More weird AI (Full Version)

All Forums >> [New Releases from Matrix Games] >> Gary Grigsby's War in the East Series >> Tech Support


Schmart -> More weird AI (10/4/2012 6:07:38 AM)

I suppose this isn't really a bug or anything, just some info to perhaps help future AI programming.

The Axis AI has created a huge and vulnerable bulge. See below. It seems to be attacking deeper into salients,
and paying little attention to flanks and rear areas. If I really wanted to, I could cut-off close to half the German
army in Russia. Rail lines into the bulge have been cut at Kursk and west of Orel so supply must be on a shoestring.
The Orel-Tula flank has been very lightly defended. Again, I could rush that sector and create havoc. And yet, the AI
keeps pushing deeper east.

The panzer ball pushing NE of Tula has finally been contained and halted, but instead the AI pulls back those forces
and throws them further east into the salient NW of Tambov and north of Voronezh, deeper into an already vulnerable bulge.
I have gradually been pressuring the bulge over the winter from the south in the Kursk-Voronezh area, and in the north in
the Orel-Bryansk sector, with seemingly little additional reaction from the AI. It was a slow slog, but I eventually cut the two
rail lines into the bulge (in the Kursk and Orel vicinities).

I know the AI has limits, but I have found the Russian AI surprisingly competent at avoiding encirclements. The same
can not be said of the Axis AI. During the first blizzard, 1 Div here, 2 Divs there, etc can fairly easily be cut off without
any pull back by the AI (even when it very easily could pull back just single hexes), eventually tearing the Axis to
shreads if one really wanted to. Now into 1943, and again very little effort on the AI to avoid vulnerable salients, bulges,
and encirclements.

Overall, I find the Russian defensive AI competent as well as the Axis offensive AI. Some "cross-importation" might be
usefull to beef up the Axis defensive AI and Russian offensive AI.


DrewBlack -> RE: More weird AI (10/4/2012 2:49:38 PM)


What levels are you playing at percentage wise for the Ai?


Schmart -> RE: More weird AI (10/4/2012 5:29:29 PM)

Been playing mostly on 110/90 in favour of the AI. I bumped it up to 115/85 for the summer/fall of 1942 to give the AI a boost, but then this crazy bulge started developing, with the AI pushing relentlessly (I just couldn't stop it) further and deeper into the vulnerable bulge. I bumped it back down to 110/90 thinking that it was just stupid to have the AI keep pushing into these unrealistic narrow salients.

I can understand the push NE of Tula (combined with the AI gradually grinding away on Kalinnin and Western Fronts) as it threatens my positions south of Kaluga with encircelment, but then the AI started pushing NW of Tambov which in my mind is a thrust into nowhere. The AI has some of its best Inf Divs NW of Tambov, but won't do anything to secure the neck of the larger bulge and re-establish the rail lines at Orel and Kursk.

DrewBlack -> RE: More weird AI (10/4/2012 7:17:44 PM)


Why do you put your side at a negative figure always find the Ai works better if you keep your percentage at 100% and increase the Ai's especially Moral past the 118% figure, and the others from 105 and up, personnal choice I suppose.

I say go on, close the bottle neck see how the Ai responds. ;-) You may be surprised or there agian you may not be!!!


Schmart -> RE: More weird AI (10/4/2012 7:46:57 PM)

I guess I put it at those settings becuase that's how the difficulty settings work with the game. I never gave it much thought, TBH. So if I wanted a 20 point spread, you're saying it's better to have it 120/100 rather than 110/90? I thought at 120% morale is when the AI 'cheating' settings kick in. I suppose I could always put it at 119 then...

Von Hindenburg -> RE: More weird AI (10/5/2012 4:13:33 PM)

There is a huge difference between 119 and 120!

DrewBlack -> RE: More weird AI (10/5/2012 7:21:27 PM)


Thats why I go to 118% on moral.... Ai sees a higher moral and chooses where to attack with a higher degree of certainty in its favour. NOw 120% + thats gonna hurt as the Ai does cheat at those levels..... i actually remember reading some while back that 111% and above starts the cheating on moral setting but I may have dreamed it. AGAIN!!!


Joel Billings -> RE: More weird AI (10/6/2012 7:14:59 PM)

From the rules:

Whenever the morale level is set to 110 or greater, then leader admin checks for movement
allowances are always successful (14.1.2) and the Soviet AI is able to build combat units so as
to maintain a force of 300 divisions and 100 brigades at no additional admin cost (see section
12.2.4). In addition, whenever the morale level is set to 125 or greater, all leader initiative
checks for movement allowances points are also automatically successful.

Those are the only "cheat" break points that I know of.

The AI when on the offensive is pretty lousy at flank security. We found that changes made to provide better flank security ended up making the AI much too timid. We were not able to get this to work any better. My suggestion would be not to take advantage of obvious mistakes like leaving a section of the front open. I'd suggest moving into the area slowly instead of just trapping the AI all at once. I realize this is not the ideal, but we're not likely to be able to solve this basic issue with the AI without neutering it so much that it would lose any ability to take the offensive.

AFV -> RE: More weird AI (10/7/2012 4:05:10 AM)

Schmart- I am curious, what is the AI/player ratio at the top of the screen (looks like xxx:xxx). I think it only displays during the AI turn.

The reason I ask is because it appears to me that when the AI calculates it has an advantage (say the ratio is 140:100 or something like that), it is fearless, does not retreat seemingly for any reason (even the one hex thing that would save a unit from destruction at least for one more turn, as you have seen). Once that calculated ratio is in your advantage (say like 100:140), then the AI will start to retreat, and seems to play "smarter"- or perhaps more cautios as the situation would dictate.

Edit, add
So the ratio it calculates might be valid for the whole map, but NOT valid locally, but it acts as if it has the advantage regardless, even in local areas where it clearly does not have an advantage at all. Where it should be retreating but instead holds or even advances.

Page: [1]

Valid CSS!

Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI