Suggestions for AI player improvements...

Gary Grigsby's strategic level wargame covering the entire War in the Pacific from 1941 to 1945 or beyond.

Moderators: Joel Billings, wdolson, Don Bowen, mogami

Post Reply
User avatar
Apollo11
Posts: 25430
Joined: Thu Jun 07, 2001 8:00 am
Location: Zagreb, Croatia
Contact:

Suggestions for AI player improvements...

Post by Apollo11 »

NOTE: I posted this some time ago on UV forum but, what the heck, why not post it here as well... <VBG>


Hi all,

Yes... we all know that the only real thrill for playing UV is PBEM but there
are many many UV users who never play PBEM and their only opponent is AI
player...

Also even we, who play PBEM, sometimes want to test some new tactics/strategy
and for that the AI is our only possible opponent.


The UV was improved greatly with all patches (thanks again Matrix/2By3 !!!)
but one aspect of it never got overhauled.

The AI in current UV v2.30 looks/feels almost the same as it was in UV v1.00
long ago.


Perhaps if we, as good UV PBEM players, can give some advice to Gary and other
good people at Matrix/2By3 for the AI player to be improved.


IMHO, we have the knowledge and experience!


If you aren't already bored to death please bear with me and read the long
document I wrote below (in fact this is something I wrote ON and OFF for some
time now)...



Leo's suggestions for AI player improvements


a) Overall strategy posture in scenarios

#1
The AI player should have at least 3 overall strategies (postures):
- offensive
- defensive
- combined


#2
Unless the scenario is predetermined (i.e. about one side markedly set to be
on offensive and other on defensive) the AI player should choose its strategy
posture.

The good idea would be that, at the beginning of each scenario, the AI player
would randomly choose the overall strategy posture from list of available ones
(except in predetermined scenarios).


#3
During scenario duration the AI player should (several times) check the
situation (bases won/lost, victory points, losses etc.) and re-check the
overall strategy posture. If appropriate AI player would then change it
accordingly (for example from offensive to defensive).

This all would ensure that AI would never behave the same (i.e. robot like) in
every situation. Randomness and variety is wonderful thing...



b) Bases in scenarios

#1
In every scenario the AI should "know" which bases are key (strategic) ones
(and this should apply to both - its own and enemy ones).

Every base should be evaluated and ranked (this shouldn't be the problem since
we already have the "value" number for every base and it's possible MAX value
when fully developed).

I think that four ranks would suffice:
- supply bases (like Turk, Brisbane and Noumea)
- primary bases (like Rabaul, Port Moresby, Lunga)
- secondary ones (like Shortlands, Lae, Luganville, Townsville, Cairns...)
- tertiary ones (all other good bases that can be reasonably developed)
- all others (minor bases with no development perspective)


#2
According to overall strategy posture the AI player would then decide what to
do with his own and enemy bases.

In other words this overall strategy would dictate whether to defend or attack
or try to do both.

Note that the friendly supply bases would always be first to defend (then the
primary and so on).


#3
After accessing current situation and overall strategy posture the number and
targets for possible attacks (in accordance with ranking of enemy bases to be
taken) should be decided.

This would also dictate the way AI player would behave (i.e. whether to go for
the enemy primary bases directly or first go after secondary/tertiary ones).



c) Defense strategy

#1
Friendly supply bases should be best defended against enemy attack.

The appropriate amount of garrison troops in them should be around one
division (i.e. several regiments together with HQ and support units like AAA,
engineers and coastal guns).

This should always be the case regardless of overall strategy posture.


#2
Other friendly bases would be defended according to their rank to minimal
capacity.

The first to come after supply bases would be primary bases and so on.

Also the additional defenses (i.e. more than minimal) would be dictated with
overall strategy posture for AI player (i.e. whether it is offensive or
defensive or combined).


#3
If attacked (and several friendly bases can be attacked at same time) the AI
player should first check the enemy attacker's strength and accordingly deploy
(if needed) additional forces from reserve (see section below).



d) Strategic reserve

#1
Whenever possible the strategic reserve of one or more divisions (i.e. several
regiments together with HQ and support units like AAA, engineers and coastal
guns) should be prepared.

The placement of strategic reserve should be at supply bases (best protected
and malaria free).


#2
Usage of strategic reserve would be dictated upon overall strategy posture and
current situation (offensive or defensive - attacked or attacking).



e) Staging area bases

#1
When AI decides to attack (and in big scenarios attack can be on several
targets) he should build forces in staging area bases.

Those bases should only be supply and/or primary bases (i.e. those which
already have sufficient development).


#2
Staging area bases should be additionally protected from enemy attack
(especially with extra air units and ASW ship assets).



f) Building and development strategy

#1
The AI player should develop bases (port/airbase/fortification) according to
their rank (starting with supply bases and so on).


#2
Certain bases (most likely supply/primary/secondary bases) would be chosen for
main air an naval power base(s).

Those bases should be given priority for building and development in order to
enhance their port/airbase facilities to maximum.


#3
Frontline (and recently won) bases should also be given priority (but less
than main air and naval bases).


#4
Since engineer units are the only units capable of building and development
(and there are never enough of them) they should be shuffled between bases
when their work is done (i.e. up to level of development decided for each base
- please note that such level should not always and in all cases be the
maximum development possible).



g) Logistics strategy

#1
The bases should be supplied according to rank.


#2
In order to better facilitate supply the best strategy would be that supply
bases directly supply primary bases and then primary bases support the
secondary ones. The secondary bases would then supply the tertiary ones and
all others.


