A Collection of AI Suggestions

[Edwin P.]

Source: JerseyJohn

In AI SC games, the U. S. makes no effort to cross the Atlantic at all if Russia is fighting and the Axis occupies Western Europe. I'd hope that will be changed in SC2; have AI make a choice, even if random, where the U. S. collects units and attemps a landing somewhere in Western Europe -- what I mean, is set it so the AI doesn't always make the effort at the same place; sometimes it would be the British Isles and other times it would be France. That would force the Axis to garrison it's Western European holdings while fighting on the Eastern Front.

[Edwin P.]

Source: EdwinP

In SC games the Allied UK AI will never make a serious effort to invade Norway or liberate Sweden; with land units supported by carriers, naval bombardment and UK based air units, even if it is defended by only 1 corps in Bergan. In SC2 I would like to see the UK AI consider an invasion of Norway if it is lightly defended, Germany is at war with Russia and the USA is neutral.

Ideally, the chance of the AI deciding to invade Norway would be influenced by the number of Axis units in Norway and Sweden. More units = reduced chance of Norwegian invasion, fewer units = greater chance of Norwegian invasion.

Example:

1 Axis Unit in Norway/Sweden = 70% UK Invasion

2 Axis Units in Norway/Sweden = 50% UK Invasion

7 Axis Units in Norway/Sweden = 0% UK Invasion

[ August 26, 2004, 08:40 AM: Message edited by: Edwin P. ]

[JerseyJohn]

Great Thread Edwin!

And one that I think Hubert will find particularly usefull.

[Edwin P.]

Source: Lars

Maybe a simple AI switch where the US would build a real navy instead of sailing to their doom would solve the problem.(OF US transports sailing into a waiting Axis Naval Units) Or have them transferred in from the Pacific at some point (sort of like the Siberians for the Russians). By 44 the naval war was pretty much over in the Pacific.

[Edwin P.]

Source: Edwin

Have a chance for the UK UK and French Navy to concentrate their forces in the Med so they are prepared to sink the Italian navy when Italy enters the war. After dealing serious damage to the Italian navy they most of the naval units return to the North Atlantic to protect the British Isles from a Sea Lion invasion. A few, depending on the surviving strength of the Italian navy may remain to guard the Eastern Med.

Vary the chance by AI Level; ie 0% at Beginner and lower level, as this strategy can really ruin the Axis players chances for supporting a campaign in the Middle East.

[ August 26, 2004, 07:40 PM: Message edited by: Edwin P. ]

[Edwin P.]

Source: Edwin

Italian Risk Free Naval Bombardment Strategy: Italian Naval Units will seek out and Bombard allied land units in Central and Eastern Med if No Allied Air or Naval Units are in the Region. In SC1 all they did was to stay at anchor and do nothing, no commerce raiding, no naval bombardment to build up experience.

[Edwin P.]

JerseyJohn, Thanks.

Any contributions to this thread by one and all would be greatly appreciated.

[Edwin P.]

Source: Edwin

AI Opportunistic Hit and Run Bombardment Strategy - AI Naval Units will bombard enemy land units where it perceives no chance of receiving damage from enemy air or naval units.

[Edwin P.]

AI Air Fleet Targeting Priority - In SC the AI air fleets would attack a port and ignore transports carring land units or HQ units that were in range. Air Fleets would attack empty city hexes just prior to their land units entering the empty hex. Air fleets would constantly attack a surrounded land unit in Brest and ignore the port, allowing the Human player to withdraw the injured unit and send in a new unit. Air fleets would not concentrate fire on one unit to destroy it.

The system needs an improved AI Air Attack Priority system.

Example: Targeting Priority For Transports at Sea

Priority 1: HQ Transports

Priority 2: Amphibious Transports

Priority 3: Armor Transports

Priority 4: Army Transports

Priority 5: Corps Tansports

Priority 6: Engineer Transports

HQ Allocation Rules - In SC you would often see HQs in areas where their benefits could not be used. 3 German HQs in Italy with no German Axis units nearby for exampe.

The rule I would use for governing the operation of HQ units is simple - Maximum of 1 HQ unit per 4 Units per Country. Example: 8 German Units in Russia = Max of 2 German HQ units in Russia, 3 German Units in France = Max of 1 German HQ in France.

