Artifical Intelligence for World in Flames

A forum for the discussion of the World in Flames AI Opponent.

Moderator: Shannon V. OKeets

Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

ORIGINAL: Rotor
DECISION MAKER: Field Marshalls (any country)
a case can be made for letting the JCS deciding globally of the necessary reorganizations
PROBLEM: Reorganize
CHOICES: Which units to reorganize with wich hq
INFORMATION NEEDED: expected remaining length of the turn, expected weather (reorg planes of not), which units could effectively be reorganized, what are we going to do in the coming impulses : advancing, retreating, heavy fighting.., do we expect a situation where we would need the hq special abilities (emergency supply & offensive/defensive support irc), would inverting the hq put it in danger (case of retreat)..
imho it will be the toughest part to do right. Grand strategy looks easy in comparison. But maybe it's just me...
I do not think figuring out reorganizations will be that hard.

First of all there are only a few unit types that can reorganize units (HQ, ATR, TRS). So, the question becomes "should this HQ/ATR/TRS reorganize units now?" Rather than look at all of the HQs at the same time, go through them sequentially. This means that even the Germans in the thick of Barbarossa only have 3 or 4 binary (yes/no) decisions to make.

There will almost always be units that could be reorganized so the criteria have to trade off: (1) the benefit of turning units face up versus (2) saving the 'reorganizer' for other duties or reorganizing during a future impulse.
Steve

Perfection is an elusive goal.
rotor911
Posts: 38
Joined: Mon Dec 13, 2004 1:59 pm

RE: Artifical Intelligence for World in Flames

Post by rotor911 »

Now, you could take one of the missions (say, port attack) and figure out the criteria for setting the #of planes, air-to-sea factors, and # of fighters to send.
ok. Please forgive me for the long post but I couldn't come with something shorter. Also it's a first try so of course it can be improved.
First some assumptions :
-weather is good else we don't bother with a port attack -> we don't count enemy scs and submarines since our carriers should always be able to choose air to sea combat (unless heavily surprised..)
-we assume that if friendly land based aviation was in range of the target we wouldn't risk our carriers so they're out of the picture.
- composing a TF is as much a feasibility study as an actual preparation : so if the Admiralty estimates that it lacks the necessary resources, it just aborts the procedure, in effect saying no to the JCS.
so
- first calculate the safest path (less enemy LBA & patrolling fleet) unless the mission is critical, then choose the shortest one (more ships to bring)
- if the target is a minor port, select only ships which can reach the 3 box of the selected sea zone, the 5 box if the target is a major port (2 & 4 respectively if the mission is critical)
- calculate how many air-to-sea factors you need to cause at least 1 X given the number of enemy ships in the target port : for instance, if there are 6 ships in the port, you need at least 5 air-to-sea factors
- select enough carriers to have these factors available. If you haven’t them, abort now.
- add the anti-aircraft factors of the enemy ships in the target port and those of the uninverted anti-aircraft land units in the hex (or adjacent hexes irc). Calculate the average result against the number of bombers we selected so far (we’ll need a table here to spare us recalculating each time the value of 1/3 or –2/5).
- For each 5 points of damage, add 1 carrier with air-to-sea factors. . If you haven’t them, abort now
- Add-up all possible intercepting or cap fighters air-to-air factors. Add enough carriers with air-to-air factors to at least achieve parity. . If you haven’t them, abort now
- Add up the maximum number of land-based enemy planes which could attack the TF at a time : add enough cruisers, light cruisers and, if you have many or if the mission is critical, battleships, to boost the TF anti-aircraft value of the TF so it gives at least 1/1 or better against such a number. . If you haven’t them, abort now
- Estimate the worst enemy carrier TF which could intercept you (if playing with hidden task forces, it could be tricky..) : add carriers until you have at least the parity in number of carriers. . If you haven’t them, abort now
- Now sum the build value of all the ships which were selected : if the total is greater than the importance of the mission (which would then be the maximum ships build points which the JCS would accept to lose in the affair) , then abort.
User avatar
Greyshaft
Posts: 1979
Joined: Mon Oct 27, 2003 1:59 am
Location: Sydney, Australia

Italian Timing for DOW v.02

Post by Greyshaft »

DECISION MAKER: Grand Strategist for Italy.

PROBLEM: When to make a declaration of war and should it be against the Allies or a neutral?

CHOICES:
  • a fixed date (eg Mar/Apr 1940) OR
  • after a specific event (eg Surrender of France) OR
  • on identifying an opportunity (eg uncommitted British&French Fleet in Med < 80% of Italian fleet attack factors)
  • Attack Allies
  • Attack Neutral (Yugoslavia / Greece / Turkey / Spain)
INFORMATION NEEDED:
  • Quantification of what surplus Italian naval/land/air forces “X Force” are available for invading other countries.
  • Allied Land: Quantification of what Allied air/land forces (France in Alps and British in Egypt) are available to counterattack Italy.
  • Allied Fleet: Quantification of what Allied naval forces are available to prevent our naval missions in the Med.
  • Political effects of our DOW
CRITERIA:
  • Game start selection of a particular strategy (east to Suez / west to Gibraltar / attack Spain / ???) which weights other factors
  • Political events (Fall of Poland completed / has Germany or Allies DOW or conquered low countries implying French forces tied to defensive role and unable to spare a Naval action in the Med)
  • Casualties (Does French/British land & air casualties outnumber German casualties by 2:1 in unit count?)
  • French Land: How many French units are available for combat against Italy ie not garrisoning Maginot Line?
  • British Land: Is Malta / Cyprus / Egypt / Gibraltar vulnerable to invasion by X Force with > 50% chance of success?
  • Generic Land: Are we vulnerable to an enemy counterattack eg Egypt => Libya?
  • British Fleet: Does British fleet surface combat factors in Med ports inc. Gibraltar > 80% of X Force surface combat factors?
  • Other Land: Is X Force value > 150% of corresponding Yugoslavia / Greece / Spain / Turkey forces (not total forces – just the forces free to counterattack our invasion)?
  • Political: Will delaying the DOW produce a significant US entry effect - eg DOW in 1941 instead of 1940


Feedback please, before I start working on v.03 [:)]

/Greyshaft
Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

Let me paraphrase your criteria into a structure that the AI could use. Again, if I make mistakes I apologize and you should let me know.
-------------
DECISION MAKER: Admiralty
PROBLEM: Choosing task force (TF) for a port attack
CHOICES: Which ships to give the mission [list all ships that could participate and each is a yes/no chioce]
INFORMATION NEEDED: land based aviation and carriers that the enemy can use against the task force, weather, importance of the mission, list of ships still available, optionally starting harbour[??]
CRITERIA
(1) If weather is not good, then no port attack.
(2) If land based NavAir available, then use them instead of sending carriers. This means no port attack.
(3) If insufficient [?] resources available, then no port attack.
(4) Choose sea box for attack depending on major or minor port and whether the mission is critical. Given the sea box, choose the best path to get there, avoiding the enemy if possible. If enemy is likely to be encountered, increase the TF to cope with the naval combat). Evaluate air-to-sea factors that can make the sea box; if too few, then no port attack.
(5) Estimate how many carriers will get through after air-to-air combat. If too few, no port attack.
(6) Calculate expected losses due to anti-air attack. If too few air-to-sea factors are going to get through, no port attack.
(7) Estimate enemy attack on TF next impulse and add naval units (Cruisers, Light Cruisers, and/or Battleships) for defense. We need to defend against both air and surface atacks. Compare expected losses (in buld points) inflicted on enemy versus expected losses we will take. If ratio (or magnitude of difference) exceeds what is justified by mission criticality, then no port attack.
-----------------

Each numbered step of the criteria is a pass/fail. If any step is fail, then there is no port attack and we don't need the rest of the steps. I think of these steps as rules where sometimes the program has to run a calculation or procedure before it can evaluate the rule. Rules are usually written as: "If ..., Then ...".

I changed the order of air-to-air and anti-air combat from what you gave (steps 5 and 6).
I do not know what you mean by "starting harbour".
I do not know how you measure 'insufficient' in step 3.

This is a good start on defining task forces for port attacks. It would appear that no one should ever do port attacks!