#3
Specific means of supply should be decided depending on enemy threat and
appropriate delivery method would be chosen (normal "Transport TF", "Fast
Transport TF", barge "Transport TF", air transport etc.) for each base
independently.



h) Air power strategy

#1
According to overall strategy posture the AI player would decide how to deploy
air units (offensive or defensive or combined).


#2
The obvious choice for main air base(s) would be primary base(s).

In such bases the AI player would amass the air units needed for offensive
and/or defensive.


#3
Other bases would have air units assigned according to threat and need.


#4
When preparing attack the AI should amass additional air units in staging area
bases (if they are not already main air bases).


#5
In case of enemy attack air units can and should be transferred when and where
needed.


#6
AI player should assign air units to specific tasks (like "Naval Search",
"Recon" and "ASW") and deploy them to all appropriate bases.

In other words all supply/primary/secondary bases should receive the "ASW" air
unit(s) in order to lessen the effect of possible enemy submarine deployment
in those areas. According to base rank the number of units would vary (the
supply base being the most important one).

The "Naval Search" and "Recon" air units should be deployed to bases from
where AI player can observe the enemy activity (whether attacking or
defending).



i) Naval power strategy

#1
According to overall strategy posture the AI player would decide how to deploy
naval assets (offensive or defensive or combined).


#2
The obvious choice for main naval base(s) would be primary base(s).

In such bases the AI player would amass the naval assets needed for offensive
and/or defensive.


#3
When preparing attack the AI should amass additional naval ships in staging
area bases (if they are not already main naval bases).


#4
The "ASW" (antisubmarine) ships should be kept in all
secondary/primary/secondary bases in order to fight possible enemy submarine
activity. This is to protect friendly shipping (i.e. supply convoys).


#5
The "MSW" (mine sweep) ships should be kept in reserve and used only when
needed. Those shops are irreplaceable.

Supply base(s) and main naval base(s) are obvious candidates for keeping those
valuable ships.

Only when enemy mines are detected the "MSW" ships should be deployed.

The only exception would be invasion when the "MSW" ships should be integral
part.


#6
The "ML" (minelayer) ships should be used only in defensive mine laying
operations (i.e. laying mines in friendly HEXes).

Those are very valuable ships and they should not be lost in offensive mine
laying.




j) Various tactics

#1
AI player must learn to use its CV TFs with caution.

If there are operational enemy bases capable of housing land based bombers
(level, dive, torpedo) near-by the AI player's CVs should not close to them
within possible lethal distance.

Also the CV TFs should not operate near coastal HEXes due to UV v2.30 new rule
that halves the air ops there.


#2
AI player should rest and/or rotate squadrons in order to achieve the best
possible moral and fatigue values.

Only in dire need the AI should use the air units with fatigue and low moral.


#3
AI player must learn to give LRCAP air umbrella over endangered friendly units
(both naval and ground).

Special care must be made to provide the LRCAP over invasion TFs.

The supply convoy TFs should also be protected with LRCAP if there is chance
of enemy air attack.


#4
AI player should only use well protected transport TFs (with at least few
"ASW" capable ships).

Convoy system is practically enabled with this (and thus it gives the best
possible protection to transport ships).

This is a must (supply ships should never travel unescorted/alone)!


#5
AI player's invasion TFs should have "MSW" assets in main force and be
supported with "Surface Combat" TFs (following it) and under LRCAP (from land
and/or CV TFs).

This is also a must!


#7
AI player should use its submarines in both offensive and defensive.

In offensive they should cover enemy choke points and in defense they should
cover approaching paths towards endangered bases(s).


#8
AI player should use its submarines to lay mines.




What do you think gentleman?


Can Matrix/2By3 please comments on this?


Leo "Apollo11"
Image

Prior Preparation & Planning Prevents Pathetically Poor Performance!

A & B: WitW, WitE, WbtS, GGWaW, GGWaW2-AWD, HttR, CotA, BftB, CF
P: UV, WitP, WitP-AE
mdiehl
Posts: 3969
Joined: Sat Oct 21, 2000 8:00 am

Post by mdiehl »

It's a very interesting idea.
Show me a fellow who rejects statistical analysis a priori and I'll show you a fellow who has no knowledge of statistics.

Didn't we have this conversation already?
Snigbert
Posts: 765
Joined: Sun Jan 27, 2002 10:00 am
Location: Worcester, MA. USA

Post by Snigbert »

The AI is being reworked for WitP, but I have no idea what their intentions are on how they will design it.
"Money doesnt talk, it swears. Obscenities, who really cares?" -Bob Dylan

"Habit is the balast that chains a dog to it's vomit." -Samuel Becket

"He has weapons of mass destruction- the world's deadliest weapons- which pose a direct threat to the
TIMJOT
Posts: 1705
Joined: Mon Apr 30, 2001 8:00 am

Post by TIMJOT »

Some very interesting ideas, Im all for anyway to improve the AI. Priorities seem to be the key. A good AI has got to know when to hold them and when to fold them.:D

One thing though.

(Quote)
"This is a must (supply ships should never travel unescorted/alone)"

I believe the Japanese did use unescorted supply ships between SE Asia and Japan. Like wise the US between West Coast and Hawaii and I believe even between West Coast and Australia later in the war. Forceing the AI to escort even rear area convoys may cause critical shortages in front line areas.

Regards
User avatar
Fred98
Posts: 4019
Joined: Fri Jan 05, 2001 10:00 am
Location: Wollondilly, Sydney

Post by Fred98 »

I too prefer PBEM.

The best way to practise is to set up a game H2H and play against yourself.

Then you can ensure that unit "A" is in position to be attacked by unit "B" and test your theories.
Post Reply

Return to “War In The Pacific - Struggle Against Japan 1941 - 1945”