SC1 AI Improvements

[Edwin P.]

Desert Dave mentioned that HC is busy at work on the AI.

Here are some areas where I think the SC1 AI could be improved.

A. Utilization of HQ units.

The use of HQ units is very important in SC, yet the AI seems be oblivious to this fact when on the defensive.

In SC1 I would often see 4 or 5 HQ units clumped together in locations where they could not support front line units, after the Allies launched their offensive.

Example: In numerous games vs the AI I have seen 4 to 5 HQ units clumped around Athens in Greece or several German HQ units in Italy where only 1 or 2 units were in range of HQ support.

There would often be no HQ units positioned to assist German units defending Paris.

Simarily, in playing against the Russians I would often see HQ units evacuated to the Ural Mountains while none were located in Southern Russia to support the Soviet forces defending Stalingrad and Rostov.

---- Solution: Use a formula to prevent HQ units from being clumped in areas where they can't support the maximum number of units or (Max-1).

The AI constantly ignored the need to move or operate HQ units to safer areas when they were in danger of being cut off by advancing enemy lines. Many times I have encountered a lone HQ unit or two just southwest of Lenningrad on the south shore of the Baltic.

Futhermore, when the Allies invade France and are pushed back if there was a HQ unit in Brest it would never be transported back to the UK. It would wait in Brest to be destroyed.

-----Solution Idea: Routine to Evacuate HQ back to UK if Allied forces are outnumbered 3:1 in France and there are 3 or fewer Allied Units in France.

Simarily, the AI in later scenarios would not transport unneeded HQ units from Northern Africa back to Europe where they could aid in the conflict vs Russia.

---- Solution Idea: AI should decide on its strategy for northern Africa in the later scenarios. Either Hold the Line and evacuate un-needed units back to Europe or Offensive and reinforce the African units so Egypt can be conquered.

[ December 11, 2005, 03:53 PM: Message edited by: Edwin P. ]

[Edwin P.]

B. Engineers in SC2

In SC2 I would like to see the Axis AI have a choice of strategies for utilizing engineer units.

Example:

40% Fortify the Eastern Border of Poland, north and south of Warsaw to halt a counter attack by the Soviet forces.

40% Fortify the tiles along the Rhine to defend Germany from a future invasion by the western allies.

20% Use the Engineer unit to build a defensive line in Russia.

RE: UK AI Engineer Units

I would like to see tha UK AI know how to use Engineer units.

Perhaps it tracks the results of several games.

Game 01: Sea Lion, No Mid East Invasion

Game 02: No Sea Lion, Egypt Conquered by Axis

Game 03: No Sea Lion, Egypt Conquered by Axis

Game 04: No Sea Lion, No Mid East Invasion

Based on this data base it uses an Engineer unit to fortify its position in Egypt 50% of the time, England 25% of the time, and in 25% of the time does not build an engineer unit.

As the human player adjusts his strategy so will the UK AI adjust its decision to build and use an Engineer unit.

If the Human player's game history shows that he launches a Sea Lion Invasion 80% of the time then the UK AI will use an Engineer to fortify England 80% of the time.

[This would be a simple text file database that would track human actions and convert this to a percentage term: # Sea Lion Launched, # Egypt Conquered, #D-Day in Early 1942, #D-Day in late 1942, #D-Day in early 1943, etc.. By tracking when the Human player launches his D-Day invasion the AI can increase the chance of being prepared for it.]

[ December 11, 2005, 07:42 PM: Message edited by: Edwin P. ]

[Edwin P.]

C. Use of Air Fleets

Basically, the AI needs to focus air fleets on destroying units and not damaging resources, units not on the front line or empty cities that Axis units will occupy.

Many times I have seen the AI land units destroy the Beligan corps in Antwerp only to see AI air fleets bomb an empty city instead of adjacent land units that were damaged.

Many times I have sent corps to tiles behind the maginot line to draw fire away from the units in the Maginot Line.

The AI should also use Air Fleets to prevent the evacuation of Allied or Axis units. Bombing the port of Brest can prevent the Allies from evaucating units from a failed D-Day invasion or constantly replacing the unit holding this city with a new unit at full strength.

In my view the AI also needs to concentrate the majority of its air units on one front. All too often I would see 4 or Axis air units in France when the Axis AI declared war on Russia. Attacking Russia with 8 air units concentrated in one front would have been far more effective than keeping half the Axis air force in France while the USA was still neutral.

[ December 11, 2005, 04:12 PM: Message edited by: Edwin P. ]

[Edwin P.]

D. Reclaiming Research Chits

The AI should know when to reclaim research chits to finance the construction/reinforcement of combat units or when it has reached the maximum or optimum research level in a tech area.

