List of possible AI improvments

[lavisj]

Am I to understand that this type of probing the community for making a decent AI was not done before release of the game? If so I will have to wait for that AI patch to be made before I can buy the game.

[Jimmer]

ORIGINAL: Murat


[3] England should set up in the Channel and only blockade fleets that are a threat (have corps in the port). The goal is destruction of enemy fleets and keeping the channel open to you and denied to your enemies.

This won't work. Someone suggested it as a strategy once, and we set it up. France simply tried to move all five fleets into sea zones all around GB (the main island). GB intercepted 2 on the way, and forced two more to retreat (one went to port, but the other retreated to a sea area that had access to Ireland). The final fleet won its battle, and landed a corps somewhere in the middle of GB. The ONLY saving grace for GB was that there were only 6-8 factors in each corps, so the one that landed only had 8. But, still, GB had a major set of battles to fight, with no obvious hope of victory.

It was worse than that. Because GB attacked at the end, and three French fleets were now in port, with the British navy scattered, he was forced to either go first and risk yet more factors landing, or go last and allow the French to go back to safety. All the way around, we concluded it was far too risky until GB had a bigger army. Any French corps with 10 boys in it would be a bad thing to leave laying around England too long.

[lavisj]

ORIGINAL: Jimmer

Why not? That's how Napoleon lost at Waterloo: He put Ney in charge of the Guard with orders to wait until he needed them. Ney sent them into combat 1000-2000 at a time. When Napoleon wanted to "commit the guard", there were few left available to him.

I think this is why Ney's third stat is a 1. At Waterloo, he was in charge of two corps, and didn't do so well.

Actually I think that Ney's rating reflects more his performance during the 1813 Campaign.

As for the Guard at Waterloo, Ney is not the one who commited it, but Napoleon itself, and it was committed at different points:
1. Plancenoit (Young and Old Guard). And set up as a local reserve.
2. Placed back in general reserve (minus a division of Young Guard)
3. During the main battle for the last ditched effort at the end.

But you need to place yourself in Napoleon's shoes. The Guard was the last reserve he had available at this point, committing it in its entirety to win the battle would have been great if it had worked, but if it had failed, there was nothing left to cover the retreat of the army.

The commitment of the guard actually seems appropriate at Waterloo. And it was not a lone raider as it was part of a 6 Corps battle (on the French side).

[bresh]

Gibraltar has no value beside a safe port in this version of the game. [France/GB dont seem to loose their dominant abilities]
 
Fleets, beside GB, fleets AI should not intercept unless weaker force [Keep in mind it can order this for stacks to].
 
AI's shouldnt pay to many $ for supply if they only have a risk militias.
 
AI's shouldnt have their corpse Freeze far away while they dont use them. This is especially seen for turkey and spain, who when they take portugal/egypt keep the force "foraging" in the minor.  Garrison maybe leave 1 corpse but not 3+.
 
AI GB use its corps more wisely. Its factors come costly so better wait and combine with minors and or "friends".
 
 
Regards
Bresh

[Odysseus]

Some interesting points here. As avid a fan of EiA as I am, I have to admit that I'm a tad surprised at the lack of diplomatic interaction available to the single player in this game that is, to a large extent, driven by diplomacy. CivIV and the Total War-series, even HoI2 (though much less developed), has some form of interface where you can assess the attitudes of the AI faction towards yourself and to others (like the incredibly accessible Civ4 diplomacy screen showing wars, alliances etc with colored lines between the factions). Also there is the possibility of actually cutting deals with the other factions - and seeing their reactions to the different proposals you make. It's not always very accurate or even logical, but it's there. I would have expected something along those lines for this game...

[zaquex]

I think this game needs a layered AI system incorporating strategies, for example something like this:

Politics layer

Agendas

reactions modifiers (alliances, stances, opposing strategies, history, national stance)

This layer will influence declarations of war, support for neutrals, requests and responses for alliance proposals etc. And it will decide Strategies.

Strategy layer

strategy/objectives

military reaction modifiers (control of areas, placements of armies and fleets etc)

Decides operations wardeclarations, support for neutrals etc

Operational layer

decides destinations of armies and fleets, placement of depots and reinforcements

Tactical layer

decides what route to use and if to engage in combat and what tactics (chit) to use and if to forage, use forcemarch etc


This is ofc very simplified but it should give an idea of how I think the AI of EiAM needs to work and if the Agendas and strategies would be modable there is no limit on how successfull this game could become covering basicly any conflict in a substantial historic period.

There is even a possibility for the game to cater for characteristics of different nations and even leaders (cautious or reckless, defensive or offensive what chits the prefere etc based on historical facts) in the tactical layer a possibility the board game never had.



