behind.

The real solution is to rework the entire unit AI setup to, as you mentioned, match units to tasks as opposed to have each unit/stack look for it's own task individually ... but that's a bit too much work!

Well, we certainly don't want to completely gut the current system. One possibility that I was considering is to evaluate its war strategy at the start of each turn and then it could do one of a few things: 1) Pick out units best suited to accomplish these goals and then push them, or 2) Set certain plots or cities as targets for units when the player cycles through all of their units.

The main problem with the first option is what if the unit hasn't completed the task by the next turn? The unit might get redirected to do something else... To be honest, I wonder how often this happens now. If the unit is checking for the best course of action each turn, isn't it likely the unit will be on the way to do one thing when it's redirected to do something else? I can't help but wonder how many tasks the AI fails to accomplish for this very reason.

The problem with the second option is that if the player stores important target plots, then each unit is going to evaluate each of those plots separately. So they AI might rebase 10 bombers to take out one Oil Well when 1 or 2 would've sufficed. So there'd have to be some way for the AI to say "No, we've already assigned a unit to bomb that plot. Move onto the next."

One of the things I'm working on now is improving the AI's diplomacy skills. In particular, I want the AI to work more intensely trying to diplomatically isolate war enemies. I've added a couple of new features so that the AI will pester their friends more frequently to cut off relations and declare war, BUT if they are initially turned down when they demand it, they will then try to strike a bargain.

So one turn you might get that familiar shot of the leader asking you to cancel your deals, but if you refuse, then a few turns later you might get another contact. This time though, it'll be a diplomacy trade table with the stop trading offer and some sweet carrot the AI will be offering in exchange for stopping trade with their enemies.

I've also added the ability for the AI to ask for peace on behalf of another civilization. Of course, I'm careful not to have the AI be a bleeding heart pacifist here. They'll only ask a civ to make peace with another if they're friendly with the losing party or if the third party is likely to lose and the AI is afraid of the bigger civ. After all, why let a civ that's already bigger than you gobble up another civ that has no chance to potentially slow the progress of the bigger civ?

 

[notque]

Two things would be a huge improvement.

1. Evaluating winning or losing a war effectively.

2. Evaluating the power of a stack effectively.

I think 2 is already done to some extent to determine if boats should leave a city that might be taken over. Am I getting that right? Where have I seen that.

 

[BobTheTerrible]

So one turn you might get that familiar shot of the leader asking you to cancel your deals, but if you refuse, then a few turns later you might get another contact. This time though, it'll be a diplomacy trade table with the stop trading offer and some sweet carrot the AI will be offering in exchange for stopping trade with their enemies.

I like that. It always bothered me that the AI wouldn't bribe humans into trade embargoes/war declarations. One more change I'd like to see is the AI only consider to ask you to join a war if your relation with that AI is better than your relations with the civ the AI wants you to join a war with. It's stupid for a Furious AI to try to ask you to join a war against someone you're pleased with. This change was made in the most recent version of the World Piece mod, although I never actually used the mod to see how it played out.

 

[Dom Pedro II]

I like that. It always bothered me that the AI wouldn't bribe humans into trade embargoes/war declarations. One more change I'd like to see is the AI only consider to ask you to join a war if your relation with that AI is better than your relations with the civ the AI wants you to join a war with. It's stupid for a Furious AI to try to ask you to join a war against someone you're pleased with. This change was made in the most recent version of the World Piece mod, although I never actually used the mod to see how it played out.

Yeah, that's something I'm going to address. Right now, the AI just basically randomly picks someone they're at war with and asks another player to declare war.

I want the AI to not only ask more frequently for assistance against a war enemy but to also approach other civs to help them against the one they're most likely to get assistance for.

So for example: Suppose England and France declare war on Germany. England is more powerful than France. Russia is friendly with England but dislikes France. Italy is friendly with France but is cautious toward England. Eventhough both are at war with Germany, only France is Germany's worst enemy.

