How AI takes strong enemy positions

Please post here for questions and discussion about scenario design and the game editor for WITP.

Moderators: wdolson, Don Bowen, mogami

Post Reply
el cid again
Posts: 16983
Joined: Mon Oct 10, 2005 4:40 pm

How AI takes strong enemy positions

Post by el cid again »

It was watching Palembang I learned the way it works. I had seen it all before - at Bataan, at Manila,
at Singapore, at Hong Kong - but didn't understand it. Finally I looked at things from the ALLIED point of view.

In the latest medium term test of RHSEOS - to my surprise AI decided to attack Palembang at the same time
as Tabaoli (Banka) - both major supply sources and both with RHS supply sinks of some size in consequence.
It was as dumb as ever - I thought - land a fairly large force - support it with nothing - no bombardment by ships -
no air strikes - not even supplies. Just put a force in the hex big enough it won't be wiped out by any attacks.
Indeed, any attacks just hurt the Allies - they lose things (squads, supplies, morale) faster.

What AI then does is - nothing!

Turns out this is anything but stupid. This was probably coded to deal with Bataan - 40 units go there - representing IRL not less than 80,000 soldiers - maybe 110,000 (depending on who you believe). How do you beat two corps (which is what was at Bataan) with a vastly inferior force (which is what attacks Bataan IRL and in the game?

Wait.

Seems - even if the hex is a supply source - it produces NO supplies if the enemy is in the hex. If you wait, the enemy gets down to nothing supply wise. His morale follows that. And he also has fatigue levels skyrocket. Not sure how long it waits - it is months: might be until attacking units are planned up 100% - up to 100 days. But THEN the AI attacks - shock attacks - and it cleans up. Either the position surrenders, or survivors retreat into the bush.
User avatar
witpqs
Posts: 26376
Joined: Mon Oct 04, 2004 7:48 pm
Location: Argleton

RE: How AI takes strong enemy positions

Post by witpqs »

Well, several versions ago in RHSEOS...

The Japanese AI, without landing anything at Palembang, landed a lone artillery unit at Banka. I did not order it attacked, but Allied bombardment attacks continued from teh opposed landing. Several turns later it was 'Destroyed through attrition'.

A few weeks later the AI did it again...

I wish the AI were always as smart as you found.
User avatar
akdreemer
Posts: 1028
Joined: Sun Oct 03, 2004 12:43 am
Location: Anchorage, Alaska
Contact:

RE: How AI takes strong enemy positions

Post by akdreemer »

ORIGINAL: el cid again

It was watching Palembang I learned the way it works. I had seen it all before - at Bataan, at Manila,
at Singapore, at Hong Kong - but didn't understand it. Finally I looked at things from the ALLIED point of view.

In the latest medium term test of RHSEOS - to my surprise AI decided to attack Palembang at the same time
as Tabaoli (Banka) - both major supply sources and both with RHS supply sinks of some size in consequence.
It was as dumb as ever - I thought - land a fairly large force - support it with nothing - no bombardment by ships -
no air strikes - not even supplies. Just put a force in the hex big enough it won't be wiped out by any attacks.
Indeed, any attacks just hurt the Allies - they lose things (squads, supplies, morale) faster.

What AI then does is - nothing!

Turns out this is anything but stupid. This was probably coded to deal with Bataan - 40 units go there - representing IRL not less than 80,000 soldiers - maybe 110,000 (depending on who you believe). How do you beat two corps (which is what was at Bataan) with a vastly inferior force (which is what attacks Bataan IRL and in the game?

Wait.

Seems - even if the hex is a supply source - it produces NO supplies if the enemy is in the hex. If you wait, the enemy gets down to nothing supply wise. His morale follows that. And he also has fatigue levels skyrocket. Not sure how long it waits - it is months: might be until attacking units are planned up 100% - up to 100 days. But THEN the AI attacks - shock attacks - and it cleans up. Either the position surrenders, or survivors retreat into the bush.
The manual states that no supplies, no resources, are generated in a hex that contains friendly and enemy units.
el cid again
Posts: 16983
Joined: Mon Oct 10, 2005 4:40 pm

RE: How AI takes strong enemy positions

Post by el cid again »

ORIGINAL: witpqs

Well, several versions ago in RHSEOS...

The Japanese AI, without landing anything at Palembang, landed a lone artillery unit at Banka. I did not order it attacked, but Allied bombardment attacks continued from teh opposed landing. Several turns later it was 'Destroyed through attrition'.

A few weeks later the AI did it again...

I wish the AI were always as smart as you found.


Well - it is dumb as a post. It routes oil and resources out near Palau en route to Osaka! That is, they go SOUTHEAST INTEAD OF NORTHEAST from Borneo and nearby points. It fails to load many ships bound for Japan with anything useful - which only has the merit of being typical of Japanese management style. It escorts almost nothing. So - yeah - it remains pretty dumb.