Regards

zaq

[Murat]

ORIGINAL: Jimmer

ORIGINAL: Murat

[3] England should set up in the Channel and only blockade fleets that are a threat (have corps in the port). The goal is destruction of enemy fleets and keeping the channel open to you and denied to your enemies.

This won't work. Someone suggested it as a strategy once, and we set it up. France simply tried to move all five fleets into sea zones all around GB (the main island). GB intercepted 2 on the way, and forced two more to retreat (one went to port, but the other retreated to a sea area that had access to Ireland). The final fleet won its battle, and landed a corps somewhere in the middle of GB. The ONLY saving grace for GB was that there were only 6-8 factors in each corps, so the one that landed only had 8. But, still, GB had a major set of battles to fight, with no obvious hope of victory.

It was worse than that. Because GB attacked at the end, and three French fleets were now in port, with the British navy scattered, he was forced to either go first and risk yet more factors landing, or go last and allow the French to go back to safety. All the way around, we concluded it was far too risky until GB had a bigger army. Any French corps with 10 boys in it would be a bad thing to leave laying around England too long.

How did a fleet land a corps when the ports with a fleet and a corp were blockaded?

[chuckj118]

The problem with the GB fleet really needs to be addressed.

I started a new game as France today. The GB AI placed just 10 LF blockaded Amsterdam and 20 HF and 12 LF blockading the French in Brest. Well, I attacked in both places and won. I was then able to place the Dutch Corp into middle England. In the two battles GB lost 17 LF to my 4 LF. So this was pretty bad for them.

The Dutch were slaughtered by the small GB Army but what the hey...most of it could have been else where.

If I had a French Corp in Brest GB would have probably gone down.

I've seen alot of good ideas here so I won't repeat them.

[Adraeth]

ORIGINAL: Adraeth Montecuccoli

A simple rule might be that the AI should try to make an average of 2 armies (escept Uk): one to protect her country in defense and one to ravage enemy territory seeking THE DECISIVE BATTLE as in Napoleonic war concepts.

Example: Russia is at war with Turkey, (both are AI); Russia keeps Benningsen with 5 corps near Moskov-Kiev but launches Alexander with 5 corps in Moldavia to reach Costantinopole; in the meantime Turkey keeps Kushanz Ali in Costantinopole and makes offensive in Podolia or Caucasus (depends on initial deploy).

This, i think, might be achieved with different initial deployement and a check on AI trajectories like the pathfinding in other games; so if Turkey is deployed with strong numbers in Caucasus the pathfinding of the AI will check the nearest way to Moscov via Sevastopol and up.... and so on; the initial deploy should be changed by the players thanks to the editor.

Otherwise another solution might be to give each Major Power AI a list of possible objective based on the "red" cities on the map, using perhaps the way the AI thinks about Diplomacy spheres of influence (this works in the game i think as i said in other topic).

For example something like: If Rus at War with Tur then: (objectives).... list of red cities in Turkish control...

Just giving examples, hope this helps

[bresh]

In this game GB's-Fleet  main purpose should be keeping others to enter GB.(later some vps to)
So have a fighting fleet around.
And primary "blockade" the ports that pose a invasion thread. Equal it in force. Mainly heavy fleets with more than 5 ships, and harbors with corps in same zone as the fleet. Essential it should still leave a sizeable force of Gb ships to hit elsewhere(or protect the mainland).

And not forget to hold the channel !

Cheers
Bresh

[dude]

Having played GB most of the time years ago the primary use of my fleets was to bottle up the fleets opposed to me.  I also would leave a fleet to transport a corps in case an opportunity arose to besiege the port where the enemey fleets where.  You don't know how many times enemies would leave a fleet in a lightly guarded port with their corps out of range.  I always felt it was worth the risk to drop a corp with 5i to 10i on a enemy port that had only a garrison.  At a minimum it would make an enemy leave a corps behind to guard the ports and one less corps at the front always was a benifit to my allies. 
 
Any remaining fleets were used to transport my forces to places like North Africa, block crossing arrows, or provide supply.  If there were enough ships left I would leave an intercept force near England somewhere.  This would all of couse vary based on who was a current enemy and how many opposing ships there were.

[chuckj118]

GB has a interesting political/military situation at the start of the game.  He should have to walk pretty softly for the first year or so.  Against Humans with will always be true...but how do we make it true against the AI.
 