Under the current system, Germany will ask both Italy and Russia to stop trading with France but not England in spite of the fact that England is the bigger threat and Italy would be more likely to agree. Germany would also be equally likely to ask Russia to declare war on England eventhough they'd never agree as they would to declare on France where they probably would. Same goes for Italy.

The logical thing would be for Germany to ask Russia to cancel deals with France and Italy to cancel deals with England. Similarly, Germany should ask Russia to declare war on France and Italy to declare war on England. And they should not bother asking anything if they think there's no chance of the other accepting because all it will do is worsen their relationship towards the player that refuses.

But attitudes can't be the only criteria. I need to find the right balance. If Russia is likely to declare war on both, the AI should pick which one would do the most good for them. In this case, that'd be England.

I also don't like the fact that the AI won't even ask other AI to declare war! The join war offer is only made to human players. I mean, that would be a real challenge for the human player if the civs they attack would ask other AI civs to come in and rescue them... Without this, they are sort of fighting with one hand tied behind their back.

 

[BobTheTerrible]

And they should not bother asking anything if they think there's no chance of the other accepting because all it will do is worsen their relationship towards the player that refuses.

Exactly.

I also don't like the fact that the AI won't even ask other AI to declare war!

I didn't know that - I've been playing Civ 4 this whole time thinking they could bribe other AIs to war as well.

 

[jdog5000]

I also don't like the fact that the AI won't even ask other AI to declare war!

Wait wait wait ... what?!? Really?!? How did I miss that? I mean, if the AI would consider joining the war, then their odds of declaring a DOGPILE warplan on their own would be above 0 ... but that's just odds, a request with a little something extra would be a sure thing. The % must be enough to have faked us out that the AI actually did bring in war allies. But that's still unbelievable.

 

[Dom Pedro II]

Yeah, I always thought that the AI got other AIs in wars, but here's the proof (below) ! Just found it today... Although actually the AI will NEVER bribe another player human or otherwise to declare war because they will only demand it. I would like to change this so that the AI will offer human players and other AI players techs, bonuses, etc. in exchange for declaring war on their enemies.

It's all in the CvPlayerAI.cpp file in the AI_doDiplo() function. This is where the AI determines when and what to ask for/offer to other players.

[b]if (GET_PLAYER((PlayerTypes)iI).isHuman() && (GET_TEAM(getTeam()).getLeaderID() == getID()))[/b]
{
	if ((AI_getMemoryCount(((PlayerTypes)iI), MEMORY_DECLARED_WAR) == 0) && (AI_getMemoryCount(((PlayerTypes)iI), MEMORY_HIRED_WAR_ALLY) == 0))
	{
		if (AI_getContactTimer(((PlayerTypes)iI), CONTACT_JOIN_WAR) == 0)
		{
			if (GC.getGameINLINE().getSorenRandNum(GC.getLeaderHeadInfo(getPersonalityType()).getContactRand(CONTACT_JOIN_WAR), "AI Diplo Join War") == 0)
			{
				iBestValue = 0;
				eBestTeam = NO_TEAM;

				for (iJ = 0; iJ < MAX_CIV_TEAMS; iJ++)
				{
					if (GET_TEAM((TeamTypes)iJ).isAlive())
					{
						if (atWar(((TeamTypes)iJ), getTeam()) && !atWar(((TeamTypes)iJ), GET_PLAYER((PlayerTypes)iI).getTeam()))
						{
							if (GET_TEAM(GET_PLAYER((PlayerTypes)iI).getTeam()).isHasMet((TeamTypes)iJ))
							{
								if (GET_TEAM(GET_PLAYER((PlayerTypes)iI).getTeam()).canDeclareWar((TeamTypes)iJ))
								{
									iValue = (1 + GC.getGameINLINE().getSorenRandNum(10000, "AI Joining War"));

									if (iValue > iBestValue)
									{
										iBestValue = iValue;
										eBestTeam = ((TeamTypes)iJ);
									}
								}
							}
						}
					}
				}

