Ai

[Velociryx]

On my way to check it out now!

[ACooper]

A customizable AI would be great.

One thought in education is about "multiple intelligences" which basically mean people have different intelligence levels in different areas. (music, math,sports, etc.)

Well why couldn't we do that with the Civs AI. Break it down into areas such as Military, Defense, Science, Domestic, Culture, etc. (we could call each module an "advisor" or "cabinet member") We would still need a master AI or rules table to determine which advisor would take precedence if there was a conflict.

What I like is the idea of being able to mix and match AI components to form a new AI. Maybe the "out of the box" Civs would have an optimized "Cabinet" of advisors but can be customized. It could even be that "advisors" could change with certain thresholds are reached. For Example the Military advisor could go from a peaceful AI to a warlike one.

To sum it all up ('cause I at work) I guess I would like to see the AI Object Oriented and customizable.

[Spaced Cowboy]

Wow,

That Civ Evolution site looks pretty awesome. It is strange how I go from nothing to keep me interested to overload. Tournament game, Vel's mod, and now my chance to write an AI; I'll be busy for weeks.

I think that it is possible to make an AI that mods itself via test files, but LIB is correct in saying that the code will not change, only the parameters, however, it is still unclear to me how much text files can be used.

Vel, I also remember playing games that touted that they will use your stategies against you. Can't remember if they actually did though or what the context was. I've been playing computer games since 1977 (on mainframes) and then later on anything that I could get my hands on.

[Velociryx]

A conjectural game for our A-Instinct model

A simplified example for our AI discussion:

Let us simplify the discussion further still so we can more completely focus on our theories for Artificial Instinct

The name of the conjectural game is: The Courts of Candle’Bre (and yes, I’m making all of this up off the top of my head, so forgive me if it’s a little unpolished….I’ll polish it as we go along).

Background: Candle’Bre is a tiny kingdom, surrounded on all sides by vast stretches of mountains. Extremely isolated, and little contact with the outside world (note – for game purposes, there is NO outside contact. Candle’Bre is an entirely closed system).

Recently, the kingdom was ripped asunder by murder most foul, when the ENTIRE royal family was slain in a single night of madness and bloodshed.

Now, the Lords of Candle’Bre stand on the brink of war, each making a bid for the throne of the tiny Kingdom.

OoO

Notes: There are five Royal Houses of Candle’Bre, each having different traits and abilities that affect gameplay. Specific notes on them are as follows:

The Furies (default human player) – Led by Lord Michael Fury, this family has a long, proud history of serving as the King’s right hand. The Furies are good, noble folk, and boast the finest, best Cavalry in the realm (cavalry attacks at +2 and defends at +1) (Two special Leaders: Sir Kenna Fury – Cav. Commander, and Lord Stephen Fury – The Grand Marshal of the King’s Guard)

The Mourngrym Family – Led by Lord Flavin Mourngrym, this family is far and away the wealthiest in the Kingdom (perhaps even wealthier than the King himself, tho this has never been proved). It is whispered in dark corners that Mourngrym sold his soul to the Devil in exchange for his fabulous wealth. Whispered, because people who speak too loudly of such things often vanish without a trace. (special powers – Mercenaries can be hired at half the normal rate, chances for espionage success are 50% more likely to succeed – money to grease the wheels).

The Castillar Family – Led by Gloriana Castillar, this family’s power base is centered around religion. For sixteen generations, the men and women of the family have found their calling in the service of the Church, and as such, they hold tremendous sway over the commoners of the realm. (special powers – provincial defection: 30% more likely to succeed for this family than any other, 50% more likely if Gloriana is in an adjacent province – resistance to enemy espionage attempts at causing provinces to rebel/defect). Special Leader: Gloriana Castillar (General/Clergywoman).

The Council of Seven: The Kingdom’s “Lower House� was long ago given stewardship over a large tract of Candle’ Bre to serve as the basis of their power. Special Power = The wise men of the council are well respected by all in the Kingdom, and attempts to make war on them would generally be frowned upon by all – warring on the Council of Seven would result in a loss of Honor, AND an increase in the likelihood of rebellion in the attackers’ territories for the duration of the war. Additionally, the Councilmen have a special unit: Praetorians (Elite, bada$$ defensive unit).

