AI behavior v1.2

Adanac's Strategic level World War I grand campaign game designed by Frank Hunter

Moderator: SeanD

Post Reply
Phisto
Posts: 13
Joined: Sat Jan 29, 2005 12:30 am
Location: Poolesville, MD USA

AI behavior v1.2

Post by Phisto »

I know that the AI has been tweaked for this version, so I thought that it would be worthwhile to report on its behavior. I've been playing as the CP with a France-first strategy vs. the TE AI, normal difficulty. As shown in the attached image, the CP armies have reached the outskirts of Paris in Sept-Oct 1914. Oddly, the AI has chosen to vacate Paris, while establishing a defensive line that would seem to run right through Paris.

On the other hand, the AI has started to abandon the frontier forts, and assaulted the southern flank of the CP attack toward Paris, which is why there is a large gap in the CP lines (a reasonable move). The gap would be much more problematic if it weren't possible to stroll into Paris on the next impulse (which is what happened; the large stack on the northern flank was inactive, and the gap on the southern was closed).

I should also note that the Serbs in this version have been much more frisky than before and are giving the AH significant trouble in the Balkans.

Image
Attachments
odd_AI_choices.gif
odd_AI_choices.gif (157.09 KiB) Viewed 358 times
FrankHunter
Posts: 2111
Joined: Fri Mar 26, 2004 6:07 am

RE: AI behavior v1.2

Post by FrankHunter »

The French AI is running out of troops.  Not sure why Paris was left open but my guess would be that the AI did allocate troops to defend it but for some reason they never made it.
User avatar
Lascar
Posts: 538
Joined: Sat Oct 07, 2000 8:00 am

RE: AI behavior v1.2

Post by Lascar »

ORIGINAL: FrankHunter

The French AI is running out of troops. Not sure why Paris was left open but my guess would be that the AI did allocate troops to defend it but for some reason they never made it.
While the CP AI has shown more aggressiveness in the latest 1.2 update, it still has the habit of leaving unoccupied fortress cites such as Koenigsberg and Przemysl, allowing the Russians to enter unopposed. I have seen this happen several times while testing the CP AI. It will occupy hexes adjacent to those cities and leave those important fortress cities unoccupied.
Phisto
Posts: 13
Joined: Sat Jan 29, 2005 12:30 am
Location: Poolesville, MD USA

RE: AI behavior v1.2

Post by Phisto »

ORIGINAL: FrankHunter

The French AI is running out of troops. Not sure why Paris was left open but my guess would be that the AI did allocate troops to defend it but for some reason they never made it.

The French AI actually had a large stack in Paris on the first impulse (presumably reinforcements for Sep/Oct 1914), and apparently sent those elsewhere. It did try to remedy the situation by moving troops back into Paris after the screen shot, but it was too late. The German armies arrived before they did, and French were thrown back with significant losses.
FrankHunter
Posts: 2111
Joined: Fri Mar 26, 2004 6:07 am

RE: AI behavior v1.2

Post by FrankHunter »

That sounds like I figured, the AI will (supposedly) use the closest troops available but if it was massing for an attack on the southern flank and assumed success then I could see how it would give priority to that mission over defending Paris.  Because normally capitals and fortresses are rated pretty high importance-wise and so another mission would have to be considered very important to drain troops from there.

Hard to tell with the fog but do you know if the British have landed any corps? 

Lascar, its probable that those fortresses were ordered reinforced by rail moving troops and simply never arrived either due to being attacked or because the fortress itself was attacked. 

If you guys have saves of these situations that I could run and see the same behaviour repeat I'd like to look at them just to make sure.
Phisto
Posts: 13
Joined: Sat Jan 29, 2005 12:30 am
Location: Poolesville, MD USA

RE: AI behavior v1.2

Post by Phisto »

ORIGINAL: FrankHunter

If you guys have saves of these situations that I could run and see the same behaviour repeat I'd like to look at them just to make sure.

Thank you Frank,

I know that I have the save from the screen-shot impulse, but I'll see whether I have the previous impulse when I get home this evening. I will email you what I have then.

I had not encountered any British forces up to that point.
User avatar
Lascar
Posts: 538
Joined: Sat Oct 07, 2000 8:00 am

RE: AI behavior v1.2

Post by Lascar »

ORIGINAL: FrankHunter

That sounds like I figured, the AI will (supposedly) use the closest troops available but if it was massing for an attack on the southern flank and assumed success then I could see how it would give priority to that mission over defending Paris. Because normally capitals and fortresses are rated pretty high importance-wise and so another mission would have to be considered very important to drain troops from there.

Hard to tell with the fog but do you know if the British have landed any corps?

Lascar, its probable that those fortresses were ordered reinforced by rail moving troops and simply never arrived either due to being attacked or because the fortress itself was attacked.

If you guys have saves of these situations that I could run and see the same behaviour repeat I'd like to look at them just to make sure.

If I recall correctly those fortresses were not being attacked and already had a unit in them, in the case of Prezmysl, but then moved it out allowing the Russians to just walk in.
I may still have the file saved and will send it to you if I do.
Thanks.
FrankHunter
Posts: 2111
Joined: Fri Mar 26, 2004 6:07 am

RE: AI behavior v1.2

Post by FrankHunter »

Okay, I fixed the problem.  Turned out not to be an AI issue.  Paris was considered to be "overstacked" even though the AI had ordered a couple of corps to move into adjacent frontline hexes.  But the units remaining behind were still considered to be overstacked even though that wasn't the case.  So they exited the hex.

This behaviour could occur elsewhere too, nothing special about Paris.
MarkShot
Posts: 7552
Joined: Sat Mar 29, 2003 6:04 am

RE: AI behavior v1.2

Post by MarkShot »

Another patch or beta patch then?
2021 - Resigned in writing as a 20+ year Matrix Beta and never looked back ...
Phisto
Posts: 13
Joined: Sat Jan 29, 2005 12:30 am
Location: Poolesville, MD USA

RE: AI behavior v1.2

Post by Phisto »

Thank you for looking into the problem, Frank. Does the residual overstacking effect occur with human-controlled forces as well? IIRC, overstacked hexes automatically exit forces during the impulse phases, even for human-controlled forces, but I haven't noticed vacated hexes during my own play.
FrankHunter
Posts: 2111
Joined: Fri Mar 26, 2004 6:07 am

RE: AI behavior v1.2

Post by FrankHunter »

Another patch or beta patch then?

I'll have to do another patch, I can't leave that bug in there.  There isn't any need to test it through a "beta" process though.  Besides, now I can fix that "Pskov" thing, add another screen resolution and fix a second bug that hjaco found in the supply penalty after Germany has started facing food shortages.

Does the residual overstacking effect occur with human-controlled forces as well?

Yes
but I haven't noticed vacated hexes during my own play.

Right, because most of the time it wouldn't happen in quite that manner.  But in this case it was the two corps ordered to defend the level 4 fort on Paris that were chosen to fulfill the stacking obligation.  If the game had chosen other units nothing would have happened as they were moving anyway.


Post Reply

Return to “Guns of August 1914 - 1918”