				if (eBestTeam != NO_TEAM)
				{
					if (!(abContacted[GET_PLAYER((PlayerTypes)iI).getTeam()]))
					{
						AI_changeContactTimer(((PlayerTypes)iI), CONTACT_JOIN_WAR, GC.getLeaderHeadInfo(getPersonalityType()).getContactDelay(CONTACT_JOIN_WAR));
						pDiplo = new CvDiploParameters(getID());
						FAssertMsg(pDiplo != NULL, "pDiplo must be valid");
						pDiplo->setDiploComment((DiploCommentTypes)GC.getInfoTypeForString("AI_DIPLOCOMMENT_JOIN_WAR"), GET_PLAYER(GET_TEAM(eBestTeam).getLeaderID()).getCivilizationAdjectiveKey());
						pDiplo->setAIContact(true);
						pDiplo->setData(eBestTeam);
						gDLL->beginDiplomacy(pDiplo, (PlayerTypes)iI);
						abContacted[GET_PLAYER((PlayerTypes)iI).getTeam()] = true;
					}
				}
			}
		}
	}
}

Here's a short list of things the AI will never offer/ask of another AI:

Convert Religion
Change Civic
Join War
Stop Trading
Give Help
Ask for Help
Demand Tribute

And here's the things the AI will trade another AI for:

Techs
Bonuses
Maps
Open Borders
Defensive Pacts
Permanent Alliance
Peace Treaty

Another interesting detail is that when the AI offers other things to other AI like bonus trades, tech trades, etc., the other AI will automatically accept. So they could hate each other and one AI could offer Divine Right in exchange for Stealth and the other AI will accept it without question. I'm going to rewrite this because that's just lazy.

I'm interested to see what will happen when the AI is able to make all trades to all players and the other AI will be able to accept or reject the offers. It makes me think that there will be the possibility of more WWII-style balanced wars between two groups of adversaries rather than dogpiles. If the AI is engaging in more diplomacy with the other AI's there's going to be much more hurt feelings. If the AI is really angry at a number of players, they're going to be more inclined to start wars with them and more inclined to join wars when offered by other players. The number of wars per game will definitely go up either way... I think it will probably drive the AIs into rival camps. Or it might just create a complete free-for-all. Should be fun! I'm going to add new code though so players know exactly when one AI player gets another AI player to cancel deals with them. After all, if the AI knows when I get someone to cancel deals with them and bring in war rivals, I should be able to know too!

 

[PieceOfMind]

Fascinating. Thank you for enlightening us Dom Pedro.

 

[MadmanAtW]

It's stupid for a Furious AI to try to ask you to join a war against someone you're pleased with.

RP-wise it's stupid, but gameplay-wise it's still a decision the player has to make for whether or not to get the -1 "You refused to help us during war time" penalty against. Maybe not a very interesting decision, but it does have a gameplay effect that would have to be weighted before taking it out.

 

[Dom Pedro II]

RP-wise it's stupid, but gameplay-wise it's still a decision the player has to make for whether or not to get the -1 "You refused to help us during war time" penalty against. Maybe not a very interesting decision, but it does have a gameplay effect that would have to be weighted before taking it out.

Although on the other hand, I would also argue that it might not be in the AI's best interest to make that demand. If the AI demands the player to make war on someone the player is very unlikely to accept, the AI will have a lower opinion of the player. This will matter when the AI weighs future diplomatic decisions such as continuing trades, making new trades or even declaring war. All of these could potentially hurt the player, but they also have a very good chance of hurting the AI as well. That -1 attitude could be as much of an impediment for the AI as the human player dealing with that AI civ.

At the moment when the AI is waging war, they should be looking to strengthen diplomatic ties rather than antagonizing them. But I guess this leads to an important question: Is the AI supposed to be a substitute for other human players struggling to survive and try to win the game? Or is the AI's primary purpose simply to be a hurdle for the human player?