Question:  Is the game using a single AI program for all the countries or does each country have its own AI program?  Since strategies vary so much between them I would think they require "special" handling for each country.  Just curious...  [:)] 
 
1)  During the diplomacy phase GB needs to try really hard to keep France, Spain or secondarily Russia from combining on him right away.  If they do and are competent GB might go down.  I estimate about a 30% chance for a France/Spain combo; France/Russia is a bit weaker; All Three will make it over 50% probably (I haven't gamed this out yet)
2) Assuming that he avoids any of the dangerous combos above, GB should concentrate on acquiring (or at least denying them to his potential enemies) the navies of either Sweden, Denmark, or Portugal.  Computer GB should bribe them each turn and concentrate on either retaining one of these three that they have gained through diplomacy or attack one of them as soon as they can.  Against human players of Russia and Spain this could be worked out in negotiation.  Try to take whatever country you attack without destroying the fleet.  This can be difficult but worth it in the long run.
3)  Build infantry and militia during your 1st and maybe 2nd builds.  This is to get boots on the ground in GB and provide a small army and/or for one of the potential target countries.
4)  The GB navies need to stay close enough to GB to attack any sea area around the islands in the early stages.  The AI will be giving the human opponents opportunities if it goes romping off into the Med.
5)  Once Wellington arrives to command the small army that has been built in GB the invasion risk will become less urgent.  He can now stomp on and crush the small armies that France, Spain, or Russia can land.  (Fr - 48 factors with the other two around 30?)
6)  From the first econ phase in 1806 GB needs to commence building 3 to 5 LF every economic phase.  Place them in a well protected port.
7)  From 1806 on GB needs to take advantage of whatever opportunites occur.  His main goals should be to reduce the number of ships, overall, in the game, especially French; and to keep the battle on the mainland going by subsidizing Prussia and or Austria during their down times.
 
Wow...programming all this has got to be amazingly complex!  Hats off to all of you guys!  [&o][&o][&o]
 
Nal 
 

[mrpharmacist]

ORIGINAL: Donegal

Not an IA improvement but interface improvement. I would like some zoom out button, i really miss it. Other possible screen resolutions could be good too

Yeah i like to zoom too.....[;)]

[yammahoper]

So far the game plays fairly well.  The AI has flaws, and I will list the ones I have seen, but it does fight decently when it gets it right.
 
     1) Cooperation between allies; as Russia I invaded France via Morroco and the nation north of holland.  I crushed the garrison protecting the french fleet, but the french ran the blockade and retreated south.  While the brits went last and returned home to replenish the fleets, then first next phase to return and prevent french movement, at no point did the brits, with a stack of corps at home and fleets in port, send me any help.  After several months, I was so reduced I had to flee.
 
     2) Appropriate aggression; when I seized the minor north of Holland, and then holland, the french came and after a few battles, chased me from holland.  I retreated north and because I had a ton of infantry to reinforce in, stayed around for several more months.  I got the attention of the french army and they came in force, again forceing me to flee, but once I fled, those french armies just sat there and did not take out the conq minor I was using to supply myself and invade from.  They also never tried to cut my supply off, and this fight lasted long enough that I rebuilt my guard corp twice and brought in 5 fresh cav plus numerous infantry.  Also, single corps or 100% infantry should not be thrown at large advancing armies.  This tactic may be worth it with a small corp of militia or infantry to see what is coming (and the AI should remember what is coming as well as a human who can see clearly if an enemy is in range of supply, thus knowing if the numbers might change) or even as a delaying tactic if such a manuever would allow a better army to arrive in defense.
 
     3) Appropriate defensive posture; Paris was left wide open with 1I for garrison, allowing me to sneak in a single corp and tackle the city in an eco phase, winning me (to my incredible surprise) a conditional from france.  I had a fleet with a 10I corp sitting in Morroco, the very corp I used to take the minor with (the cav corp left to join the rest of the army), so it would be reasonable for the french AI to know what was there and were it could reach, while almost the entire french army was north in the german states and prussia, fighting russia and prussia.  Since paris can be reached via naval invasion in two rounds, the AI should defend paris with enough force to repel any invasion force in range wjen its corps are out of range, even for a double move.
 
yamma
 
 

[jamo262]

Should (or does) the AI take into account the following when declaring war
[ol][*]Natural alliance modifiers[*]Who is the dominant/rampaging land power in Europe at present. Perhaps % level of VP towards winning total modified by relative land strength.[*]That the web of alliances and declarations of war should primarily be influenced by[/ol]

• The British agenda to not allow a dominant power to emerge in Europe. British natural enemies are France then Russia. (Pre-revolution Russia was considered the dominant land power in Europe FredII legacy notwithstanding).
• Britain can influence with it's money the alliances of other powers to unite against who it considers to be its major geopolitical threat (Normally Napoleonic France 1, Russia 2).