What we will most likely do is make a lot of the pass/fail decisions probabilities so that sometimes the AI takes risks. Avoiding combat unless things are just right makes the AI too inactive and predictable. It then becomes easy to defeat the AI by simply putting up a good defense (deterents) everywhere and concentrating any forces left over as a hammer to beat on the weakest part of the AI's position.
Steve

Perfection is an elusive goal.
rotor911
Posts: 38
Joined: Mon Dec 13, 2004 1:59 pm

RE: Artifical Intelligence for World in Flames

Post by rotor911 »

I do not know what you mean by "starting harbour".
I do not know how you measure 'insufficient' in step 3.
"starting harbour" : the port where will be composed the TF. Composing several TF in several ports and assigning them a meeting point is maybe uselessly complicated (and consumes several naval moves). Generally, the navies are concentrated in some major ports and it's not a problem.
"insufficient" : less than the calculated number of air-to-sea factors (or ships or..) required to do the task. Or than a percentage of the calculated number, the said percentage depending of the importance of the task.
Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

ORIGINAL: Rotor
"starting harbour" : the port where will be composed the TF. Composing several TF in several ports and assigning them a meeting point is maybe uselessly complicated (and consumes several naval moves). Generally, the navies are concentrated in some major ports and it's not a problem.

"insufficient" : less than the calculated number of air-to-sea factors (or ships or..) required to do the task. Or than a percentage of the calculated number, the said percentage depending of the importance of the task.
Ok.

Then there should be something about whether the TF is all starting in the same place or starting in different places and joining up at the destiation.

Insufficient has the same meaning as in the later steps: enough air power to achieve the desired results.
-------------------
As you probably are beginning to see, I break everything into little pieces and then try to define each piece as precisely as possible. Eventually, the program is going to have to 'look' at the units and the map and generate a number for each little piece. The AI will then use the numbers in the 'If' portion of a rule to decide whether to go on to the 'Then' portion or not.

There will be small procedures for combining the numbers and boolean logic as well (If A and B but Not C, Then ...). In order for this to work, I have to be able to translate the thought processes we use when we play the game into the rules, numbers, and boolean logic for the AI. Actually, I view this as the most interesting part of the project. Not necessarily the most important, just the most interesting.

There are many things about developing MWIF that are important, and most are interesting. Regretably, many tasks are tedious and require a lot of attention to make sure all the details are done perfectly. I work on the last as long as I can, but once I lose my sharpness, I skip to things that are a little less demanding.
Steve

Perfection is an elusive goal.
User avatar
c92nichj
Posts: 345
Joined: Fri Jan 14, 2005 1:15 pm
Contact:

RE: Artifical Intelligence for World in Flames

Post by c92nichj »

ORIGINAL: Shannon V. OKeets
It would appear that no one should ever do port attacks!

Which is not that bad rule of thumb, Port Attacks are usually not that effective, escpecially since the port attack phase comes before naval movement which means that the opponent have the chance to leave the port before he is attacked. Usually it effectiveness is best during a suprise impulse.
User avatar
Neilster
Posts: 3002
Joined: Mon Oct 27, 2003 1:52 pm
Location: Devonport, Tasmania, Australia

RE: Artifical Intelligence for World in Flames

Post by Neilster »

Hello Steve. I'm very happy with the recent progress and best of luck.

If you haven't already, contact Incy man. There was an old thread on this and he was talking about 15-20 (from memory) intelligent agents co-operating to form the AI and the cat sounded like a real gun.

I've played a fair bit of WiF and other wargames and I'm finishing my computer science degree this semester with 2 years of AI. My instincts tell me that the AI will have to be as flexible as possible and that is probably not best attained by a rule heavy approach.

I've got some ideas if you're interested.

Cheers, Neilster
Cheers, Neilster
Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

ORIGINAL: Neilster

Hello Steve. I'm very happy with the recent progress and best of luck.

If you haven't already, contact Incy man. There was an old thread on this and he was talking about 15-20 (from memory) intelligent agents co-operating to form the AI and the cat sounded like a real gun.

I've played a fair bit of WiF and other wargames and I'm finishing my computer science degree this semester with 2 years of AI. My instincts tell me that the AI will have to be as flexible as possible and that is probably not best attained by a rule heavy approach.

I've got some ideas if you're interested.

