Vary the AI Strategy at Expert Level

[Edwin P.]

When I play the AI at Expert level it always selects the same countries to attack. First Vichy France > Spain > Portugual > Hungary.

It would be a big improvement if the AI did not always select the same strategy.

Example:

Beginner Level

---- 100% Strategy 1

Intermediate Level

---- 50% Strategy 1

---- 50% Strategy 2

Expert

---- 25% Strategy 1

---- 25% Strategy 2

---- 25% Strategy 3

---- 25% Strategy 4

[ May 05, 2005, 07:00 PM: Message edited by: Edwin P. ]

[Exel]

Why be so restrictive? Make the AI more dynamic, so it actually considers what would be the most preferable thing to do and not just select from a few pre-determined strategies and then stick to it no matter what.

[Edwin P.]

I agree, a dynamic would be preferable; however, it would also take more time to program and, based on SC1 the AI was very scripted at the Strategic Level. I would also like to point out that in going to war nations choose strategies to follow as a guide before they begin operations.

Futhermore, at the lowest level you want to keep it simple until the human player learns how to play the game.

Perhaps at the lowest AI level you would only see the Axis prepared to attack Russia when it readies for war.

Then at the next level the AI might attack Russia early in the game (25%) or later when it prepares for war (75%). I found that an early attack on Russia really changes the dynamics of the game in Sc1. Of course, the Attack Russia Early option would only occur if France fell quickly to the Axis powers.

Then at the next level and higher levels the AI would become much more dynamic and adjust its strategy based the opposing sides actions. Choosing between an Early Sea Lion, Conquest of the Middle East, Conquering the Nordic States, Battle for the Atlantic, Attacking Spain, Using Diplomacy and more.

It should also be noted that as HC said (and I am paraphrasing his comments) something to the effect of - its best for the AI do to a few things great than many things poorly.

[ May 07, 2005, 10:00 AM: Message edited by: Edwin P. ]

[Exel]

Well I agree that in the easier difficulty levels the AI could (and should) be simpler, and then progressively get more dynamic in its actions with higher difficulty settings. But I don't think the AI's options should be limited to only a few scripted strategies. Of course they could be used as a base for AI planning, but the AI should also be able to change its strategy according to the situation. And even the scripted base strategies shouldn't be so few and so transparent that the human player could predict what the AI is going to do.

[Edwin P.]

I agree that the scripted scripted strategies should not be few in number, but each one takes time to develop and bugtest, I believe that HC will have to focus on the most important ones or develop an AI routine that can develop its own viable strategies.

An important issue with an AI changing its strategy or operational deployments is tha you don't want an AI that constantly switches from one strategy to another every turn or every few turns.

Example: In SC1 I played a few games where the AI constantly operated its HQ between France and Italy and another game where units constantly operated between France and Russia.

Example: You would not want the AI to initially prepare for an Invasion of Russia and then switch to preparing for a Sealion in 1941. That would leave it exposed to an Russian attack and waste MPPs from operating units West to East and East to West and then West to East again.

Additionally, the Russia First Strategy is not just one strategy, it is the focal point of a set of supporting operational strategies.

Example:

Strategy: Defeat Russia First

Force Structure Options (that also affect research decisions):

Option 1: Balanced Air and Land Force

Option 2: Armor Heavy Force

Option 3: Infantry Heavy Force

Option 4: Air Heavy Force

Axis of Advance Options:

Option 1: Broad Front

Option 2: Stalingrad

Option 3: Moscow

Research Level Options

Option 1: High

Option 2: Medium

Option 2: None

The result of a Russia First Strategy could be:

Force Structure: Armor Heavy; Axis of Advance: Moscow; Research: Medium

or

Force Structure: Air Heavy; Axis of Advance: Broad Front; Research: Low

Ideally, you would want the AI to maintain a file containing the win/loss record for each strategy. There would be a greater chance for it to select a strategy that has a a record of winning. Thus as the human learns to counter one strategy the AI would likely switch to another strategy.

[ May 08, 2005, 09:56 AM: Message edited by: Edwin P. ]

[Retributar]

Applause to Edwin P. ...yes!!!, fabulous idea!...and yes!, thats the best way it could work!...i like it very much!.

[JerseyJohn]

Agreed. Great Ideas, Edwin, as always!