For example, in the last game I was playing, I got started on an island. I was lagging pretty far behind the lead civilizations in the game, but I was climbing the ranks at a fairly decent clip once I'd been able to get out and explore the rest of the map. If the AI was in the game to win it, one would expect that they'd all dogpile on the leading civ which wasn't me. However, instead civilization after civilization (some weaker than me) declared war and invaded. I was thousands of miles away from them neither being a real threat or an ideal target, but they declared war on me just the same while leaving each other alone. I mean, the maintenance costs for cities that far alone would've made trying to conquer me a deleterious endeavor, but that didn't stop the AI from sending waves and waves of ships and troops that I (for the most part) easily repulsed. It was only towards the end when several of the biggest civs with fully upgraded units started landing on my shores that my chances started to look bleak.

So for the goal of harrassing me, the AI had ultimately been successful. But in doing so, they wasted valuable resources trying to conquer me that left them open to attack on their home continent and also ensured that Rome was going to win the game through any victory choice it wanted.

So when considering decisions to improve the AI, we have to ask ourselves, do we want the AI to be a better player or a better antagonist? I don't think there's really a right answer. Just a matter of preference.

 

[Dom Pedro II]

In case you weren't disgusted enough already

I was looking into the tech value code and I discovered that the AI isn't the least bit interested in what a tech actually does when weighing its value. It's only really concerned with it's cost subtracted by their current progress and how many civs already have it.

The main reason for this is that the AI doesn't actually have any kind of a function for simply determining the value of a tech. The only thing it has is a function that determines the best tech. What I would like to do is extract a lot of the code used in figuring out the best tech and have it simply return a numerical value for a given technology so it can be applied in other places such as this. I really don't know why they didn't do that in the first place since they did it with just about every other game feature.

 

[r_rolo1]

At the moment when the AI is waging war, they should be looking to strengthen diplomatic ties rather than antagonizing them. But I guess this leads to an important question: Is the AI supposed to be a substitute for other human players struggling to survive and try to win the game? Or is the AI's primary purpose simply to be a hurdle for the human player?

For example, in the last game I was playing, I got started on an island. I was lagging pretty far behind the lead civilizations in the game, but I was climbing the ranks at a fairly decent clip once I'd been able to get out and explore the rest of the map. If the AI was in the game to win it, one would expect that they'd all dogpile on the leading civ which wasn't me. However, instead civilization after civilization (some weaker than me) declared war and invaded. I was thousands of miles away from them neither being a real threat or an ideal target, but they declared war on me just the same while leaving each other alone. I mean, the maintenance costs for cities that far alone would've made trying to conquer me a deleterious endeavor, but that didn't stop the AI from sending waves and waves of ships and troops that I (for the most part) easily repulsed. It was only towards the end when several of the biggest civs with fully upgraded units started landing on my shores that my chances started to look bleak.

So for the goal of harrassing me, the AI had ultimately been successful. But in doing so, they wasted valuable resources trying to conquer me that left them open to attack on their home continent and also ensured that Rome was going to win the game through any victory choice it wanted.

So when considering decisions to improve the AI, we have to ask ourselves, do we want the AI to be a better player or a better antagonist? I don't think there's really a right answer. Just a matter of preference.

If I had a € for every time that happened to me , I would be rich :/ I do remember quite well one of the Lonely hearts game where everyone in the game DOWed me ( including Mansa ), while the Khmer was cruising to a Dom win ( I managed to turn the tide by a great ammount of diplo work that ultimately led to QSH, the #2 in the world, to turn against khmer and to see a nice nuclear firework exchange while I was going away from the planet, but that is besides the point )

The point of gamer AI vs "stone in the shoe" AI in the end is matter of balance. My personal position is that the AI should be able to pass a Civ IV Turing test, but I can see that kind of AI would not be welcomed by a lot of people. But I think that when the designers aproached the problem of the Civ IV AI, they have gone too much in "Designing a AI to lose" path ( like Soren said.... don't know the official Blake position on this, but his comments about the BtS Agg AI option make me think that he would had prefered a AI with more teeth ( not that I consider Agg AI in any way superior to the normal AI... it is just more agressive, most of the time not for their own good )), and that the Civ IV rarely has either self preservation instinct or a long term plan ( besides culture, and even there things are ugly.... for a example, if a AI changed cap for any reason, it will not consider a culture win viable ) .
I would definitely like to see a AI that fights to not lose, and in fact I already proposed elsewhere that the fact that a civ is aproaching a victory condition should give diplo demerits with everyone else ( it is the only realistical way of aproaching this in Civ IV IMHO ) to push the rest of the AI to the fight. Allies of that civ would probably continue to be Allies for a while ( a little more reluctant, ofc .....no one should be happy to work for their own demise ) and it would push more neutral civs to the other side........