Last Ditch Survival Strategy - A country should reclaim tech chits if their survival is threatened. For example: Russia has 4 tech chits in Anti-Tank. Germany has taken Moscow and is moving on Stalingrad - So reclaim these 4 chits for 500MPP and purchase 4 Corps.

What's Last Ditch?

For the UK - London has been taken.

For the US - Axis Troops are landing in North America.

For Italy - Allied Troops have take 2 Italian Cities or Germany has surrendered.

for Russia - One of your key cities has fallen

For Germany - Paris has fallen or Russian Troops have taken Bucharest or Russia has taken Warsaw or Italy has surrendered.

Now let's refine this:

For Italy

Allied Troops have taken 2 Italian Cities after 1941 or Germany has surrendered.

Why after 1941 - sometimes the Allies will launch a surprise invasion in the early part of the game and take Sicily and Bari. The date condition prevents Italy from losing its Tech chits due to this Italian gambit.

[ August 28, 2004, 08:39 PM: Message edited by: Edwin P. ]

[Retributar]

Sounds very-good Edwin P...i like what i am reading...keep it up!!!.

[Edwin P.]

Source: Edwin

Bomber Intelligence - The AI will strategically place a bomber or two outside the range of enemy air to spot enemy troops movements.

Example: UK AI may station a bomber in England or Egypt (!) to spot any Axis naval action.

Example: Italian AI may place a bomber in Sicily or Bari or Libya (!) to spot allied troop and naval movements.

[ August 29, 2004, 08:21 PM: Message edited by: Edwin P. ]

[Edwin P.]

Early Game Italian Naval Survival Strategy: If the Allies have moved their entire naval fleet to the Med to engage the Italian Fleet, the AI viewing descretion as the better part of valor, will temporarily move the Italian Naval forces to the safety of the Adriatic instead of engaging the UK fleet and suffering irreplaceable losses.

(AI Chance: Beginner 0%, Intermediate 50%, Expert 80%, Genius 90% )

Then, once Paris falls the AI will operate air units to Sicily, if the UK naval units are in the area, so that on the following turn the Italian Navy and German airforce can engage and sink the British Navay.

(AI Chance: Beginner 0%, Intermediate 50%, Expert 80%)

Summary: 0% at Beginner, 25% Both Strategies at Intermediate, and 64% Both strategies at Expert.

This means that at Intermediate Level a human Allied player that attempts the Sink the Italian strategy will face a varied AI response, not the predicatable response seen in SC1.

Mass the Fleet - The Italian Naval AI will position its fleet so that the entire fleet can strike one massive blow in one turn, not several weak blows over two turns. It will use a low cost Corps Transport as a spotter that will travel ahead of the fleet to spot enemy naval units and keep the fleet from being surprised.

AI Early Game Sea Lion:

25% if Allies try the Sink the Italian Fleet Strategy and Fewer than X units are in the UK.

[ August 30, 2004, 10:13 AM: Message edited by: Edwin P. ]

[Edwin P.]

Source: Edwin

AI Sealion

An AI Sealion is perhaps the hardest AI routine to program as its chances of success, even for a human player can be doubtful.

So, lets give a few conditions for an Axis Sea Lion:

1. 3 or fewer land combat units in England and less than 300MPP in the UK Bank.

2. Most of UK Navy in Med or has been sunk.

3. France and Denmark have surrendered.

Even so the AI may choose not to launch a SeaLion if it is following another strategy.

Even so, lets make it easy for the Axis AI to launch a Sea Lion:

IF No UK units in London and No UK Naval Units within 1 turn of the the Channel

Axis Amphibious Units will embark from Belgium and land to take London in 1 Turn. Supporting Axis Unit will transport and land besides London. Axis will move additional Land units to Belgium Port to reinforce Landing on the following turn.

Axis 1st Turn Denmark Attack - Most experienced players attack Denmark on Turn 1 yet the AI in SC never does. I would like to see the AI able to execute this strategy to 1) give its naval units combat experience and 2) perhaps take Denmark early in the game for more plunder and aid in the conquest of France.

[ August 30, 2004, 11:55 AM: Message edited by: Edwin P. ]

[Retributar]

Why-not!...'Spys' would be able to transmit information on the obvious.

If their were too few units in England or Britain...surely it would be easy for anyone to see that fact!.

The 'Spys' could then let the German High Command know that resistance in England would be low enough or limited enough to permit an invasion...even though they might not know the exact compostion of those forces that are available.