Cheers, Neilster

I am. Could you send me (Steve@PatternDiscovery.us) your email address?
Steve

Perfection is an elusive goal.
User avatar
Neilster
Posts: 3002
Joined: Mon Oct 27, 2003 1:52 pm
Location: Devonport, Tasmania, Australia

RE: Artifical Intelligence for World in Flames

Post by Neilster »

Done
Cheers, Neilster
herwin
Posts: 6047
Joined: Thu May 27, 2004 9:20 pm
Location: Sunderland, UK
Contact:

RE: Artifical Intelligence for World in Flames

Post by herwin »

Actually, an actor-critic approach where the AI learns the player's styles might be the most effective.
Harry Erwin
"For a number to make sense in the game, someone has to calibrate it and program code. There are too many significant numbers that behave non-linearly to expect that. It's just a game. Enjoy it." herwin@btinternet.com
User avatar
Manack
Posts: 30
Joined: Mon Aug 29, 2005 4:47 am

RE: Artifical Intelligence for World in Flames

Post by Manack »

What would be great for the extreme geeks amongst us would be to have the AI implemented as much as possible via a interperated script which could be user edited. Events instructing the AI to react certain ways in certain conditions, ratios of CVs, BBs, and CA in tasks forces which are assigned different objective. At least be able to have access to variables which weigh how the AI behaves in the game. For example being able to mod the amount of fighters it produces or the land forces Germany guards Frances beaches with.

It would be great if you were able to mod the AI to the extent that you could pursue a different Grand Strategy which might be a bit radical and probably not implemented by Matrix. Such as a German attempt on taking India or US invasion through the Balkans.

That way the user community could continually improve the AI constantly fine tuning it and experimenting. The best changes floating around the forums could always be reabsorbed back into the core release during the patch process.

This would greatly enhance the solo replay value of the game as people trade AI enhancements as well as allow the community to continuing to improve the game way past the time when it's commercially possible for Matrix to do so.
Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

ORIGINAL: Manack
What would be great for the extreme geeks amongst us would be to have the AI implemented as much as possible via a interperated script which could be user edited. Events instructing the AI to react certain ways in certain conditions, ratios of CVs, BBs, and CA in tasks forces which are assigned different objective. At least be able to have access to variables which weigh how the AI behaves in the game. For example being able to mod the amount of fighters it produces or the land forces Germany guards Frances beaches with.

It would be great if you were able to mod the AI to the extent that you could pursue a different Grand Strategy which might be a bit radical and probably not implemented by Matrix. Such as a German attempt on taking India or US invasion through the Balkans.

That way the user community could continually improve the AI constantly fine tuning it and experimenting. The best changes floating around the forums could always be reabsorbed back into the core release during the patch process.

This would greatly enhance the solo replay value of the game as people trade AI enhancements as well as allow the community to continuing to improve the game way past the time when it's commercially possible for Matrix to do so.

I gave an answer to another post that applies to what you are suggesting. See #82 in PBEM Standing Orders.
Steve

Perfection is an elusive goal.
User avatar
Froonp
Posts: 7998
Joined: Tue Oct 21, 2003 8:23 pm
Location: Marseilles, France
Contact:

RE: Artifical Intelligence for World in Flames

Post by Froonp »

What Manack wrote reminded me the good old SSG games (from the Battlefornt Series to the Carriers at War series), and compelled me to chime in.

These old SSG computer wargames had the best AI I encountered in computer wargames, because it was designed like Manack described.
The whole system was called "Warplans" if I remember correctly.
It ws designed with the user able to input things in it, and with the user ability to design "warplans".
And at that time there was no Internet, no worldwide communication, and the thing worked.
SSG even had a magazine issued (was call Run5) where you had scenarios with warplans you could add to the games.

Having the old SSG chaps working now for Matrix, and having Matrix doing a computer version of WiF FE, I thought that I should at least tell you : Why not ask the SSG guys if they can adapt their "Warplan" AI to WiF FE ?
Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

ORIGINAL: Froonp
What Manack wrote reminded me the good old SSG games (from the Battlefornt Series to the Carriers at War series), and compelled me to chime in.

These old SSG computer wargames had the best AI I encountered in computer wargames, because it was designed like Manack described.
The whole system was called "Warplans" if I remember correctly.
It ws designed with the user able to input things in it, and with the user ability to design "warplans".
And at that time there was no Internet, no worldwide communication, and the thing worked.
SSG even had a magazine issued (was call Run5) where you had scenarios with warplans you could add to the games.

Having the old SSG chaps working now for Matrix, and having Matrix doing a computer version of WiF FE, I thought that I should at least tell you : Why not ask the SSG guys if they can adapt their "Warplan" AI to WiF FE ?

There is an English expression "being a wet blanket", which is my role for this idea.

