How to REALLY improve the AI in the next patch!

[Hartmann]

We do not have to wait for EU2 to have these things tweaked:

1. ECONOMY: This is the cornerstone.
- AI should be tweaked to handle inflation better.
- It should be tweaked to go for better infrastructure decisions. (We already had the Pleskau case. I recently discoverd Kleve building a ship manufacture. Kleve has no access to the sea and the province is just providing grain.)

2. PEACE TREATIES:
- Minor AI often agrees to silly peace treaties giving provinces/indemnities despite having achieved a major victory. This should be changed.

3. CONSCRIPTION/DISBANDMENT:
- AI should not conscript so heavily on the isles, neglecting the mainland (still a problem with 1.05b)
- AI should not disband all their troops after having concluded SEPERATE peace while still at war with other countries (occurs MUCH too often).

3. WAR:
- AI should not declare war, when having no troops on the mainland and no ships to transfer them from the isles (Venice)
- AI should not invade isles, unless it has gathered and shipped a superior force. (By now it wastes it´s troops peacemeal: Turkey/Venice)
- AI shouldn´t assault while there´s a considerable enemy force in the neighbouring province
- AI should try to attack an assaulting force from a neighbouring province

I´m convinced, that if these details will be tweaked, the AI will provide a REAL challenge.

Hartmann

 

[unmerged(541)]

I just opened the same topic in 'General Discussion'. I support your ideas and just want to add some points.

1.AI countries shall be more active on the diplomatic front. They shall make it hard for the human player to advance militarily, to build alliances AND to keep them stable.

2. AI countries shall do more against inflation. Now runaway inflation cripples their economies, so that there is no money left for research etc.

3. For the superpowers there should be special challenges. Expanding empires create new problems. Wastage, slowliness and huge bureaucracies are specific disadvan-tages. To make it easy for implementation one could say that for every province exceeding a specific number(steps for little, mediumsized and big empires)there is a financial fine to reflect that(higher transaction-costs for empires).

N.B. I made some remarks on the German Forum in 'General Discussion'

Greetings

In hoc signo vinces!

 

[ar73]

Hartmann:

- AI should not disband all their troops after having concluded SEPERATE peace while still at war with other countries (occurs MUCH too often).

Hm, I never did disband any troops.
A minor pays ~10d for 1k of infantry. Upkeep for 1k of infantry is <<1d.
For cavallery/cannons the disband/buy vs upkeep cost ratio is even more disadvantageous.

=> AI disbanding troops should be the
exception.
=> AI should disband inexpensive infantry
first.

 

[Hartmann]

ar73: The main problem is, that AI often disbands huge armies while still being threatened (and EVEN when it would be easy to walk them home).

basil: WHICH German forum do You mean?

Hartmann

 

[unmerged(330)]

AI must react faster on growing powers , alliances must shift often and not be based on relations as much as power balance.

Backstabbing should accur more often , first land attack second land , backstabb from third land.

 

[Hartmann]

Good points, netraM!

Hartmann

 

[unmerged(541)]

Originally posted by Hartmann:
ar73: The main problem is, that AI often disbands huge armies while still being threatened (and EVEN when it would be easy to walk them home).

basil: WHICH German forum do You mean?

Hartmann

I mean the Forum at europa-universalis.de



In hoc signo vinces!

 

[unmerged(452)]

Originally posted by basil:
1.AI countries shall be more active on the diplomatic front. They shall make it hard for the human player to advance militarily, to build alliances AND to keep them stable.

Agreed IF this rule is not applied if the human player behaves normally. I would hate to see the usual AI stragtegy of banding together against the human player, regardless of circumstances.

I find the ability of the human player to build AND maintain stable alliances with AI countries one of the most refreshing aspects of EU

/zwingli

 

[unmerged(542)]

- AI shouldn´t assault while there´s a considerable enemy force in the neighbouring province
- AI should try to attack an assaulting force from a neighbouring province

Well, in the Northern Campaign Russia had 47k troops to counter-attack Swedish forces. The KI didn't use them to capture provinces (instead it let me do the work )

And assault/not-assault decision should also depend on the fortress size. Why should I care if there is an enemy 20k army if I have level 4 fortresses but I could capture the level 1 fortress with my army. It would be stupid if the AI would say: Hey, I attack the army instead of the territory. After all, captured territories are the key.

