The AI (Full Version)

All Forums >> [New Releases from Matrix Games] >> Empires in Arms the Napoleonic Wars of 1805 - 1815



Message

NeverMan -> The AI (6/7/2008 4:52:46 PM)

I just started my second AI game (I had played France before with v1.00) and now I am playing Austria with v1.02k.

The AI seems to be improved in the sense that it doesn't do nearly as many "stupid" things. By stupid, I mean anything that puts itself in harms way and is totally moronic.

That said, the AI doesn't do any "smart" things either. I am in June 1805 and right now it looks like this:

France and GB start at war: normal.

Russia DOWs Turkey: OK, that's not out of the ordinary, except Russia does not have any of it's Corps in position to do any combat with Turkey any time soon, so it basically is just eating PP at the moment.

France declares war on Russia. ????? Why? France is not at war with Prussia or Austria, has no means AT ALL to get to Russia to do any damage to the Czar. This move is just another "eating PP" move. ALONG with this then Prussia came to Russia's call against France. This might not be a total bone head move depending on how you see the situation, but it's not a move I would make this early in the game and since France cannot put any direct pressure on Russia, Prussia is giving France it's only means of getting PP back from DOWing Russia.

Spain DOWs Turkey. Again, not a bone head move, except that now Turkey has called France and then Spain has called GB. France is now at war with GB, Spain, Prussia and Russia all in June 1805.

Now, here comes the crazy part: Prussia DOWs Turkey. ???????????? WHAT????? LOL. This makes NO SENSE at all. Some might employ this move later in the game if they are doing REALLY WELL as Prussia, personally, I've never had the inclination to go all the way down to Turkey to do combat considering France and Russia are so close, especially not in 1805.

OK, so now the world is in Chaos right? WRONG. None of these nations are actually in any position to do combat with each other. There are no land combats happening. Turkey is sitting with it's corps spread all over Allah's brown earth, Russia doesn't have Corps anywhere near Turkey, France is just chillin out, Spain is doing nothing, GB is doing nothing and Prussia is doing nothing.

What am I doing you might ask? Gathering minors, what else? Since no other country seems all that interested in getting money and manpower I will. I have all of Italy at the moment (Spain didn't contend for Naples AT ALL). France doesn't seem interested in Italy either. I have Baden, Bavaria, Wuttermburg and Swabia, no contention there either.


My bottom line: the AI doesn't seem as boneheaded as it once might have, but it has a long way to go. I hope it gets there.



borner -> RE: The AI (6/7/2008 4:56:54 PM)

In a game this complex, I doubt the AI will ever get close to the quality of an average human opponent, or even a below average one. Understanding that, I think if all you are trying to do is get a feel for how things operate, the AI is ok. Then you can try to take on a real game as Prussia, Spain or maybe Turkey.



NeverMan -> RE: The AI (6/7/2008 5:45:44 PM)


quote:

ORIGINAL: borner

In a game this complex, I doubt the AI will ever get close to the quality of an average human opponent, or even a below average one. Understanding that, I think if all you are trying to do is get a feel for how things operate, the AI is ok. Then you can try to take on a real game as Prussia, Spain or maybe Turkey.


Yes, I understand this, I am in several PBEM games now, which I believe one you are in, I think you are playing Russia, not sure though.

I agree about the current state of the AI; however, I think that there are some things here that can be addressed (are totally possible to code) that can make the AI much better.



NeverMan -> RE: The AI (6/7/2008 5:51:34 PM)

OK, so now I am in 1806.

GB DOWs Russia: not totally out of the ordinary, but you have to wonder anytime a GB player does this what that advantage is. For me, I have to be within striking distance of the Czar's navy and have a solid ground force so that I can make a run for Sweden. You will probably never get a surrender out of Russia so Sweden and the Russian fleet is all you are going to get for your PP.

HERE is something I have a SERIOUS ISSUE WITH: Why does France have so many damn corps here? (In case the attachment is too small for you to see, 90% of those corps are just CAV factors with no INF/MIL in them). There is a Cav Corps that is not even all the way full. Needless to say this brought down Murat's Tactical quite a bit (personally, we would always play with the rule that a leader can only have twice as many corps has is limit will allow, ie. 12 for Charles and Nappy, etc.). I'm not sure how many CAV factors France got that reinforcement phase, but Murat only had 3 corps in the previous month when we did battle so he didn't have a lot of CAV then either. Now he has 28 CAV.



[image]local://upfiles/11088/D0678CB0ABFF4910A22AC63A6352BF18.jpg[/image]



pzgndr -> RE: The AI (6/7/2008 5:58:44 PM)

quote:

In a game this complex, I doubt the AI will ever get close to the quality of an average human opponent, or even a below average one.