The Council of Five: The Kingdom’s governing “Upper House� is made up of five mysterious, reclusive wizards, who were also given stewardship over a large tract of land during the King’s reign. Little is known of their powers, but they are greatly feared nonetheless (special powers: Choose one at game start: Pestilence, Probability, or Divination)
Pestilence: Requires X spell points to cast. Any military units moving through the target province are subject to triple normal attrition rates. Leaders suffer a 2% per turn (cumulative) chance of dying. Every turn the pestilence remains, the territory has a 3% (cumulative) chance of rebelling against its controller.

Probability: Requires X spell points to cast. The wizards may use their magic to impact the outcome of a battle (this amounts to “redoing� the battle (new random number seed) for a better or more desirable result). (x is paid on a per battle basis, no battle can be refought more than twice).

Divination: Requires X spell points to cast. The wizards may “spy on� any other house, and gain perfect information on their enemies. (is is paid on a per territory basis)


Game Concepts to introduce to the AI:
Honor: X # of Honor Points will win you the game by default.

Combat (standard A/D/M valuations for middle ages units, and giving the AI recognition of them)

Rebellion % (and other espionage abilities)

Special Leaders (attributes, and the AI should be given a desire to actively destroy them – adjustable).

Diametric tensions: Two ways to win the game: Accumulation of Honor and Controlling 2/3 rds of the provinces that make up the Kingdom.

How can we make Honor a different, and viable form of victory?

Honor is gained by:
Combat
1) Mercy (captured leaders are ransomed, not killed)
2) Captured troops are released or disarmed, not slaughtered
3) Allies are aided with gifts of money
4) Alliances are not broken (+x for every N Turns that an alliance is kept in good standing –note that these gains are doubled then subtracted if an alliance is broken….meaning the longer the alliance is in good standing before the betrayal, the deeper the impact on Honor).

Honor is lost by:
1) Getting caught using Espionage techniques (penalty depends on the severity)
2) Slaughtering captured troops/leaders
3) Attacking the Council of Seven

Thus, it should be theoretically possible to ally yourself with the Council of Seven (or anyone, really) and win by gifting them units and cash (the family that can hire cheap mercs could hire an army and gift it to their allies,for example, letting the allies do the heavy lifting, while they accumulate honor by being good and worthy allies).

More later, I gotta go to lunch, but it strikes me that this would be a MUCH simpler game to write effective AI for. Smaller in scope, less variability (we could limit units to just a handful, but I’ll flesh that out when I get back!)

-=Vel=-

[Analyst Redux]

Re: Ai

Originally posted by Velociryx
Soren's AI kicks a$$! Witness the number of people, GOOD Civ2 and SMAC players who started out on Monarch and reported that ummm...they had to tone it down to Regent or Chieftan for the first couple of games.
-=Vel=-

I wouldn't read too much into that. I was one of those who admitted to experiencing a learning curve that forced me to drop down to Regent to experiment for a game or two.

My problem was more a failure to grasp the implications of new rules than the devastating quality of the AI. I was rarely pop-rushing, for example, and it took me a bit of time to draw the line connecting the AI's startling new capacity for early growth to the pop-rushing rule. I'd have saved myself a lot of brain-work if I'd just read Absimiliard's thread on the subject--an outstanding bit of analysis there.

Between the different terrain-improvement options and pop-rushing (not to mention the fresh-water irrigation rule), early game managment is very different, and has to be relearned, but that's a learning curve issue unrelated to AI. For obvious reasons, I preferred to back down the pressure a bit to re-analyze the best early game strategy.

After I mastered the econometrics of working people to death the only surprise I had after that was the somewhat greater level of warfare aggression, most especially, the hyper-agressive counter-attack pattern. Somewhere along the line, I also realized it was a good idea to try to anticipate when I'd need a right of passage agreement, and secure it sooner, rather than later. Again, that was more learning curve adaptation to a new rule than AI difficulty.

Doing good AI is a really really hard task. As nostalgically attached as I am to the cool and quirky game line of Sid Meier (my nostalgia goes back to Pirates!), I think that the time when a single programer, or small group working above the proverbial garage, could startle us with AI ingenuity has come and gone. It just ain't as easy to do as it looks.

[Velociryx]

Economic Model:
To add a layer of complexity to the game, I’ll attempt to broadly outline the economic model here. The goal is to create a system strong enough to facilitate a good many divergent strategies, but not so overwhelmingly complex that the AI is bogged down and unable to fight its way out of the proverbial wet paper bag.

Candle’Bre is composed of Provinces. For the moment, we’ll say that there are a total of 60 provinces in the game. A player going for a military win would need 2/3rds under his control, or 40.

