Matrix Games Forums

Forums  Register  Login  Photo Gallery  Member List  Search  Calendars  FAQ 

My Profile  Inbox  Address Book  My Subscription  My Forums  Log Out

1.07 AI observation

 
View related threads: (in this forum | in all forums)

Logged in as: Guest
Users viewing this topic: none
  Printable Version
All Forums >> [Current Games From Matrix.] >> [Napoleonics] >> Empires in Arms the Napoleonic Wars of 1805 - 1815 >> 1.07 AI observation Page: [1] 2   next >   >>
Login
Message << Older Topic   Newer Topic >>
1.07 AI observation - 10/2/2009 4:15:09 PM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline
The AI is doing some things better. But I've noticed an odd behavior I haven't seen before: a flight-to-the-capital panic when there's no occasion for it. So far I've seen both Russia and Turkey do this, rather disastrously.

Russia did it in Jan 1805. Having set up forces in the Caucasus and on the Turkish frontier, it immediately pulled them back to Moscow, evidently foraging them willy-nilly, with the high losses you'd expect. Russia wasn't being invaded, or even at war with any MP.

Turkey did it later on, pulling its forces out of Egypt (it had crossed the border and then sat there; it did a better job of this in 1.06) and Palestine, rushing back towards Constantinople and foraging with high losses. As above, not invaded or at war with anyone.

What's it trying to do? I've seen this sort of thing in previous versions only when the capital is besieged. Sometimes in those cases the army was too distant to save the city, so maybe this is an attempted improvement to make it anticipate threats? If so, its tripwire tension is set way too high.
Post #: 1
RE: 1.07 AI observation - 10/2/2009 7:27:34 PM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline
Following up on this, I started another test game, and saw the same thing with Russia again. In both cases, Russia had DoWed Sweden, and a Swedish corps had set up in Finland on the Russian border. So my guess about the AI's reasoning here is that the Swedish corps threatens St Petersburg (which it's counting as a capital).

In this case, Russia went so far as to let a DoW on Circassia lapse so that the corps there could march north. It also foraged its cavalry to bring them home from the Prussian border. And (this is really silly) it drew most of the corps it had massed on the Finnish frontier back to Petersburg, presumably to hunker down for the coming siege of the Swedish hordes. (It did send one corps over the border, at least.)

As a side note, the AI didn't build even depots for the corps it was marching north, which it could easily have done. I thought it had been doing better with that in 1.06, so I don't know what's happened here.

Other random note: Austria DoWed some minors it could technically reach, but only with cav, and let those lapse.

(in reply to ndrose)
Post #: 2
RE: 1.07 AI observation - 10/2/2009 7:27:53 PM   
sw30

 

Posts: 397
Joined: 9/20/2000
From: San Francisco, CA
Status: offline
Just saw Spain do it the same turn that it DOWed Portugal.  It's at war with nobody, DOWed Portugal with units around (Good!) ran all forces towards Madrid (Bad!) and lapsed the war (Oh noes!)

(in reply to ndrose)
Post #: 3
RE: 1.07 AI observation - 10/3/2009 3:11:58 AM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline
That's interesting; so that's several cases now where it's in reaction to a perceived threat from a minor the AI is supposed to be invading.

I just saw another case with Turkey, but there were no enemy corps (even minor ones) anywhere near Constantinople. Looking at the log, the only thing I see is that at the beginning of the month, Austria and Russia became allies. On the one hand, it's good that the AI for Turkey recognizes this is a threat (if that is indeed what it was responding to). On the other hand, it doesn't warrant foraging all the Turkish corps back from distant parts of the empire. (And in the process lapsing wars with Egypt and Dalmatia.)

Russia did DoW Turkey the following month, but as all its corps were in the far north (see above), it was meaningless.

(in reply to sw30)
Post #: 4
RE: 1.07 AI observation - 10/3/2009 3:26:23 AM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline
Later in the same game, France DoWed Austria, and Russia came to Austrian support. Here's where things really get weird.