But now (some weeks later) I see it pounding Palembang and Taboali with bombers - prepratory to its assaults.
It ain't totally stupid.
el cid again
Posts: 16983
Joined: Mon Oct 10, 2005 4:40 pm

RE: How AI takes strong enemy positions

Post by el cid again »

ORIGINAL: AlaskanWarrior
ORIGINAL: el cid again

It was watching Palembang I learned the way it works. I had seen it all before - at Bataan, at Manila,
at Singapore, at Hong Kong - but didn't understand it. Finally I looked at things from the ALLIED point of view.

In the latest medium term test of RHSEOS - to my surprise AI decided to attack Palembang at the same time
as Tabaoli (Banka) - both major supply sources and both with RHS supply sinks of some size in consequence.
It was as dumb as ever - I thought - land a fairly large force - support it with nothing - no bombardment by ships -
no air strikes - not even supplies. Just put a force in the hex big enough it won't be wiped out by any attacks.
Indeed, any attacks just hurt the Allies - they lose things (squads, supplies, morale) faster.

What AI then does is - nothing!

Turns out this is anything but stupid. This was probably coded to deal with Bataan - 40 units go there - representing IRL not less than 80,000 soldiers - maybe 110,000 (depending on who you believe). How do you beat two corps (which is what was at Bataan) with a vastly inferior force (which is what attacks Bataan IRL and in the game?

Wait.

Seems - even if the hex is a supply source - it produces NO supplies if the enemy is in the hex. If you wait, the enemy gets down to nothing supply wise. His morale follows that. And he also has fatigue levels skyrocket. Not sure how long it waits - it is months: might be until attacking units are planned up 100% - up to 100 days. But THEN the AI attacks - shock attacks - and it cleans up. Either the position surrenders, or survivors retreat into the bush.
The manual states that no supplies, no resources, are generated in a hex that contains friendly and enemy units.


Sure does. And I forgot that tactically - but whoever wrote code didn't.
bbbf
Posts: 490
Joined: Sun Jul 16, 2000 8:00 am
Location: Hobart, Tasmania, Australia

RE: How AI takes strong enemy positions

Post by bbbf »


[

Well - it is dumb as a post. It routes oil and resources out near Palau en route to Osaka! That is, they go SOUTHEAST INTEAD OF NORTHEAST from Borneo and nearby points. It fails to load many ships bound for Japan with anything useful - which only has the merit of being typical of Japanese management style. It escorts almost nothing. So - yeah - it remains pretty dumb.

But now (some weeks later) I see it pounding Palembang and Taboali with bombers - prepratory to its assaults.
It ain't totally stupid.
[/quote]


Actually, I don't think this is too dumb. Most Allied human players will stock the NE route with plenty of subs (actually I think the AI does this as well). The AI doesn't play the ASW game like a human player (who will tend to make life difficult for those Allied subs), so routing via Palau actually probably will result in less losses to subs - unless the Allied player works out that is where they are.
Robert Lee
Mike Scholl
Posts: 6187
Joined: Wed Jan 01, 2003 1:17 am
Location: Kansas City, MO

RE: How AI takes strong enemy positions

Post by Mike Scholl »

"Actually, I don't think this is too dumb. Most Allied human players will stock the NE route with plenty of subs (actually I think the AI does this as well). The AI doesn't play the ASW game like a human player (who will tend to make life difficult for those Allied subs), so routing via Palau actually probably will result in less losses to subs - unless the Allied player works out that is where they are."


I wouldn't give the designers too much credit for this. The basic rule for routing ships seems to be for the US to go West, then South or North; and the Japanese to go East, then South or North. I doubt there is anything more subtile than that involved. Which is why so many people want "waypoints"
Hornsby
Posts: 7
Joined: Mon Jul 24, 2006 4:31 pm
Location: Glasgow, Scotland

RE: How AI takes strong enemy positions

Post by Hornsby »

Actually I have a newfound respect for the code. I recently discovered that TFs on a transport mission will alter course to avoid sighted submarines, also if a TF suffers an attack by land-based air, subsequent TFs will alter course to give an extra 2 - 4 hexes buffer to the base from which the attack came.
However there do have to be attackes before the AI will alter routes, so care must be taken if you wish to limit losses.
User avatar
Sardaukar
Posts: 12464
Joined: Wed Nov 28, 2001 10:00 am
Location: Finland/Israel

RE: How AI takes strong enemy positions

Post by Sardaukar »

Also noticed that AI "routine" TFs tend to change route slightly if suffering sub attacks. That came as surprise.
"To meaningless French Idealism, Liberty, Fraternity and Equality...we answer with German Realism, Infantry, Cavalry and Artillery" -Prince von Bülov, 1870-

Image
Post Reply

Return to “Scenario Design”