Well, this is just to show you that implementing rules is a difficult thing, there is always a situation where it would be a bad thing

 

[unmerged(23)]

Originally posted by Hartmann:
1. ECONOMY: This is the cornerstone.
- AI should be tweaked to handle inflation better.
- It should be tweaked to go for better infrastructure decisions. (We already had the Pleskau case. I recently discoverd Kleve building a ship manufacture. Kleve has no access to the sea and the province is just providing grain.)

Infrastructure decisions are sometimes very bad for some major countries too (Spain comes to my mind).

2. PEACE TREATIES:
- Minor AI often agrees to silly peace treaties giving provinces/indemnities despite having achieved a major victory. This should be changed.

I don't fully agree. A country can have achieve a major victory but can have a depleted manpower and/or treasure meaning he's heading srtaight to defeat.
Your points are valid if the behavior is really silly. It's up to paradox to check if it's a silly one or rationale one (like 'I'm going to be annexed even if, now, I'm victorious').

3. CONSCRIPTION/DISBANDMENT:
- AI should not conscript so heavily on the isles, neglecting the mainland (still a problem with 1.05b)
- AI should not disband all their troops after having concluded SEPERATE peace while still at war with other countries (occurs MUCH too often).

Agree although I didn't see that often during the beta. Can be linked to various AI patches that followed the realease.

3. WAR:
- AI should not declare war, when having no troops on the mainland and no ships to transfer them from the isles (Venice)
- AI should not invade isles, unless it has gathered and shipped a superior force. (By now it wastes it´s troops peacemeal: Turkey/Venice)
- AI shouldn´t assault while there´s a considerable enemy force in the neighbouring province
- AI should try to attack an assaulting force from a neighbouring province

Agree although, as it was said, such subtlety in the AI can be very difficult to implement. Justz hope it can be done.


PS : I also answered to Basil post (same topic). I didn't check the fact the topic was posted twice before answering.

------------------
Pierre
Retired EU Beta Tester

 

[Guest]

Military AI weaknesses:

- stacking huge armies for months beyond supply value with the result of incredible attrition losses: Spain by example moved two armies numbering 88,000 men in Valentia then let them here. In a few months, only 56,000 remained.

-AI doesn't place huge armies under best leaders: during the same time, Cordoba was the leader of a proud 4,000army in Roussillon, 2 hexes away.

- when AI capital is besieged, I've never seen AI move back troops to relieve the fortress.

- AI seems to privilegiate seizure of province rather than fighting against enemy forces. AI should adopt a more balanced strtegy.

And by the way: loan mechanism has to be modified, because it's too lenient for the player...unless the problem is only the fact AI doesn't agree to take loans in any case...

[This message has been edited by laurent Favre (edited 20-12-2000).]

 

[unmerged(358)]

As netraM said I believe backstabbing is the way to go...
I've seen this in the boardgame and I regularly use it myself...
What I mean is that the AI countries should be better at grabbing a great opportunity when it has one... Let's take an example in my recent campaign, when I was playing England. I was in a war with Venice and had been for a few years, then both countries like France, Spain and Scotland should have attacked. I was moving larger bits of army out of England and my fleet was all down in the mediterean. France could finally have grabbed Calais, Scotland could have kicked my butt and Spain could have captured larger parts of America that I had seized. All while I was occupied with Venice. The same goes for Turkey, they are fighting a lot with Venice anyway, why couldn't they have joined me in trashing them ??
This kind of behaviour will make it much more difficult to turn into a proper superpower, since whenever you are fighting you should expect some oppurtunistic minor (or major) to grab some provinces or assets in your backyard... That way you can't concentrate on your single enemy. This shouldn't be too difficult to implement, I even think it can almost be hardcoded on the most difficult setting. Like NEVER attack on the continent without having some fleet and some major forces to protect against Scotland when you are playing England.

Another thing I want improved is Spain's handling of the treaty of Tordesillas, if it works like it does in the BG why doesn't Spain grab all my non-fortified colonies in south America ??