Well I disagree completely. Sure it's complex, but it's not impossibly complex and there are still some fundamental rule sets even a human player mentally goes through to make reasonable decisions. Most everything a human player might consider based on the current game situation prior to making a decision could be scripted for the computer opponent, with some randomness included. Depending on the state of Marshall's AI algorithms, sooner or later he should be able to develop a decent and challenging AI for each MP. Noting the expectations from solitaire players and even some multiplayer groups for permanent/temporary replacement players, everyone should be supporting this goal. Yes?

France DOWs Russia. Prussia DOWs Turkey. These are obviously not right. But the AI is in fact doing things, which is good, so it's more a matter of making some adjustments for the AI to make more "smart" decisions and fewer "stupid" decisions. I'm optimistic these things are eventually possible, but also fully aware of how difficult and time consuming AI scripting and playtesting can be. So while I doubt the AI will ever get close to the quality of an experienced veteran human opponent, with difficulty level bonuses included the computer opponent(s) should be a fair match for an average human opponent and certainly a below average one.



NeverMan -> RE: The AI (6/7/2008 6:04:41 PM)

I agree completely. I think we all know that the average AI in any commercial PC game isn't going to match the wits/knowledge of an experienced player, they should at least provide some challenge. I also agree that this can be accomplished by not doing silly things like the things mentioned above.

ALSO, it still seems that the AI is making "one corps" runs at capitals (with like 1 INF in it, it would be different if france had 25I and 3C going after Vienna when none of my corps were around, but instead he runs Ney with 1 INF or 1 MIL factor just to get destroyed by my 20 INF factors, it's needlessly silly and risky with a leader).

ALSO, the AI still needs to do a better job of stacking corps. I see Nappy running around way too often with 1 or 2 corps under him just asking for it.

I think that DOWs need to be better coordinated with tactical positioning and geographical location. France can DOW Russia, but it needs to make sure it can even reach Russia (ie. have Poland or some Prussia/Austria province bordering Russia, or have unlimited access through Prussia by means of a surrender, etc, etc..).

Also, there are too many DOWs going around too early. Minors need to be more of a priority to the AI. I grabbed all the minors while the other countries were doing nothing.

EDIT: Jan 1807, Spain DOWs Prussia? Again, why? Can Spain really have any effect on Prussia whatsoever? What can Spain do to Prussia? The AI simply needs to be more focused in it's efforts. This isn't "Diplomacy" after all.



NeverMan -> RE: The AI (6/7/2008 6:31:04 PM)

Ok, I'm not sure how this happened because I wasn't paying direct attention to it, but France got Davout in London with 1 corps. GB currently has several fleets in the channel (but they might not have been there before). GB surrenders conditionally to France. Does the GB/France start at war not come with the unconditional option automatically? If not, then there needs to be an option for this.



Marshall Ellis -> RE: The AI (6/7/2008 6:39:55 PM)

Neverman:

Mandatory GBR-FR unconditional is NOT in the game. How important is this?




NeverMan -> RE: The AI (6/7/2008 7:27:49 PM)

IMO, It's pretty important. Right now, every group I am in is using this as a "House Rule". It's pretty standard. I'm not saying it has to be hard coded, and it's certainly not a priority, but it would be a great option that would probably be pretty easy to code in.

The problem comes in AI games. It doesn't matter much now since the AI is easily dominated, but if the AI gets wise then it will become a problem in AI games, since the AI doesn't really care what House Rules you have in play.



NeverMan -> RE: The AI (6/7/2008 8:09:19 PM)

France again for some reason decides to DOW GB, then a few months later France gets Soult into England.

The problem is this: Soult is not even in LONDON, he has just come across the Channel (again while it seems there are ships there), and GB decides to unconditionally surrender. Who knows why? Picture is taken during Reinforcement phase of same month, so no one has moved since GB surrendered.

[image]local://upfiles/11088/F2D1D6075E6C489FADA3FF74BE990EB0.jpg[/image]



NeverMan -> RE: The AI (6/7/2008 10:17:35 PM)

This game ended June 1812. My final Money/Manpower was ~226/110, I had 3 provinces from Turkey, Russia and Prussia and 6 from France. I had all but 1 minor in middle europe and all of Italy. These were the only countries I DOWed. I took my time, spending some time acquiring minors and building my army, just like I would in a normal game, so if it seems like it took me a long time, that's why. I wasn't playing for speed.

Also, I played Austria with 18VP for the bid, so my total was 348. It ended with me at 103% and Spain in second with 78%. FYI.