Likewise for the Allies their 'Spys' could let them know if their were too-few German units guarding the French coast!...or, other places!.

What constitutes too-few???...i like 'Edwin P.'s' suggestion of 3 or fewer 'GROUND-Units'.

This concept could help the AI player greatly in this game as well as give the Human player a chance to take advantage of such a situation.

[Edwin P.]

Retributor, I agree.

The AI should know if there are few or no enemy units in a zone, but not their location or composition. Thus the human could still surprise the AI, and the AI could better plan its actions. this is especially true in the Med when Axis agents (in Spain, Spanish Morrocco, Vichy France, Egypt) notified their superiors of any ships entering or leaving the Med via Gibraltar or the Suez Canal.

For example: I would allow the AI to know the number of enemy naval units in the Meditterrean and the Atlantic (given that once you know the number in the Mediterrean you can then deduce the number remaining in the Atlantic) Then the AI could decide whether to maintain an aggressive posture in either area or a defensive one.

Simply put:

AI Italian Naval Forces in Med have 2:1 advantage over allied forces = Aggressive Posture (ie 5 ships vs 2 ships = hunt them down)

AI Italian Naval Forces in Med have 1:2 disadvantage over Allied Forces = Defensive Posture (ie 5 ships vs 10 ships = withdraw to the Adriatic).

Now the AI simply stays put and does not search for enemy naval units or engages the first enemy naval unit it sees, even if it is outnumbered.

----------------------------------------------

Personally, I would like to see the AI change its strategy depending on the current situation and the global strategy of the Axis player.

Italian AI Naval Strategy 1 - Hide in the Adriatic

Italiaqn AI Naval Strategy 2 - Search for the Enemy Fleet

Italian AI Naval Strategy 3 - Spot the Enemy Fleet and Then Sink It

Italian AI Naval Strategy 4 - Take Malta

This rountine gives the AI a strategy with which to take Malta. Basically, all Italian Surface ships bombard the Island then then when their experience is up to about 3 the Axis sends in Air units to support them and finish off the defender.

When should the AI adopt this strategy - when it has Naval superiority in the Med.

Italian AI Naval Strategy 5 - Breakout to the Atlantic

Italian AI Naval Strategy 6 - Defend the Western Med. In this strategy the AI attempts to surprise any allied ships that enter the Western Med. Most of the Italian Fleet, waits off the Spanish Coast, out of range of any aircraft in Gibraltar, ready to strike. A single transport or Sub sits off the Vichy Coast to alert the fleet to any UK ships entering the Med. A bomber sits in Vichy Algeria.

Italian AI Naval Strategy 7 - Defend the Central Med

When should the AI use each strategy?

1. No Allied Naval Ships in the Med

- Defend the Western Atlantic

- Take Malta

- Breakout to the Atlantic

- Conquer Vichy France

- Conquer the Middle East

- Defend Italy

2. Allied Naval Superiority in the Med (2:1)

- Retreat to the Adriatic

- Defend Italy

[ August 31, 2004, 11:44 AM: Message edited by: Edwin P. ]

[Edwin P.]

Source: Condor

I got Poland Breakthru, LC gambit, Norway gambit, Rome invasion, MED ship mayhem, Spanish gambit, Ireland training, France blitz, Scandinavian blitz, Barbarrossa 1st turn mayhem, Sweden gambit, Portugal gambit, Denmark defense, Iraq takeover, so those ..... that beat the AI would find themselves, when playing some veterans, saying: oh man, game over man, game over!

Unless the SC2's much improved AI incorporates a library of strategies such as these.

[ August 31, 2004, 12:32 PM: Message edited by: Edwin P. ]

[Liam]

For AI Aggression, you're talking about adding a bunch of strategies into it's existing AI Membrane You'll always find a flaw in AI strategy and how to exploit it. Making the AI vicious is good! All the things suggested are going to make the AI harder to beat but still not a human challenge. Just a bit of time to get around it. I think HC programmed the original AI for an education to PBEM and IP rather than a stand alone. I think SC2 AI will be similar. Even handicapped to death, I think most players will beat the AI. Unless it's sneaky enough to get to your Capitols and not in a predictable fashion.

[Edwin P.]

Source: Edwin

In the Med I think that the Allies have 6 basic Med Strategies for the opening game:

1. Withdraw the forces to the Atlantic

2. Mass Allied Naval forces in the Med in an attempt to sink the Italian Navy when it enters the war and withdraw them to the Atlantic after 2 or 3 rounds of combat.