I think it is universally accepted that creating an AI for a wargame is difficult to do, that they are almost always embarrassingly stupid, and that it would be really surprising and quite wonderful if MWIF included a good AI opponent.

At this point in the development of MWIF I think I can write a good AI. Time will tell whether I am right or wrong. However, I have a pretty clear vision of how it will work and that includes having a very good understanding of the nuances of WIF (the rules). These fundamental underpinnings do not lend themselves to review and tweaking by people who are not familiar with how they all tie together. In a nutshell, WIF is complicated and the MWIF AI opponent code/logic/rules/data structures/procedural elements will be too. Designing it for parsing and manipulation by others is not on my task list. And since my task list is already long and quite challenging, it is not going to become part of my task list either. As I said earlier - I am being a wet blanket.
Steve

Perfection is an elusive goal.
User avatar
Manack
Posts: 30
Joined: Mon Aug 29, 2005 4:47 am

RE: Artifical Intelligence for World in Flames

Post by Manack »

I've done a little bit of prolog and it probably shines in my recursive intensive approach to an AI problem. I know this is a little more computer resource intensive than other approaches but I thought I'd throw it into the ring anyhow.

Decision Maker: Admiralty

Problem: Convoy raiding Decision making.

Choices: One or more sea areas in range of a naval unit or plane with naval factors.

Information Needed:
  • Aggression policy on convoy raiding
  • Weather
  • Units available to reach sea area
  • Enemy Convoys in a sea area
  • Convoy utilisation in a sea area
  • Enemy naval escorts in a sea area
  • Enemy air power in a sea area
  • Enemy units able to intercept raiders enroute to convoys
  • Enemy units able to counterattack raiders next impulse.
  • Seabox advantage
  • Available Naval moves

Criteria:
Preferably, the AI should combine the likelyhood of finding an enemy convoy and the predicited damage to loss ratio to derive a value which can be ranked for every possible combination of sea area(s) and convoy raider makeups within the available number of naval moves.

In ranking a attack the AI would need to consider the likely makeup of the opposing sides in a naval battle (multiple if interceptions/counter attacks are likely) and make a prediction about the likely type of battle (sub, surface, air) and the likely outcome of the battle.

This outcome ranking would then be modified by the likelyhood of it occuring with the AI preferencing positive outcome battles with a likely occurance.

If the best ranked combination meets the convoy raiding aggression policy criteria then the order is executed/recomended.

Other:
A utility such as this to test opponent vulnerablility should be part of the decision making process in choosing if a land, naval or combined action should be taken.
User avatar
Manack
Posts: 30
Joined: Mon Aug 29, 2005 4:47 am

RE: Artifical Intelligence for World in Flames

Post by Manack »

ORIGINAL: herwin

Actually, an actor-critic approach where the AI learns the player's styles might be the most effective.

I would love to see a well done actor-critic wargame done one day. But I can see how it could go horribly wrong. An AI that might seem balanced in beta might slowly become irrational once released. Or a bug may surface in the program which can only be reproduced due to a particular evolutionary path of the AI making it almost impossible to debug.

AI actor-critic also only works very well when a action can be measured and there is limited interdependence between the actions. In Wif lots of stuff can't be measured and that's not even counting the actions the AI doesn't attempt which can never be accurately measured. How can you measure something like placement of ground units on the Russian front when interelated factors factors like weather and airpower are what turns the tide.

Lots of effort would be required for fine tuning. What if for example the AI got unlucky and rolled a one 5 times in a row for a 3:1 blitz attacks it might then decide to not try those attacks again. You would need to develop some very robust automated test cases and go over the AI to weed out those problems.

I think in the limited avenues where an actor-critic approach can be easily implemented in WIF would end up being no more effective than a simple rules based/table lookup anyhow. 1:2 attack bad 3:1 attack good.

But still... I'd love to see it done. [:)]

Shannon V. OKeets
Posts: 22165
Joined: Wed May 18, 2005 11:51 pm
Location: Honolulu, Hawaii
Contact:

RE: Artifical Intelligence for World in Flames

Post by Shannon V. OKeets »

ORIGINAL: Manack
ORIGINAL: herwin
Actually, an actor-critic approach where the AI learns the player's styles might be the most effective.
I would love to see a well done actor-critic wargame done one day. But I can see how it could go horribly wrong. An AI that might seem balanced in beta might slowly become irrational once released. Or a bug may surface in the program which can only be reproduced due to a particular evolutionary path of the AI making it almost impossible to debug.