Each Province has an Income Valuation, expressed in a number visible on the main map (similar to Axis and Allies). This is the BASELINE per turn gold value that the Province rakes in. The sum-total of these values represents each players’ income.

Income is used for a variety of things:
1) Paying upkeep costs of your standing forces
2) Building Improvements in your Provinces
3) Hiring Mercenaries
4) Hiring Special Units
5) Training New Troops
6) Research

Each of these will be covered in turn.
1) Paying Upkeep for your standing forces: Each company of soldiers in your service requires 1g in upkeep each turn. If this fee is not paid, the unit is not lost, but simply deactivated. It may not move. It will not initiate an attack, but it WILL defend (half strength).
2) Improvements: A Province may contain no more than three (3) improvements, forcing the player to make strategic choices about what to build where. Building Barracks near the front line of a looming battle will have the effect of allowing you to churn out troops close to the action, but you run the risk of losing the territory AND your ability to defend those behind it. Available builds depend on your level of research.
3) Hiring Mercenaries. Normal combat units (ones you train yourself) have static combat values, mercenaries do not. Their combat values are floating, depending on the level of salary you provide them (the higher you set your per turn upkeep, the better they fight). Note though, that even the best paid mercenary force is no match for an elite “standing� fighting force. 1:1, the Mercenaries will lose every time. Their main strengths are a) They do not require a Barracks to train, and b) They can be hired en mass to achieve a specific objective (adjusting their salaries to whatever level of effectiveness is required). Note too, that attrition rates are higher for Mercenary units (double)
4) There are, at present, only four types of special units available (but more may be considered later: a) Sages (increase your research rate – Each Sage has a speciality – see below) b) Justicars (decrease the chance of rebellion in the province they are assigned to) c) Spy (used to conduct espionage missions) d) Taxman (used to increase the baseline revenue value of the province they are assigned to)
5) Training New Troops – This is, when the dust settles, a war game, so training troops will be where the bulk of your money ultimately goes. All the Provincial Improvements are simply a means to an end…allowing you to field a large enough army to claim the Kingdom of Candle’ Bre as your own.
6) Research – Is broken into four categories: Military Research, Infrastructural Research, Espionage Research, and Lifestyle Research. Military Research gives you better weapons (expressed as attack bonuses, and sometimes new combat abilities…also the occasional new unit). Infrastructural Research gives you additional buildings for provincial improvement. Espionage Research opens up more “sneaky stuff� to do to your opponents. Lifestyle Research increases the quality of life for those living in Provinces you control. This leads to happier people who are more able to resist attempts made by others to sow the seeds of rebellion. It also (over time) increases the per turn gold values of your Provinces.

“Techs� For the moment, we’ll leave the techs unnamed, simply referring to them as M1, M2, M3, etc(military techs), I1, I2, I3 (Infrastructure techs) and so forth. Eventually, we can get around to giving them names, but for now, this gives us a good “thumbnail reference.�

Builds: It is unknown at this time just how many total builds will be available, or what techs will make them available. For the moment tho, the ones in my head are:
Bank: Each bank pays a 1% interest on any money carried over from turn to turn. (taxmen may only be assigned to provinces with a bank)
University: Each University contributes 2% per turn to research in the fields you are currently researching in (Sages may only be hired and assigned to provinces with a university)
Barracks: Allows for the training of troops
Fortress: Provides defense bonuses in the event of an attack
Castle: Provides greater defense bonuses than a fortress in the event of an attack
Courthouse: Provides a +5% resistance to rebellion factors (justicars may only be hired in provinces with a courthouse, but may be assigned to any province that is adjacent to the province containing the courthouse, or the courthouse province itself).
Temples: Each temple constructed provides you 1 influence point per turn (influence is covered later, and has one, and only one very specific use)
Thieves Guild: spies may only be recruited from provinces with thieves guilds in them. The thieves guild itself provides counter-espionage abilities for that province, making it more difficult to conduct espionage therein.
Guilders’ Hall: Reduces the cost of non mercenary units by 10% (may only be built in provinces that contain a barracks)

· Note that no province may contain more than two special units, and the special units in question may not be of the same type!