First, all the Russian corps that had been heading for St Petersburg veered off to Moscow, and most of those that were already in St Petersburg moved towards Moscow. (Actually leaving it open enough that it could be vulnerable to a Swedish attack.)

France didn't have any corps (apart from the Swedish one) east of Venice, but OK, it's at least nice to see the Russians not sitting in Grodno waiting to be annihilated by Nappy prior to an unopposed march to Moscow.

But then, once they had gathered in Moscow, they all set out for different fronts again: some to Prussia (again, nice to see them actually crossing the frontier, but it would have been better if they hadn't foraged to Moscow and back first), and some back to the Caucasus, presumably to attack the French-controlled Persians. I say presumably, because they all foraged to death before they got there, the last factor expiring just as it entered Circassia, whence it had started so many months before....

I can see that the AI is actually trying to think further ahead than it used to, so that's good. But at the moment the results are worse, because the left hand doesn't know what the right hand is doing (DoWing a minor then running away from it; gathering in Moscow then fanning out to seek the enemy), and because of mechanical troubles (not building depots). On that last point, it's not easy to tell from the log exactly what it's doing, but I believe that in one example, it fed several infantry corps from the depot they had marched away from rather than building a new one where they marched to (I believe this because the log showed no foraging losses in those corps although it was Russian winter and they'd used their movement allowance), then (out of money, I guess) foraging the cavalry and losing factors there.

(in reply to ndrose)
Post #: 5
RE: 1.07 AI observation - 10/5/2009 12:56:55 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Can someone send me a game when they start moving to their capital?


_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to ndrose)
Post #: 6
RE: 1.07 AI observation - 10/5/2009 2:36:33 PM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline
Posted at Mantis.

(in reply to Marshall Ellis)
Post #: 7
RE: 1.07 AI observation - 10/6/2009 12:47:48 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Good stuff guys!
This is the feedback that I need from the BETA!



_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to ndrose)
Post #: 8
RE: 1.07 AI observation - 10/6/2009 4:00:20 PM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline
Strangely, when I played Russia and actually DoWed Turkey, it did not rush back to the capital, and in fact let its corps sit idly by in the Balkans while I besieged Constantinople; but in another game, Turkey ran home for no reason at all, lapsing wars in Egypt, the Caucasus, and Dalmatia.

(in reply to Marshall Ellis)
Post #: 9
RE: 1.07 AI observation - 10/7/2009 1:14:44 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
The AI seems to be freaking at some point in a major war???
Odd indeed! Will be looking at this!


_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to ndrose)
Post #: 10
RE: 1.07 AI observation - 10/7/2009 1:50:36 PM   
ndrose

 

Posts: 612
Joined: 10/13/2006
Status: offline

quote:

ORIGINAL: Marshall Ellis

The AI seems to be freaking at some point in a major war???
Odd indeed! Will be looking at this!




I would say it's being *less* active in war. It's at times of peace that it suddenly panics and sends everybody back to the capital. Looking at my last few AI games, it seems as if most of the countries do this at the very beginning, effectively negating whatever set-up position they've carefully chosen, costing themselves money, and often incurring foraging losses. The problem's especially bad with Russia and Turkey, since they have so far to go, and end up losing a lot to foraging. (Almost nobody makes it home from Palestine or the Caucasus.) But the other countries are doing it too.

(in reply to Marshall Ellis)
Post #: 11
RE: 1.07 AI observation - 10/7/2009 1:58:40 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Good info! Thanks!


_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to ndrose)
Post #: 12
RE: 1.07 AI observation - 10/8/2009 8:04:13 PM   
Skanvak

 

Posts: 577
Joined: 4/3/2005
Status: offline
I still play in 1.063 but I have seen a lot of what you refer with the AI : Staking on the Capital (Russia at war with Turkey : one corps in Turkey, 5 on St Petersburg). It seems that the unique strategy of the AI is to stack on the capital and wait. Austria did that, and Austria and Prussia made succide Attack on my besieging Forces (anout 50 factors against less than 5...)

For the lapse of War : Russia dow Sweden and never moved in. Spain dow Algeria from Morroco, the Army stopped in the first area an never moved to Alger for several month until the current turn (no change). Same thing for Turkey in Egypt One Army in moved at limit of supply but do not push to Cairo.

