The Tactical AI in SC3 is quite good. It tactically deploys land and air units quite well and give you; whether the Axis or Allied, a solid challenge in the Eastern Front.
It is weak at the strategic layer when it comes to land areas not connected to Europe (UK and North Africa) and non-reactive. And not at all devious in the Diplomacy area.
Example: AI will never consider invading other ports in the UK, besides London. (the invasion sites are scripted)
Example: Allies forces in Egypt need reinforcements, but AI will not send any units there unless it is scripted.
Example: Axis really needs a few more air units in NA to drive back the British, but AI will not send additional units there.
Example: Axis AI will never focus on a Battle for the Atlantic. Its main focus is always on the Eastern front.
Note: I modded the standard campaign to add research scripts that give the AI a 20% (GV1(1,20)); 1 in 5 games, to make a major effort in researching advanced subs. If GV1(1,20) is triggered, woe to the Allied player that did not research anti-submarine warfare.
AI is not reactive.
Example: If Axis reaches Advanced Subs 4, Allied AI does not increase research in Anti-Submarine warfare.
Example: If Allied AI is advancing in Advanced Fighters, Axis AI does not respond.
Example: If Allied Human players send extra air units to Egypt, the Axis AI will never counter this.
Example: AI will never attempt to counter enemy diplomatic successes. See example below.
In the diplomacy area while an Axis player may attempt to reduce US or USSR war readiness, the Axis AI will never do so. Likewise, while a human Allied player may invest in diplomacy to increase US or USSR war readiness, the AI will never do so.
Tip: If you are considering an Axis Sea Lion use Diplomacy to reduce Soviet war readiness.
Having said that, for the combat in Continental Europe - the invasion of France and the war in the USSR - the AI is quite good and will give a human player a challenge.
I would like to see the designer look at improving the AI so that it can (1 in 5 games) launch a successful Sea Lion or make a stronger push in North Africa (1 in 5 games).
I'm interested in your comments. Are they based on several play throughs?
Thanks for your interest and I'd have to suggest that I'm not really sure how valid Edwin's post really is for new players as his point of view is from having had access to the Beta development version for the last few months. Essentially I doubt any game could hold up to high standards of scrutiny for up to five months without finding some faults, and especially so if you tinker down and open up the root system to see exactly how and why the AI does some of what it does. Which is of course possible since the AI scripting system is available for all to see.
This is not to say the AI is perfect, or that things could not be adjusted and improved, but at the same time from a developers point of view I'd have to disagree with most of the assessments above, and at least for now which I'll explain further below.
Also, don't get me wrong, it would be great to have 2-3 more years to develop a purely reactive AI system and also the time to simply work only on AI code, but the realities of development require us to spread out our efforts (for our small niche market developmet team) on the entire game as a whole. Despite this, and despite not having an AI level compete at the competitive level of Google's Alpha Go, we still feel very strongly that you'll get a solid game experience regardless of any tradoffs we may have had to make.
For example, while the AI doesn't specifically react to singular items such as a German player strongly focusing his naval effort on Advanced Subs, the Allied AI is aware that Advanced Subs are a likelyhood and a distinct possibility, and therefore focuses on Anti-Sub Warfare by default. Perhaps not rapid counter investments, but it will focus on it throughout the course of the war so on average you'll see pretty good levels, and then counter levels throughout the war.
This is similar to the idea that Advanced Tanks will play a large role in the game and therefore the AI understands that this is a key research for both sides and plays, purchases and invests accordingly, and again by default.
It all boils down to having the right amount of time and where to best spend your efforts. Critically we felt that land and naval combat are the most important, so a lot of time and effort spent there, and there are certain areas that will give players the most bang for their buck.
For example, the AI will send its fleets out into the Atlantic, i.e. you will see Axis subs, the raiding of convoy lines and Allied counter sub Destroyer and Cruiser efforts and this should give most players a feel for the Battle of the Atlantic and so on.
What you generally won't see are very high risk strategies taken by the AI as past experience has told us that players can be very critical of any AI mistakes and so while the Axis AI will engage in a Battle of the Atlantic to an important degree, it tends not to put its eggs all into a high risk basket and will therefore spread out its efforts towards very key areas of the game, i.e. it will most definitely focus on the Eastern Front as a poor eastern invasion by the Axis AI will never be well received by the player community.
Similar reasons for why an Axis Sealion was given a lower development priority and why our efforts were more focused on what players will tend to see in every game which would be a D-Day, or Torch or Husky landings and so on.
I hope this helps,