In case you weren't disgusted enough already

I was looking into the tech value code and I discovered that the AI isn't the least bit interested in what a tech actually does when weighing its value. It's only really concerned with it's cost subtracted by their current progress and how many civs already have it.

The main reason for this is that the AI doesn't actually have any kind of a function for simply determining the value of a tech. The only thing it has is a function that determines the best tech. What I would like to do is extract a lot of the code used in figuring out the best tech and have it simply return a numerical value for a given technology so it can be applied in other places such as this. I really don't know why they didn't do that in the first place since they did it with just about every other game feature.

Can you say where is the code? Indeed that looks a big lapse that would hinder the AI a lot.

 

[Dom Pedro II]

Can you say where is the code? Indeed that looks a big lapse that would hinder the AI a lot.

Sure. It's in the CvTeamAI.cpp file. The function is AI_techTradeVal. It actually probably works fairly well under most circumstances because there is at least a tech-specific trade modifier for it in the XML. But if the AI is at war and there's an expensive tech that's critical for economic improvement, the AI might want to trade for that before they trade for a valuable military tech. Or they might be at peace with no serious threats and they'll demand a military tech before a good research-boosting tech.

Posted here:

int CvTeamAI::AI_techTradeVal(TechTypes eTech, TeamTypes eTeam) const
{
 FAssert(eTeam != getID());
 int iKnownCount;
 int iPossibleKnownCount;
 int iCost;
 int iValue;
 int iI;
 iCost = std::max(0, (getResearchCost(eTech) - getResearchProgress(eTech)));
 iValue = ((iCost * 3) / 2);
 iKnownCount = 0;
 iPossibleKnownCount = 0;
 for (iI = 0; iI < MAX_CIV_TEAMS; iI++)
 {
  if (GET_TEAM((TeamTypes)iI).isAlive())
  {
   if (iI != getID())
   {
    if (isHasMet((TeamTypes)iI))
    {
     if (GET_TEAM((TeamTypes)iI).isHasTech(eTech))
     {
      iKnownCount++;
     }
     iPossibleKnownCount++;
    }
   }
  }
 }
 iValue += (((iCost / 2) * (iPossibleKnownCount - iKnownCount)) / iPossibleKnownCount);
 iValue *= std::max(0, (GC.getTechInfo(eTech).getAITradeModifier() + 100));
 iValue /= 100;
 iValue -= (iValue % GC.getDefineINT("DIPLOMACY_VALUE_REMAINDER"));
 if (isHuman())
 {
  return std::max(iValue, GC.getDefineINT("DIPLOMACY_VALUE_REMAINDER"));
 }
 else
 {
  return iValue;
 }
}

 

[r_rolo1]

That is pretty much skeleton code indeed.....

 

[phungus420]

So when considering decisions to improve the AI, we have to ask ourselves, do we want the AI to be a better player or a better antagonist? I don't think there's really a right answer. Just a matter of preference.

Blake already answered this, as did Soren when CIV first came out. The designed role of the AI is to be a player, that plays to win, but does so in a defined "Role Play" based strategy.

All of your examples of the AI acting stupid that break this description are due to a lack of time/resources on Firaxis's part. I don't fault them for it either btw, that's not a complaint, just a fact. CIV: BtS has the best AI of any strategy game I am aware of. I'm simply pointing out the fact Firaxis is a buissness, with a profit motive and shareholders to consider. As such they have real time and resource issues, and thus every single aspect of the AI could not possibly be adressed by Blake in the dev time alotted.