I wonder, Marshall, that you should make template for the AI to invade Minor (ie choose a plan an compell the AI to follow it; not bright but to take garrisonned only or 1 corps minor that should be enough).

England Dow me (France) with her fleet in port, moving naval phase after me (lille crossing enabled...) so I just have to blockade her fleet.

Other strange behaviour already reported : Englande blockade the French fleet with only the Portuguese fleet, that is a 10 ships fleets against about 40... During this war England let her corps in Portugal even when I was besieging London with only one corps.

I have a Question : does the AI memorized the size of the player corps? It seems that the AI just caculated relative strength with the corps number and not the factors in it. AI should have a data base of its estimation of the player corps if it does not already have it.

I have a question too : How to do diplomacy with a non-allied AI? I will try to think about an AI scheme that is simple an enjoyable but not smart (as I really believe that someone alone cannot make a smart diplo AI look at the quaote below from wikipedia).

quote:

The Diplomacy Artificial Intelligence Development Environment (DAIDE) project is a hobby project to produce computer AIs capable of playing Diplomacy, and allow them to play against each other and against humans. The AIs being produced by this project are mostly only capable of playing games without negotiation, and are somewhat weaker than most human players, although several of the AIs are significantly stronger than the AIs in any of the commercially released Diplomacy games.


I was looking to see how the AI behave in the Diplomacy Computer game as I though it might help in designing the AI for EIANW, but looks like it is a failure...

One suggestion : implement the Uncrotrolled Major Power rules of the original EiA for multiplayer. That will save this aspect of the game from this abyssal AI.

_____________________________


Best regards

Skanvak

(in reply to Marshall Ellis)
Post #: 13
RE: 1.07 AI observation - 10/10/2009 6:21:25 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
NDRose stole my title...hmmmph!

France Hard Solitaire

Have not witnessed the wholesale retreat Rose sees, but have always seen variations on it. Particularly the no reason lapsing ala Portugal he described.

British AI....Godawful would be a complement. As in Previous 2 versions it is throwing its army cross channel endlessly. 3-4 factors vs. 17+ is not getting it done...but its still trying! Then in late 1807 it decided the Navy should park off Lille abandoning ALL blockades. After several turns of this a Dutch Corps landed in Edinburgh, then a Hessian and then a French. While this happened the Brits promptly attacked Lille as 3 Non-English capitals occupied. New army wiped out London now under siege by 3 Corps 9/08. Brit AI hasn't been this bad in 5-6 versions. but at least it didn't retreat to capital....as it should have.

Russ concentrated at St Petes and locked their 10+ Corps vs. 2 Swed, 1 Fr. Fr advance, take Capital, garrison it. Defeat Russ counterattack. Retreat to Vyborg, rebuild and repeat as necessary. Have taken and lost St. Pete's 3-4 times, Russ merely have to advance to Vyborg, wipe out 10-12 factors and be done but no, they are passive letting shattered force rebuild and losing tons of VP's.


Turk's blew war against Egypt. Stayed neutral to mid '08 then attacked Russ gaining Aust enemy.

Aust stayed neutral till CTA by Russ.
Desultory maneuvering on frontiers while French quickly march to Vienna sieging it between bouts of beating Aus army. Fr & Wurt Corp wiped out of INF but arrival of full 1st Corp keeps Fr on Vienna despite sacrificing 9 Gds after tuning Aust 3-4 times and loss of all Reg INF.

Spain is quiet took Port and gained Naples.

Russ entered war early after botching Finn invasion it declared on Fr. Sent 1 Coss to Pruss who have since UNCOND 2x. Kept maybe 1 corps near Turks, none in Caucasus (may have pulled a Rose and run early). Lost 1-2 corps in Grodno to Fr early. Bulk of Army near St. Pete's where it is gaining ignominy for itself.