The AI needs to be more aggresive and in force. The few times the AI attacked Charles with some number of force, it faired just like any other player would have, and I felt it. When you lose 1 or 2 battles at 3 and 2 PP each, that starts to hurt. France somehow managed to get Nappy killed/captured early in the game, I had casualties turned off. I also played with winter movement.



ndrose -> RE: The AI (6/7/2008 10:38:12 PM)

As to France getting across the Channel: I've noticed that after a while, instead of blockading, Britain falls into a pattern of putting its whole fleet into the Channel one turn, then moving them all to London the next, then back to the Channel, and so on. So I'm guessing that the turn before this, the fleet was in London, Soult moved across, then Britain closed the proverbial barn door.

When I play France against the AI, I just pretend I can't get across the Channel until I've built up a navy and actually defeated the British fleet. (This also is too easy to do: why doesn't Britain lay down ships? But I think that's on Marshall's list.)

Nathan



NeverMan -> RE: The AI (6/7/2008 11:33:13 PM)


quote:

ORIGINAL: ndrose

As to France getting across the Channel: I've noticed that after a while, instead of blockading, Britain falls into a pattern of putting its whole fleet into the Channel one turn, then moving them all to London the next, then back to the Channel, and so on. So I'm guessing that the turn before this, the fleet was in London, Soult moved across, then Britain closed the proverbial barn door.

When I play France against the AI, I just pretend I can't get across the Channel until I've built up a navy and actually defeated the British fleet. (This also is too easy to do: why doesn't Britain lay down ships? But I think that's on Marshall's list.)

Nathan


LOL.

Even if that were the case then why did Soult just sit there for a land phase?

Naval Phase 1: GB moves back to London
Land Phase 1: Soult moves across Channel
Naval Phase 2: GB moves back to channel
Land Phase 2: Soult stays put?

Either way, it's not good. Either France is getting across the channel because GB is moving back to London even though there is a French Corps in the crossing or the French get in England and decide to do nothing.



StCyr -> RE: The AI (6/8/2008 11:15:15 AM)

the "ai" reflects somehow somebodys "i", doesn´t it ? And if you hardly know what this game is about, then you get suicide corps going to die for the enemy occupied capital ( not even the Fuhrer could dream about), crazy dows, nations acting without any concept but only to go for the human player, etc. Oh- you mean sombody DID KNOW what this game is about ? Well, so you want to say we got that braindead ai by intention ?

BUT- as long as there are still guys around who simply argue "this game is just very complex" there is still hope that even the non-ai design mistakes (ie no naval evasion, no naval pursuit) may stay unharmed.

(11.8 would be very helpful for a "nation grande design")



pzgndr -> RE: The AI (6/8/2008 5:11:58 PM)

Yes, 11.8 Alternate Dominant Powers should be considered.  It would help the AI focus on what France and Great Britain should try not to lose, and on what the other MPs should try to gain.






Mike Scholl -> RE: The AI (6/8/2008 6:18:15 PM)


quote:

ORIGINAL: pzgndr
France DOWs Russia. Prussia DOWs Turkey. These are obviously not right. But the AI is in fact doing things, which is good, so it's more a matter of making some adjustments for the AI to make more "smart" decisions and fewer "stupid" decisions. I'm optimistic these things are eventually possible, but also fully aware of how difficult and time consuming AI scripting and playtesting can be. So while I doubt the AI will ever get close to the quality of an experienced veteran human opponent, with difficulty level bonuses included the computer opponent(s) should be a fair match for an average human opponent and certainly a below average one.


You seem to be a supporter of my theory that "AI" stands for "Animated Ignorance". Do SOMETHING, even if it makes no sense at all! It shows the player that the "AI" is "active", and makes him think it might have a plan.

Thus you get declarations of war against nations that can't be reached.., or without having any units in position to actually WAGE the war just declared. Personally I'd love an AI that would simply make rational decisions.., they don't have to be clever or brilliant---just not stupid and illogical. Seems this one has a long way to go...



pzgndr -> RE: The AI (6/8/2008 7:06:18 PM)

Whatever gave you an impression that I "support" an AI that makes no sense at all?  I mentioned the fact that the AI is doing things to highlight that it is functional, as opposed to not being programmed to make any DOWs or move units or engage in combats, etc.  The fact that the AI is making some stupid and illogical moves (ie, doing things) indicates there is some programming there, it just needs its parameters adjusted.  That's easier to do than adding in new programming from scratch. 



Thresh -> RE: The AI (6/8/2008 10:06:56 PM)

I would say it surrendered because it had nothing to oppose the French other than some garrisons. The French Corps moves to London, where it can siege and forage with impunity.