AI actor-critic also only works very well when a action can be measured and there is limited interdependence between the actions. In Wif lots of stuff can't be measured and that's not even counting the actions the AI doesn't attempt which can never be accurately measured. How can you measure something like placement of ground units on the Russian front when interelated factors factors like weather and airpower are what turns the tide.

Lots of effort would be required for fine tuning. What if for example the AI got unlucky and rolled a one 5 times in a row for a 3:1 blitz attacks it might then decide to not try those attacks again. You would need to develop some very robust automated test cases and go over the AI to weed out those problems.

I think in the limited avenues where an actor-critic approach can be easily implemented in WIF would end up being no more effective than a simple rules based/table lookup anyhow. 1:2 attack bad 3:1 attack good.

But still... I'd love to see it done. [:)]

I considered the actor-critic approach an extremely poor idea for WIF. It usually is applied either in games where similar situations arise frequently (e.g. Go) or to a repetitious set of games (e.g., playing 40 chess games in a row). In both cases, the starting premise is that a lot of data is provided by the opponent that can be analyzed. WIF does not give the AI the luxury of a learning period.

If Germany screws up the first two turns of Barbarossa, there is no recovery. The same can be said for almost every other major power in every scenario. The AI has to play well from the very start. Even if the first couple of turns were hardcoded, the AI would not learn enough about the enemy during those 2 turns to predict the enemy's actions during the 3rd. You don't know any of the units he has in production - for just one obvious example. Also the dynamics of the war have great variability. From fighting a 100% land battle in China, Japan turns to building an expansive naval empire. How the AI is suppose to predict this based on the Japanese actions early in the game is beyond my understanding. There is very little that is repetitous in WIF - every thing is in flux.

If you would like to get more involved in designing the AI for MWIF, send me an email (Steve@PatternDiscovery.us)
Steve

Perfection is an elusive goal.
User avatar
Froonp
Posts: 7998
Joined: Tue Oct 21, 2003 8:23 pm
Location: Marseilles, France
Contact:

RE: Artifical Intelligence for World in Flames

Post by Froonp »

There is an English expression "being a wet blanket", which is my role for this idea.

I think it is universally accepted that creating an AI for a wargame is difficult to do, that they are almost always embarrassingly stupid, and that it would be really surprising and quite wonderful if MWIF included a good AI opponent.

At this point in the development of MWIF I think I can write a good AI. Time will tell whether I am right or wrong. However, I have a pretty clear vision of how it will work and that includes having a very good understanding of the nuances of WIF (the rules). These fundamental underpinnings do not lend themselves to review and tweaking by people who are not familiar with how they all tie together. In a nutshell, WIF is complicated and the MWIF AI opponent code/logic/rules/data structures/procedural elements will be too. Designing it for parsing and manipulation by others is not on my task list. And since my task list is already long and quite challenging, it is not going to become part of my task list either. As I said earlier - I am being a wet blanket.
I'm sorry, I do not know what the english expression "being a wet blanket" means [X(]
OK, I must be dumb, I found what it meant [:D]
You're not a wet blanket, I am happy you are making the AI [:D]
User avatar
Anendrue
Posts: 817
Joined: Fri Jul 08, 2005 3:26 pm

RE: Artifical Intelligence for World in Flames

Post by Anendrue »

Unfortunately I can not help with your AI due to No Compete and NDA.

Your use of "decision makers will be familiar to all WIF players: Grand Strategist, Commander in Chief, Manufacturing Council, Foreign Liaison (coordination with allies), Joint Chiefs of Staff, Admiralty, Air Marshall, and Field Marshalls. Note that there may be several Field Marshalls with different ones assigned to different theaters of operation" seems very similiar to HOI2 and that would be great.

Still for location specific attack and defense values a possible solution might be to create a graduated extract of combat results tables of minimum defense and adding in additional units until max stack limits are reached. This could be extracted into a workable set of tables looking for optimum defense which theater commanders use to assist for specific location defenses.

However this would require the creation of a database containg all the combat/terrain variables and unit statistics iterrated through umpteen millions of combat results storing and averaging results resulting in the graduated tables. Quite a project in and of itself. So unless someone not under a legal hammer could take this on .... [;)]
Integrity is what you do when nobody is watching.
Post Reply

Return to “AI Opponent Discussion”