    4. A check for declarations of war might be triggered when another country declares war on a minor power considered to be part of the national aspiration goal of the first country. i.e. France attacks a German minor triggering checks from Austria or Prussia. This is perhaps already incorporated due to the influenced minors concept.

    5. Prussia should be coded to not declare on Britain or Russia. It should only do so against Russia if
a) russia has replaced France as the dominant land power of Europe
b) It is supported by Britain and at least one other major power that borders Russia.
c) Russia should not go bashing Austria and /or Russia while France is still dominant.

6. this is a repeat but Britain should view Russian resurgance with only slightly less alarm than French.
7. Countries should be more likely to declare war on powers that control previously ceded home provinces
8. when deciding to declare war the AI should conduct a comparison of relative strengths (including allies from both sides)
9) for a particular senario where Austria is inactive and france attacks Prussia (who in turn is supported by Russia) the Ai should follow the outline of the 1807 campaign. The AI should construct an optimum army of russian/ Prussian corps and fight in Eastern Prussia defending Konigsburg.
10)The English AI should support allies that have lost their capitals but still fight on  against Britains 'designated enemy' with money.

Are these ramblings of any use?

[bresh]

The AI should really change the way it gives cash. As prusia i had Russia and GB as allies begged me to $, so had 150-250$ in echonmy phases.
 
 
In general guard corps with low strenght, shouldnt walk off solo, within range of the enemy corps.
 
I wasted a solo Austrian guard corps and the solo French-corps+Napoleon 5G3C same turn.
 
Cheers
Bresh
 

[yammahoper]

Solo guard corp was how I captured Nappy, with only 5g and a a few cav in it, I chittted Esc Ass and eliminated the force.
 
Oh yeah, and this lone corp with nappy was sitting in Corsica.  I aborted all my plans to invade turkey from naples and went after nappy, knowing if I won, I would catch him.
 
So, AI should remove leaders to be reinforced later if they are in no position to be of use or in risk of being captured easily.
 
yamma

[carnifex]

The AI should act! I am right now looking at Spain that has been at war with Portugal since turn 1 and my France since turn 4 and it's turn 10 and they haven't moved a single corps anywhere. Their fleets are blockading but not a peep from the land forces. I've seen other AI nations go passive in previous games, like Russia that never takes Finland, that sort of thing.

This is probably a bug where the AI just gets stuck in some loop.

ORIGINAL: Jimmer

ORIGINAL: carnifex

The Guard corps is not to be used by the AI as a lone raider. The Guard corps is not to be used by the AI as a lone raider. The Guard corps is not to be used by the AI as a lone raider.

Why not? That's how Napoleon lost at Waterloo: He put Ney in charge of the Guard with orders to wait until he needed them. Ney sent them into combat 1000-2000 at a time. When Napoleon wanted to "commit the guard", there were few left available to him.

I think this is why Ney's third stat is a 1. At Waterloo, he was in charge of two corps, and didn't do so well.

That would be in the context of a battle. I mean seeing the Prussian guard corps come charging down through the minors towards Paris with Brunswick strapped on while the rest of the Prussians are fucking off chasing Nappy in Ansbach. I almost felt bad stepping on that corps but I had too.

The AI is either passive (my first point above, and probably a bug), or super-reckless. As a veteran EiA player I know job one is always to preserve the stack. Without an army what are you? They take so long to build and to just throw them away well that's a lot of weekends spent munching Doritos watching others pick chits.

If the AI should do one thing, is just to build up a nice stack and move it slowly toward the enemy capital paying attention to terrain. Nothing fancy, pick a spot to concentrate and then just creep. If it has a surplus then some small corps can grab minors. But not this headlong rush.

[Jimmer]

ORIGINAL: lavisj

Actually I think that Ney's rating reflects more his performance during the 1813 Campaign.

As for the Guard at Waterloo, Ney is not the one who commited it, but Napoleon itself, and it was committed at different points:
1. Plancenoit (Young and Old Guard). And set up as a local reserve.
2. Placed back in general reserve (minus a division of Young Guard)
3. During the main battle for the last ditched effort at the end.

But you need to place yourself in Napoleon's shoes. The Guard was the last reserve he had available at this point, committing it in its entirety to win the battle would have been great if it had worked, but if it had failed, there was nothing left to cover the retreat of the army.

The commitment of the guard actually seems appropriate at Waterloo. And it was not a lone raider as it was part of a 6 Corps battle (on the French side).

Perhaps I need to read my history books better.

Thanks for the info. I'll correct my misperceptions.

[Jimmer]

ORIGINAL: bresh

Gibraltar has no value beside a safe port in this version of the game.

Oh, yes it does. It's a safe port from which a sizable British fleet can monitor entrances and exits to and from the Med. You never know when an enemy fleet might ...