Influence: Influence is a means of assisting your allies. Temples provide you a per turn influence total which is stored until used (ie – you don’t lose it at the end of each turn). For the price of 20 Influence points, you may interrupt any in-game attack (the clergy in your provinces send representative to the field of battle and demand a halt of hostilities….given the power of the Church, such decrees are always heeded!). Caveats to this are as follows:
1) If you use your influence to spare an ally an attack, you gain honor
2) If you cowardly use such tactics (hiding behind the coattails of the Church) to prevent an attack on one of YOUR territories, or on YOUR troops, you lose honor

Plenty more stuff of course, but this should be enough to at least provide a skeletal frame for the game. So my question now is, given the above, (and granted, this stuff is nowhere nearly as complex as SMAC or CIV), could we reasonably expect to make an AI that could be “taught� the ins and out of this (admittedly, somewhat hastily constructed) game world to such a degree that it could give a human player a decent run?

(and, if there’s interest in persuing this further, I’ll throw some additional ideas out, including an event engine to toss odd happenings into the game for unpredictability, and a few other things!) OTOH, if nobody really gets into the idea much, we’ll just let it quietly die out….

-=Vel=-
PS: Not to Analyst: I hope you're wrong about the little guy not being able to surprise us with decent AI. Admittedly, as the complexity of a game increases, so too, does the complexity involved with getting the AI to understand how to survive in the game world, but it is my fervent hope that the little guy has one or two tricks left that the monied folk have not thought of yet....

-V.

[Heliodorus]

Trying to follow along with Vel, here are some questions:

Is your map static, as in Axis and Allies?
Is your mercenary pool also static in a game? (I imagine a Magic: TG deck in which you know who's in there, but not who will be available next turn...).

I'm not a programmer, so it would be ludicrous for me to say it can or can't be done. However, I can see from the human perspective a major prioritization scheme occurring around the idea that you can only build 3 infrastructure builds per province. It strikes me that strategic implications begin there, and filter down into the decisions about combat, about influence, honor, and whatnot.

Interesting idea so far.

[Velociryx]

Hey man! Glad you're liking the idea so far! I'm currently thinking about even more builds, and perhaps a few "wonders" as well, to make the strategic placement a bit dicier (ie - Let us suppose that if you built six universities, you could then construct an Great Library/Great Academy kinna build....it would cost about triple what a University would, and would count as a University for purposes of answering the question "what all has been built in this province?" but would provide additional benefits as well--in this case, perhaps it would provide a 5% bonus on accumulated research in all categories currently being researched in, and, if you weren't doing any research in a particular category, it'd still give you a flat number of research points IN that area....something like that).

I envision the map as being flat and fixed, like Axis and Allies, and the Mercenaries being an infinite (and somewhat generic pool), with the following exception:

Sometimes, the games "event engine" will make special Mercenary units available to you for that turn only. If you hire them then and there, you can bring them into your fold. If you can't afford it, the event passes, and you lose your opportunity. This hybridized approach would allow for rapid, "in bulk" hiring of cheap troops along with highly skilled specialty units (per your M:TG example).....Now, if we can pique the interest of a programmer or two....

-=Vel=-

[Analyst Redux]

Originally posted by Velociryx
PS: Note to Analyst: I hope you're wrong about the little guy not being able to surprise us with decent AI. -V.

Actually, so do I.....

[Libertarian]

Libertarian, why [can't the computer take meaning from events or behavior]?

Because it has no ontological reference frame through which meaning can be derived. In other words, it is unable to formulate any sort of new concept other than what has been programmed into it. A computer is not analogous to a brain. If it were, it would rewrite its own code as its experience increased.

Take, for example, the famous scenario in which the computer sends a fleet of attackers to your weakest coastal spot. You can simply move a sufficiently threatening force around and around the coast just ahead of the fleet such that the AI will behave as though you have a strong and thoroughly fortified border without a hole anywhere. It will behave this way even though you are threadbare all the way around, except the one mobile spot.

Unless it has been specifically programmed to respond a certain way to your behavior, it is powerless to do anything about it.

There simply is no such thing as meaning outside a cognitive conscience.

[Velociryx]

More on our conjectural game....

Units:
The following units are available at game start:

Skirmisher/Scout – Lightly armored infantry based unit – The only foot unit that can keep pace with cav (2 moves)
Archer – Ranged attacker, lousy at hand to hand (1 move)
Cavalry – Speed unit, devastating for their ability to execute flank attacks (2 moves)
Pikemen – Provide defense against cavalry, extra vulnerable to archers (1 move)
Infantry – The grunts of the army that do most of the real fighting (1 move)
Siege Engine – Abstracted to represent a variety of engines, the presence of these units do not affect combat per se, but weaken the defensive modifiers gained by enemy fortresses and castles. (1 move)

*Technological Advance through the Militaristic branch of the tree may allow for other types of units as the game progresses.