Anyway, any fix or change to AI behavior should be made with that design principle in mind: That the AI plays to win, but will only take actions that fit for it's designed role play behavior.

Edit: Also the often mentioned every AI should attack the player that is winning is a bogus argument that the AI isn't designed to win. As it ignores the prisoner's dilema, if the AI's were behaving as seperate entatities trying to win. Such behavior actual goes against the playing to win strategy, and is just a gamy derivative and gives the AI player a very artificial texture, real humans do not behave that way.
And given that Firaxis's lead BtS AI programmer has stated the goal behavior of the BtS AI, I see no reason to rehash or deviate from the stated principle design. Though, arguably that may be because I agree with it in principle, as I think that sort of AI makes for the most entertaining as well as challenging opponent.

 

[Dom Pedro II]

Blake already answered this, as did Soren when CIV first came out. The designed role of the AI is to be a player, that plays to win, but does so in a defined "Role Play" based strategy.

All of your examples of the AI acting stupid that break this description are due to a lack of time/resources on Firaxis's part. I don't fault them for it either btw, that's not a complaint, just a fact. CIV: BtS has the best AI of any strategy game I am aware of. I'm simply pointing out the fact Firaxis is a buissness, with a profit motive and shareholders to consider. As such they have real time and resource issues, and thus every single aspect of the AI could not possibly be adressed by Blake in the dev time alotted.

I didn't mean to give the impression that I was criticizing the game developers. They put out a great product with an AI that was certainly good enough for a finished commercial product. Thankfully, they've provided us with a way to tweak the AI to be more according to the visions of each modder.

Anyway, any fix or change to AI behavior should be made with that design principle in mind: That the AI plays to win, but will only take actions that fit for it's designed role play behavior.

Edit: Also the often mentioned every AI should attack the player that is winning is a bogus argument that the AI isn't designed to win. As it ignores the prisoner's dilema, if the AI's were behaving as seperate entatities trying to win. Such behavior actual goes against the playing to win strategy, and is just a gamy derivative and gives the AI player a very artificial texture, real humans do not behave that way.

Alright, good point. Perhaps I should've fleshed out my statement a bit. But the second-highest player has a compelling interest in knocking the strongest player out of first place. And a player has an interest in knocking anyone further down the ladder. I played a game last night where the two most powerful civs went to war, and I declared war on one of them because it served my interests to help dismantle them. By the end of the war, I was the second most powerful civilization and was also the largest.

My main point was that the AI shouldn't make decisions intended to hurt the human player without regard to whether it was beneficial for itself. The original point started about the AI picking stop trading teams the human is unlikely to accept, but it pertains just as well to the AI sending dozens of units into the meat-grinder to try to take a mediocre civilization halfway across the map and then in spite of catastrophic losses being unwilling to sign a treaty.

EDIT: I should also note that this was playing with Better AI. Personally, I don't think the AI would've actually done this in the base game.

And given that Firaxis's lead BtS AI programmer has stated the goal behavior of the BtS AI, I see no reason to rehash or deviate from the stated principle design. Though, arguably that may be because I agree with it in principle, as I think that sort of AI makes for the most entertaining as well as challenging opponent.

Well, I think we're in general agreement though. We both want an AI that plays as a player and not just an obstruction.

 

[Yakk]

Note that a Balance of Power approach both fits the "trying to win" approach, and might make more interesting games.

Of course, this requires that the powers you are balancing take this into account.

The most powerful empire "needs you less", and the less powerful empire "needs you more".

So the more powerful empire might have a tendency to bribe their allies less, and the less powerful empire might have a tendency to bribe their allies more.

On the other hand, we should also keep track of what "trying to win" means. Someone in 10th place in 1500 is not going to win the game, short of everyone else being idiots and ignoring the attempted cultural victory of the 10th place player.

That 10th place player may, however, attempt to have the highest score it can at the end of the game. This may involve slowing down other teams victory drive, and boosting their own points as high as possible.