While I'm at nation-specific AI...Something has to be done with the Portugal AI... Why should anyone explore at all, when you can simply send 15k troops to Portugal, smash their little 5k army they have protecting their main provinces and siege and grab their capital... free maps... Portugal hasn't even improved their capital fortress !! That should be the classical thing for them to do, since they've got little troops but usually quite a bit of money...

Well that's all for now

Cobos

------------------
If you are not part of the solution you are part of the precipitate.

 

[Oranje]

A few thoughts:

It's my feeling that, although some ideas put forward for improvement of the games' challenge to the player may prove too hard to implement in a patch (but let's wait to see what Johan thinks when he gets back), there are certainly some options that would matter a lot and donot seem very hard to implement.
Just a few thoughts:

1.
If the AI would make better use of available military leaders, the outcome of many conflicts would presumably be different.
In my last game I've witnessed a large English fleet get slaughtered by the french, while their only admiral was hiding in a close-by harbor -does ring a few dutch 1665 bells though .
What was even worse was seeing Spain's excellent Duke de Alba withering away in Western Pommerania for a very long period, after peace was signed in a war, Spain having ample naval transport available, but apparently the AI doesn't care about getting troops back home after the war.
It got even worse for Spain, as the dying Duke de Alba was succeeded as commander of a zero men army in the same abandoned spot by Spinola, another well known Spanish commander. Both good leaders spent numerous inactive years like this.

2. AI should get a quick course in assessing a LOT better when to combine forces if necessary and splitting them again after the action to avoid massive attrition.
To me it seems that a few simple extra rules would do most of the trick.
a) AI should not build, send or station more troops than a province can support (not counting the enemy troops in its territory).
Only exception: in order to quickly storm a fortress, or attack an enemy army.
b) AI (of a country, or the combined AI of an alliance) should not engage in futile attacks with a small army against a strong opponent, or in overkill actions (seen it send 50.000 plus armies into an enemy province that could only support 3.000, just to besiege a level 1 fortress) but gather a strong (enough) army first if possible.
c) AI should not dismiss troops unneccesary

Other ideas that might help to create an even better (challenging) game:
d) Armies should not have eternal life after conscription, 15 years seems maximum (would slow down expansion); leaders retire and/or die when old, so do soldiers (but sooner).
e) Garrison strength should not only be linked to fortress level, but (for provinces with low pop) also to province population.

greetings, Oranje

[This message has been edited by Oranje (edited 21-12-2000).]

 

[unmerged(258)]

Some comments:
*I agree with the minors sometimes accepting unfavorable peace. Kazan had TAKEN Moscow, and though it was at war with both Russia and Denmark, I was at war with Denmark as Sweden ... yet Kazan made peace agreeing to PAY Russia an indemnity?? As I had a royal marriage with Russia, I could see they had no other armies in their territory to retake Moscow anytime soon.
*I also do not favor just a blanket 'AIs band together to attack player country' solution. If we find a way to put in a change of historic terms, that would be a solution. (e.g. The other nations of Europe worked together against Louis XIV's France; Richelieu's France assisted Protestant nations to curtail the power of the Habsburgs, etc.)
*As far as leaders go, it would make sense that the biggest armies are commanded by the highest-ranking leaders (which should be the special leaders) where practical.
*I agree that many things that have been suggested in the past may be difficult to patch (esp. quickly). Hopefully we can find some things that can be adjusted/changed/corrected relatively easily that can help make the AI function better.

My 2 cents' worth ...

[This message has been edited by Tom (edited 21-12-2000).]

 

[unmerged(258)]

Ignore this message ... hit the wrong button.

[This message has been edited by Tom (edited 21-12-2000).]

 

[unmerged(23)]

Originally posted by Tom:
*I agree with the minors sometimes accepting unfavorable peace. Kazan had TAKEN Moscow, and though it was at war with both Russia and Denmark, I was at war with Denmark as Sweden ... yet Kazan made peace agreeing to PAY Russia an indemnity?? As I had a royal marriage with Russia, I could see they had no other armies in their territory to retake Moscow anytime soon.

Too difficult to say. We should know what determine how the AI concludes peace to be able to judge.
Kazan could have no manpower left. And YOU knew Russia didn't have any army left. Kazan probably didn't.
My opinion is : when you are a minor power at war with a major, with no ressources to continue the war left, it's better to negociate a peace rather than continuing it, allowing the major power to recover, invade and annex you.