3. Maintain a small force in the Eastern med to defend Egypt and the Middle East. (SC1 AI's Strategy)

4. Italian Invasion Gambit.

5. Tobruk Gambit (where the Allies take Tobruk to foil an Axis invasion of Egypt)

6. Reinforce the Egyptian fleet and support it with Airpower in Egypt.

Of these 4, strategies 1, 2, and 3 should be relatively easy to incorporated into the software and strategy 2 would make for a much more exciting game, if it is not used all the time.

Strategy 1: Simple, move all naval units to the Atlantic.

Strategy 2: Simple, move all French Naval units to the Central Med and most UK Naval Units to the Central Med

Strategy 3: Just needs to be fine tuned - ie perhaps move only Carrier to the Atlantic and hide other ships in the Red Sea.

Strategy 4: Perhaps the hardest for the AI to execute well.

Strategy 5: Moderate, taking Tobruk in North Africa, would make any Axis conquest of the Middle East more difficult but is harder for the AI to execute well than routines 1,2, or 3.

Strategy 6: Simple, send an extra 1 to 3 warships to the Eastern Med to aid in the defense of Egypt and operate a bomber to the hex east of Cairo so that approaching Axis transports can be spotted for Naval interception.

Novice AI - Access to Strategy 3

Beginner AI - Access to Strategies 1,3

Intermediate AI - Acces to Strategies 1,2,3,6

Expert AI - Access to All Strategies 1,2,3,4,5,6

[ September 02, 2004, 09:55 AM: Message edited by: Edwin P. ]

[Retributar]

What i would like to have available is about a dozen individual complete sets of AI Strategies...that could be randomly loaded into the AI...to help avoid 'Predictablity'.

Also, depending on the current winds of change...the AI should be able to change or shift strategies according to the new situations as they arise.

This would help to keep the AI as a fresh and unscrutable opponent...and as well to keep the Human player guessing most of the time...as it was in reality.

This should help prevent the Human player from finding work-arounds for the AI most of the time hopefully.

Edwin P. can you make enough sets and variations of AI strategies 'to-make-it-so?'.

[ September 01, 2004, 01:03 PM: Message edited by: Retributar ]

[Edwin P.]

Excellent points Retributor and I agree with them all.

The key issue with the SC AI, as you pointed out is its "predictability". It will never invade Egypt, never send the Italian Navy to Cairo, never wage a war in the Atlantic, never take the Nordic countries, never mass its entire air fleet on one front, etc. This gives the the human player an unbeatable edge.

As for variations of AI strategies, the inclusion of them in the game is a design choice that only HC can make. All that I am attempting to do here is to give HC food for thought and perhaps, just perhaps, among them is an idea that he will find worthy of including in his game.

[ September 02, 2004, 02:55 PM: Message edited by: Edwin P. ]

[SeaMonkey]

I would like to add one thing here. As Liam refered to, the AI strategies will always be recognized as a player accumulates game after game against it. To offset this recognition factor of experience I would like to see the AI feint or distract the human player with moves that mirror the other strategies that it has to choose from. So now the AI randomly selects a strategy, but also randomly chooses a diversion strategy for the "feint".

[Edwin P.]

SeaMonkey, Excellent Idea.

[pzgndr]

The key issue with the SC AI, as you pointed out is its "predictability".

Bingo. So how to make SC2 unpredictable? I have my doubts about AI capabilities. I believe it can be made to be a competent and challenging opponent, but unlikely to ever become brilliant or nefarious with sneaky feints. SC2 will probably have a decent generic AI, and the AI scripts should allow it to be customized somewhat. Perhaps with some randomness. Like everyone here, I am also curious to see what magic Hubert weaves into the new! SC2 AI.

Beyond that, the new event scripts WILL ensure more opportunities for unpredictability. Some players will prefer a historical game will minimal randomness, some will prefer an all out what-if game of pure surprises, and others will prefer something in between. Change the game situation, and the AI (or human) opponent must react accordingly. Game difficulty settings may affect scripted randomness probabilities, and/or these can always be edited for custom games.

My point is that between the event scripts and a decent AI with some AI scripting, we should be able to get the unpredictability we seek. Definitely NOT the same as playing another experienced player, but certainly an interesting and enjoyable game nonetheless.