Fr owns every minor of Germany, 1/2 of Prussia and all Italy but Palma & Naples plus all of North Afr. Sp lapsed vs, Morr as Fr Morr. saw no troops nearby so Morr took empty Gibralter, which should, with Malta, start with Garr of 1-3 INF at each point. Removed Corps by sea something Brits never allowed, they always attack Fr fleets in that area.

Have large arrows showing my moves pop up today. No idea where they came from and no option to be rid of them. They weren't here on 2 previous sessions of Fr game.

Interface very insensitive, forced to click and click to activate a corps then can't deactivate it w/o boatloads of right-clicks. Getting old fast


(in reply to Skanvak)
Post #: 14
RE: 1.07 AI observation - 10/10/2009 6:26:57 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
Good Stuff 9/1808

Revised VP is working.

Fr 51%

UK 49

Sp 58

Pr 39

Tu 57

Rs 48

Au 55

Status
Fr pegging the meter Dominant
Sp Normal
Pr Fiasco
All other Instab

(in reply to easterner)
Post #: 15
RE: 1.07 AI observation - 10/14/2009 5:46:04 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Easterner:

The rt-clicking issue is odd??? Can you tell me more???


_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to easterner)
Post #: 16
RE: 1.07 AI observation - 10/17/2009 2:04:21 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
Similar to interface issue I had 1.04 ish.

Lets say I want to move a Corps. I'll click on it, nothing happens. Several frustrated clicks later it asks if I want a FORCED MARCH. Next click shuts that off and one more click and I (usually) get to move it. Then I go to move next corps and original one moves again if possible. So I fix it, hit right click to disengage it and several clicks needed to turn off 'lit-up' possible moves. Full moves cause no problems but if say I move Fr. Corps 3 areas it still has a MP left and doesn't want to disengage as I try to move next corp in stack.

Speaking of which, any reason we don't have a Corps Stack Move button ala the Fleet Stack Move button?


If no one else suffers this it may be mouse issue here of course.

< Message edited by easterner -- 10/17/2009 2:05:16 PM >

(in reply to Marshall Ellis)
Post #: 17
RE: 1.07 AI observation - 10/19/2009 1:31:31 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Hmmmmm? That is a new one on me!
We cannot move land units on a stack basis (Original EiA and EiANW) since there are several single corps movement triggers (Insurrection, foraging) that require corps to be moved one at a time.




_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to easterner)
Post #: 18
RE: 1.07 AI observation - 10/19/2009 6:45:15 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
Turkey Hard Solitaire

Prussia Allied with Turkey 1/05. Declared war on Russia and Austria.
Russia & Austria attacked Pr hard. Russian army badly mauled. Austria surrendered UNCON to Pr & Cond to Turk.

France somnolent as usual.

VP not so good.

59 Fr
62 Br
64 Sp
78 Pu
39 Tu
59 Ru
57 Au

Current 1/09

(in reply to Marshall Ellis)
Post #: 19
RE: 1.07 AI observation - 10/20/2009 12:53:46 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Dang hard to win as Tu!
Do the VPs seem to be getting alloted accurately?


_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to easterner)
Post #: 20
RE: 1.07 AI observation - 10/21/2009 3:12:35 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
VP

Difficult to say. Need play a game with each country 1st. But Turks seemed to have way to hard a time of it. Pr won the game and it was well deserved victory smashing Russia and CONQ Au simultaneously despite vigorous assaults on Pr.

New
HARD Brit Solitaire

10/06
25 Fr
20 GB
29 Sp
31 Pr
26 Tu
30 Ru
28 Au

Brit is only Dominant, all others Normal.

Au is deep in Tu.

< Message edited by easterner -- 10/21/2009 3:26:23 PM >

(in reply to Marshall Ellis)
Post #: 21
RE: 1.07 AI observation - 10/21/2009 5:26:58 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
AAR becoming really annoying. Needs be moved to top tool bar. Generates severe Rt-clk problem, especially in naval phase.

Brit Game

15 or so Russ corps are stacked on St. Pete's
10 or so Tu at Const.
All Sp but 1 at Madrid
5-8 at Paris This cost Fr a trip to INSTAB in Ital/Swede Campns vs. GB.
Entire Au army less 1 at Temesvar, none at Vienna.
Pr not stacked heavily on Berlin.