Combat:
Each unit represents a company of 100 Men. Units can be stacked and combined into groups with the following considerations:
1) The group moves at the rate of the slowest member of the group.
2) The group fights as a cohesive unit (ie – all a/d values are added together).
3) Losses taken by the group are spread out as evenly as possible
4) In the absence of a leader, no more than 4 units (400 men) can be stacked together.
5) With a leader, this number jumps to eight (8) units as a maximum.

Combat occurs in rounds, and in the following order:
1) Ranged units fire (attacker/defender)
2) Losses are taken
3) Cavalry Flanking maneuvers/charges are performed (morale check for those receiving the charge)
4) Losses are taken
5) Ground troops attack
6) Losses are taken
7) Morale Check and Repeat.

Note that a charging/flanking cavalry unit may only attack “every other round.� (spending the off rounds re-grouping and gaining position).

Combat continues each round until:
1) One side is eliminated
2) One side’s morale breaks

If morale breaks, some portion of the breaking army is captured, and the rest retreat to the nearest friendly province. Captured units may either be ransomed or killed at the player’s option.

Weak Units:
Can be restored to full strength by spending a turn not moving, attacking, or defending, and for a gold value equal to x, where x is the % cost of the unit (ie – let’s say you have a company of Infantry that’s down to 50% strength. It’d cost you 50% of that unit’s full price to bring it back to full strength).

Unit Experience:
The REASON for wanting to restore weakened units is to keep their experience. Every three victories the unit participates in, they gain a “level� of experience, expressed in terms of higher unit morale (+10%). As the unit is weakened through attrition and later replenished, the raw recruits DO negatively impact morale, but only 1% per 20 Recruits added to the unit.

Leaders: Have a combat and morale enhancing attribute (both expressed as percentages). Units or groups with leaders at the head do more damage pound for pound, defend better, and are less likely to break in battle.

Elite units (150%+ morale) that win battles have a 20% chance of spawning a War Leader (the elite unit is lost, the War Leader is gained).

Mercenary units can NEVER generate War Leaders.

More later, if interest in this conjectural game continues….almost time to go home!

-=Vel=-

[Arrian]

Hey Vel, I have a idea for you. Feel free to use or discard

In the Civ series, research was simply a matter of choosing what percentage of your empire-wide trade went to science. What if you could (if you wanted to) adjust it province-by-province, such that if you had a province with good trade and a great library type wonder thingy built in it, you could set it to 100% research - it would be the research center for your empire. Thus, you could maximize something like the Civ "SSC" while the rest of your empire paid the bills. Of course, that's putting all your eggs in one basket, so you'd damn well better protect that province.

Just a thought.

-Arrian

[Libertarian]

I think it's a great idea. Naturally, I dislike being unable to model a libertarian government, but your idea would go a long way toward helping in that regard.

[Velociryx]

Oh! I like it! And in fact, you could just have a toggle switch for the player to employ.

If you just wanted to control your research at the "top level" you flip the switch and set your percentage. OTOH, if you wanted "Full, Province-Level Control" you flip the switch, but then must micromanage each individual Province. VERY cool idea!

-=Vel=-
(glad to see people getting into the spirit of the conjectural exercise....and who knows, if we generate some real excitement talking about this game that exists only on this thread, perhaps some brave soul will offer to put some code together and see what happens!)

I think the key thing is, if we stick to a relatively simple concept and design, we should be able to construct a game which has:
1) Good, player moddable AI
and
2) Lots of replayability

I dunno about the other folks reading this, but in my head, I'm already contemplating various approaches to the game using all five of the houses outlined...lol

[Arrian]

Originally posted by Velociryx
Oh! I like it!

I dunno about the other folks reading this, but in my head, I'm already contemplating various approaches to the game using all five of the houses outlined...lol

First off, I'm glad you like it. The last bit there is because you're a strategy game junkie like the rest of us. Of course your mind is already contemplating how you would play.

What I would really like to see is what you (and many others here) would do with a good editor for Civ III, which is supposedly on the way. Sure, the AI would still be dumb, but I've played games that had dumb AI all my life.

-Arrian

Edit: I just saw your post, Lib. Thanks. How would you model a libertarian gov't in a game such as this, btw?