Surrender early and theres no risk of losing the capital.

Todd



Mike Scholl -> RE: The AI (6/8/2008 11:32:16 PM)


quote:

ORIGINAL: pzgndr

Whatever gave you an impression that I "support" an AI that makes no sense at all? 


Your statement "But the AI is in fact doing things, which is good". Implies that being "animated" (doing something) is in and of itself a "good" thing. With which I disagree, as I see "doing something" as a "smokescreen" to cover up for the fact that the AI doesn't know what to do...., so it just does a lot of things to give the impression of usefull activity.



Marshall Ellis -> RE: The AI (6/9/2008 12:01:24 AM)

I just wanted to clarify a little bit here...

I'm not making the AI do any type of smokescreen to fool you into thinking that it's doing something useful???????
Not my style!
I've got too much blood involved to try and do that to you guys???
Hope everybody understands that???






NeverMan -> RE: The AI (6/9/2008 1:18:19 AM)


quote:

ORIGINAL: Marshall Ellis

I just wanted to clarify a little bit here...

I'm not making the AI do any type of smokescreen to fool you into thinking that it's doing something useful???????
Not my style!
I've got too much blood involved to try and do that to you guys???
Hope everybody understands that???






I do. I know this is a work in progress and that it will get there someday. Like I said before, the AI doesn't have to be great, or even good for that matter, but it must at least make some sense and on "Hard" (with all the VP/PP etc giving) it should force the game to the end.

I didn't start this thread as a bitching post (it kind of got hijacked it seems). I started it as a "what's wrong with the AI and what it needs to do to help make it better".



Marshall Ellis -> RE: The AI (6/9/2008 1:12:49 PM)

Neverman:

I'm with you. I liked this thread! I could learn something from the input! I'll keep watching.
Appreciate your support.



Minedog -> RE: The AI (6/9/2008 1:27:11 PM)


quote:

ORIGINAL: Marshall Ellis

Mandatory GBR-FR unconditional is NOT in the game. How important is this?



I would say very important as long as France and Britain are Dominant Powers. The French/British conflict is the primary driver in the boardgame, with Britain propping up one Coalition after the next with money and a few troops, and dominating the naval war.

In my head I would try and separate the tactical AI (i.e the manoeuver and battles) from the strategic AI. I think the tactical AI is slowly getting better, but there are some important tactical concepts the programmers need to grasp, foremostly what kind of war the power wants to fight (wide front or concentrated), the type of battles you want to fight I.e high casualty or morale breaker & pursuit battles. Within the battles, reinforcement and guard commitment are important choices not just to win the battle, but to limit the harm of a loss.

In the strategic (or diplomatic AI) I think the game should start with predispositions towards the historical alliances (France ,Spain & Turkey vs Britain, Austria & Russia), but swing towards defeating the leader based on %VP as the game goes on. However, at this point, even a strategic AI that did nothing offensive and called allies when DOWed would be better than the random war generation that seems to happen.



NeverMan -> RE: The AI (6/9/2008 3:20:28 PM)

Minedog, I agree with most of your post but I disagree with the tactical and strategic being separate. Personally, I think one usually dictates the other and vice-versa. For me, sometimes what I am doing strategically depends on how I look tactically (the ability to adapt), but usually my tactics are dictated by my strategic goals.

It seems right now that the AI will send a lot of little Corps out trying to capture Minors from me. It's annoying yes, for awhile it works, but eventually the AI doesn't know when to surrender so it always gives me time to crush those little Corps scattered all over the place and take back all my minors.

The AI also seems to not attack very often with enough Corps/Factor Strength. I have seen Nappy and other good leaders thrown away by attacking with too little forces.

The AI also seems to continue attacking in hopes of a "last battle". For example, France kept attacking me with Militia under Davout even though I had large forces (Charles with 8 Corps) sieging Paris and even a month after I got into Paris it was still attacking me. The same thing happened with Prussia and Blucher. The AI just doesn't know when to cut it's losses.

On the other hand, when I went after Turkey, it let me walk right to Constantinople unattacked. I had a supply chain unguarded of 5 depots behind me and it did nothing. John went to Constantinople while Charles went around and destroyed all the little Corps that were left alone. Turkey had a stack down south with a leader not far from John, but it made no attempt to attack John. Turkey was also at war with GB and Prussia at the time. GB had blockaded the port where these Corps were, but the AI needs to prioritize in that situation.



Marshall Ellis -> RE: The AI (6/9/2008 3:40:54 PM)

A couple of confirmations:

The AI will try to protect a capital with anything that it can. This does create suicide attacks. It is difficult for the AI to understand that surrender is maybe better than losing one's capital. I call this the Japan syndrome!