GB declare war on Genoa in hopes Ally got it creating a buffer zone across N. Italy giving GB a free hand. But it was Winter and impossible to get there

(in reply to easterner)
Post #: 22
RE: 1.07 AI observation - 10/21/2009 7:02:52 PM   
pzgndr

 

Posts: 2688
Joined: 3/18/2004
From: Maryland
Status: offline
quote:

AAR becoming really annoying. Needs be moved to top tool bar. Generates severe Rt-clk problem, especially in naval phase.


I agree. I like the new AAR feature but the right click needs to be reconsidered.

(in reply to easterner)
Post #: 23
RE: 1.07 AI observation - 10/22/2009 1:09:41 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
LOL! I was thinking the same thing while playing the other morning!
I will look at a button option somehow??? I'm running low on button space so give a me a bit to figure this out!


_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to pzgndr)
Post #: 24
RE: 1.07 AI observation - 10/23/2009 2:21:10 AM   
Dancing Bear

 

Posts: 1003
Joined: 2/21/2008
Status: offline
Can you guys post a picture of the AAR problem? I'm still on 1.06, but would like to see what the problem is.

(in reply to Marshall Ellis)
Post #: 25
RE: 1.07 AI observation - 10/23/2009 12:31:45 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
DB:

The problem is that the AAR menu pops up on a rt-click and since there are so many rt-clicks already implemented on the map (de-select unit) the menu is popping up more and kind of irratating!

_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to Dancing Bear)
Post #: 26
RE: 1.07 AI observation - 10/23/2009 8:44:52 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
There is no 'problem' the pop-up jumps out while moving as AAR needs right-click to activate and a moving unit does too to stop moving. So it is annoying and frustrating.

(in reply to Marshall Ellis)
Post #: 27
RE: 1.07 AI observation - 10/26/2009 4:56:25 PM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
Brit Game Completed

Prussia took 1st & Sp 2nd. Neither ever fought a war.

Fr- 86
Br- 71
Sp- 96
Pr- 100
Ru- 95
Tu- 82
Au- 93

Au surrendered to Fr in France's one war.
Tu surrendered to Au, Pr & GB
Brit spent a lot of time in DOM zone

Ru has not only lost the ability to CONQ Sweden, it can't even handle an unoccupied Finland.

Sp continually declares on Morocco but has no troops in range causing a lapse everytime. I believe ndrose noted this tendency.


< Message edited by easterner -- 10/26/2009 6:59:04 PM >

(in reply to ndrose)
Post #: 28
RE: 1.07 AI observation - 10/26/2009 10:39:54 PM   
Marshall Ellis


Posts: 5630
Joined: 10/2/2001
From: Dallas
Status: offline
Easterner:

Dod you have a save where Sp is declaring without forces in range? It should NOT be doing this so I would love to see this game if you can?



_____________________________

Thank you

Marshall Ellis
Outflank Strategy War Games



(in reply to easterner)
Post #: 29
RE: 1.07 AI observation - 10/27/2009 4:16:58 AM   
easterner

 

Posts: 179
Joined: 10/9/2004
Status: offline
No, I just note that Morr becomes Free and no sign of Sp in area.

Just started an AU game. Au has already had Tu UNCON and had a White Peace with Pr. Yet lags badly on VP.

I'll start a Pr game and see where Sp starts and what it does.

(in reply to Marshall Ellis)
Post #: 30
Page:   [1] 2   next >   >>
All Forums >> [Current Games From Matrix.] >> [Napoleonics] >> Empires in Arms the Napoleonic Wars of 1805 - 1815 >> 1.07 AI observation Page: [1] 2   next >   >>
Jump to:





New Messages No New Messages
Hot Topic w/ New Messages Hot Topic w/o New Messages
Locked w/ New Messages Locked w/o New Messages
 Post New Thread
 Reply to Message
 Post New Poll
 Submit Vote
 Delete My Own Post
 Delete My Own Thread
 Rate Posts


Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI

0.153