Your comments can be right. It mainly depends of how the AI makes its choices during peace negociations.

------------------
Pierre
Retired EU Beta Tester

 

[Hartmann]

Pierre: Please have a look at this thread:
http://www.europa-universalis.com/forums/Forum7/HTML/000128.html

Especially the Marocco example. I encountered NUMEROUS other incidents, where there was money payed or even a province given to a country which was utterly defeated and had no means to retaliate! Maybe this didn´t occur in the beta, but in 1.05b it´s a fact.

I want to add another problem to the list:

- The setting is this: We have two warring countries A and B. A besieges fortress b of B and B beseiges fortress a of A. The level of both cities goes down to say -6. Now A is somewhat luckier and occupies fortress b a little sooner, then moves to another province. Now B, instead of first capturing the soon to fall city a, lifts siege of city a immediately and moves to b trying to liberate that city! I have seen a country loosing a war because of this odd behaviour much too often!

Hartmann

 

[unmerged(358)]

I've thought a bit more about that backstabbing tweak I mentioned earlier above...
Maybe finding which countries should easily backstab could be connected to having a CB and relations ? Just a suggestion ?

Cobos

------------------
If you are not part of the solution you are part of the precipitate.

 

[unmerged(23)]

Originally posted by Hartmann:
Pierre: Please have a look at this thread:
http://www.europa-universalis.com/forums/Forum7/HTML/000128.html

Especially the Marocco example. I encountered NUMEROUS other incidents, where there was money payed or even a province given to a country which was utterly defeated and had no means to retaliate! Maybe this didn´t occur in the beta, but in 1.05b it´s a fact.

I want to add another problem to the list:

- The setting is this: We have two warring countries A and B. A besieges fortress b of B and B beseiges fortress a of A. The level of both cities goes down to say -6. Now A is somewhat luckier and occupies fortress b a little sooner, then moves to another province. Now B, instead of first capturing the soon to fall city a, lifts siege of city a immediately and moves to b trying to liberate that city! I have seen a country loosing a war because of this odd behaviour much too often!
Hartmann

I've read those examples.
Peace resolutions are resolved not alliance vs alliance but country vs country.
That leads to some curious results when globally an alliance is winning but individually, some countries agrees to make peace, crippling their alliance power and utterly leading to defeat.
Grouped, some countries can have an edge but individually, they, maybe, can't negociate a good peace treaty.
Spanish mainland occupied ? Unless in 1492, there are still colonies to help Spain to recover. Colonies most minors aren't able to attack.
Only the minor who control a capital should be a little more inclined to ask a better peace for himself.
Others can control provinces and individually have few stars in peace negociations. So individually, the currently defeated country can refuse peace.

It can be critized yes. I also think some slight fixing is needed there and there.

But the individual peace negociation system is needed.
It's a gameplay feature but also simulate the fact an alliance is not a monolith following its leader will at all cost.
Probably a good compromise is to let such individual peace negociation feature and add a modifier evaluating the overall victory of your alliance against the ennemy alliance.
Such modifier has to have a low impact tho otherwise some countries could let their alliance fights and ask for indemnities even if they didn't fight at all.
Or, another solution is evaluate each alliance overall sucess and members of the winning alliance shouldn't agree for a bad peace.
Problem is now to adjust the system to what is 'bad' or 'good'...

------------------
Pierre
Retired EU Beta Tester

 

[unmerged(258)]

Agreed, Pierre. It is difficult to determine any rules that can be used for the AI in making peace....

A little more information on the example I gave of Kazan having taken Moscow. It had an army still occuping Moscow and there were no Russian armies - but as you said, perhaps Kazan didn't know this. Still, I do not see a way that Kazan should have taken an unfavorable peace. It still does not seem reasonable that a country whose capital is taken without them having taken anything from the enemy should get an advantageous peace. At a minimum, it seems they would have settled for a Status Quo. In this case, the war had been relatively short, so I do not really see how Kazan could have 'lost any stars' ...

I don't think wholesale changes need to be made, and yes even tweaks can be difficult in this type of situation (unlike just adjusting attrition rates for various unit types). However, some change in this area would be an improvement, in my opinion.

Thinking about it, just the fact that we have a game where these are the types of issues that come up is a good thing!!!