The AI also is NOT taking minors like it should!

The AI is getting lost after successfully taking a minor. Right now it is not answering questions like, "What do I do with my forces in Egypt after Egypt is conquered?" or "What is the next minor I should go after?"

These are all things that I am looking to make better in 1.03









ndrose -> RE: The AI (6/9/2008 4:15:41 PM)

The way the AI protects the capital (though sometimes overdone, as when France alls back the entire Armee to obliterate a lone Spanish cavalry corps that, if it were unfortunate enough to get a breach, would be wiped out by the 25-strong Paris garrison) is not so bad, generally. The real problem is the way it goes after an opponent's capital--like that lone cavalry corps, or the way France attacks Berlin: the army will sit, a huge stack of corps, in Lausitz, and send one corps at a time to face the Prussian army.

And a peculiar habit with minors: even when it does declare war on them, the AI will do nothing about it if they get run by an MP it's already at war with. Now, on the one hand, this shows the AI is thinking a bit: I don't have to worry about lapse of war, since I'm at war with the MP, and can take this minor at my leisure. However, it never does do anything. Egypt is a good example. If Turkey is at war with the controlling MP, it just sits in Jerusalem. Conversely, the MP controlling Egypt starts launching suicide attacks at the Turkish army until there is no Egyptian army left. But even then the Turks don't move. And, as far as I can tell, though they may be left unmolested for years, the Egyptians never rebuild their army--not even to garrison Cairo. Both sides seem to be doing everything they can to lose....



Tater -> RE: The AI (6/9/2008 9:26:59 PM)

quote:

ORIGINAL: pzgndr
France DOWs Russia. Prussia DOWs Turkey. These are obviously not right. But the AI is in fact doing things, which is good, so it's more a matter of making some adjustments for the AI to make more "smart" decisions and fewer "stupid" decisions. I'm optimistic these things are eventually possible, but also fully aware of how difficult and time consuming AI scripting and playtesting can be. So while I doubt the AI will ever get close to the quality of an experienced veteran human opponent, with difficulty level bonuses included the computer opponent(s) should be a fair match for an average human opponent and certainly a below average one.


In my experience with EiA (and any other wargames) it isn't even a matter of making "smart" decisions. The players that do well are usually just not making "stupid" decisions. IMHO, just getting the AI to make as few bad choices as possible will, by itself, make the AI a better opponent. I think this is reasonably achievable and some of the improvements to date already demonstrate this.

The single biggest step now is getting the AI's to build "good" armies. So far they just aren't doing that. Over and over, regardless of which side I play or play against I go into battle after battle vs crappy armies:
1) To much militia
2) No militia
3) Half empty corp
4) No cav...
5) All cav...
6) etc...

IOW, there is no balance at all to the AI's constructing of an army.

I think two rather simple concepts that the AI needs...force balance in building an army and creating multi corp stacks using said balanced army. I get the impression from the AI actions that supply costs are much more important to the AI than having a usable army.



Hornblower -> RE: The AI (6/12/2008 9:17:05 AM)

bump



delatbabel -> RE: The AI (6/12/2008 1:46:21 PM)


quote:

ORIGINAL: ndrose

The way the AI protects the capital (though sometimes overdone, as when France alls back the entire Armee to obliterate a lone Spanish cavalry corps that, if it were unfortunate enough to get a breach, would be wiped out by the 25-strong Paris garrison) is not so bad, generally. The real problem is the way it goes after an opponent's capital--like that lone cavalry corps, or the way France attacks Berlin: the army will sit, a huge stack of corps, in Lausitz, and send one corps at a time to face the Prussian army.

And a peculiar habit with minors: even when it does declare war on them, the AI will do nothing about it if they get run by an MP it's already at war with. Now, on the one hand, this shows the AI is thinking a bit: I don't have to worry about lapse of war, since I'm at war with the MP, and can take this minor at my leisure. However, it never does do anything. Egypt is a good example. If Turkey is at war with the controlling MP, it just sits in Jerusalem. Conversely, the MP controlling Egypt starts launching suicide attacks at the Turkish army until there is no Egyptian army left. But even then the Turks don't move. And, as far as I can tell, though they may be left unmolested for years, the Egyptians never rebuild their army--not even to garrison Cairo. Both sides seem to be doing everything they can to lose....


I think that the current work that Marshall is doing, based on some of the early testing I've seen, is on the way to fixing both of these problems. Here's hoping, anyway.



Marshall Ellis -> RE: The AI (6/12/2008 2:57:15 PM)

This is my intent!



Page: [1] 2 3   next >   >>

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.9550781