Example: If the AI reaches Level 5 Anti-Tank Tech (the maximum level in this area) it should reclaim all of the chits in this area AND NOT Invest more in this area.

Example: If the AI knows that it does not want to build Lvl 4 Anti-Tank units (because they are too expensive) it should reclaim chits in this area when it reaches level 3 AND NOT Invest more in this area.

Example: If London has fallen, the UK AI should reclaim all research chits, except for perhaps AA, so that it can better defend Manchester.

How to implement this?

Each AI stategy could have a linked Optimum Tech Tree Chart. When a tech level reaches the optimum level in a tech area the AI reclaims all chits in this area and does not invest more in this area.

When the Western Allies launch their D-Day invasion in France the AI should consider reclaiming tech chits to reinforce damaged units IF doing so will likely give them victory in the West. This can be decided by looking at the Ratio of Units in France. If the Allies have the advantage then the AI should consider this. If the Axis outnumber them and the chance of victory is small then the AI should not do so, but consider a withdrawal of units back to the UK.

[ December 11, 2005, 07:40 PM: Message edited by: Edwin P. ]

[Edwin P.]

Middle East Allied Strategy

In Sc1 the Allies (and also the Axis) lacked a strategy for North Africa, Egypt and Iraq.

Perhaps in Sc2 the Allied Forces will select from a range of Strategies to employ in this theater:

01. Defend the Middle East

A strong defensive force is positioned to defend Egypt with reinforcements arriving from the Cape and Allied forces withdrawing via the Cape only if the cause is hopeless

02. Abandon the Middle East

Minimal forces are sent to Egypt to defend it against an Axis attack. Naval forces are withdrawn to the Atlantic.

03. Conquer North Africa & Iraq

A large task force is mobilized in Egypt to liberate North Africa from the Axis and conquer Iraq.

04. Liberate a Conquered Egypt

After the Axis conquers Egypt and redirects its attention elsewhere, a large Allied task force arrives via the Cape to liberate Egypt, Iraq and North Africa from the Axis.

Ideally in case 4 the AI would decide between liberating Egypt and Liberating Algeria (Operation Torch), perhaps landing a diversionary force at the site it does not intend to invade, thus forcing the Axis to waste MPPs operating units to meet a non-existent threat.

[ December 12, 2005, 03:51 AM: Message edited by: Edwin P. ]

[Edwin P.]

Italian Subs

In Sc1 the Italian sub would never attack Merchant shipping in the Med, even if there were no UK ships or air units in the region.

I would like to see this change in Sc2. If Allied ships approach it can always retreat to the Adriatic.

[Edwin P.]

Battle for the Atlantic

In the battle for the Atlantic I would like to see the German subs not only target convoy routes in the North Atlantic but also those in the South Atlantic.

It would be interesting if a sub on occassion positioned itself to hit transports heading to or from the Cape instead of always staying in the North Atlantic.

[Edwin P.]

Surprise

Perhaps the hardest task for the AI to learn will be how to surprise the enemy.

Example: If you move a ship in range of an AI airfleet it immediately attacks. Whereas a human will delay an attack for a turn or two in hopes of luring more targets into range and proceeds to quietly moves more air units into range for a surprise attack that will sink, not just damage the enemy unit.

Once in a game vs a human opponent I stationed 3 air fleets and a bomber in the Sinai. They held their fire as enemy transports approached the Syrian coast. When their HQ unit came in range my air units sunk the HQ unit and then damaged the transports carring the combat units.

[ December 14, 2005, 06:43 AM: Message edited by: Edwin P. ]

[santabear]

One from left field: The AI in SC1 isn't all THAT bad. He functions like a typical unimaginative British general...amass superior forces then advance methodically. He excels at methodical advances and retreats...if you leave a unit exposed (at "expert" level), it's DEAD.

In my opinion, the problem is not how the AI can surprise his enemy, it's how he can avoid being suckered by quick, human-like (i.e. unpredictable) decisions. Once you know how he thinks, it's too easy to sucker-punch AI. You can lure him into advancing in one direction and then counter attack. *poof* AI dies.

It IS fun to pretend to be Rommel vs. Auchinleck for a while, but it would be nice to have AI think like Patton on occasion...

[jon_j_rambo]

@Eddie --- Dude, are you really serious?

[J P Wagner]

This site might prove an interesting read for some of you, especially concerning some older games like Fields of Battle:

http://www.gameai.com/games.html

The game was in fact released but I never came across it..

[ December 29, 2005, 02:35 PM: Message edited by: J P Wagner ]