I have a long-standing policy on not critiquing games I review until the review is published. I do this in deference to the publishers and am not changing. However, I’m playing Crown of Glory and World War II: The First Blitzkrieg for review and questions raised about AI on forums and news groups have made me think why such games may not play out historically. Here are some thoughts:
The AI, dumb as it is tactically, will never be as stupid as the Third Alliance in 1805 or the Allied High Command in 1940. Its side’s strengths and weaknesses will be known to it and will allow a more rational concentration of force and effort. We won’t see a General Mack squandering time and resource at Ulm nor will the units of four armies just sit around as in 1940.
Perhaps more importantly, the AI won’t be overawed when the player is Napoleon, Lee or the Wehrmacht. The AI has no psyche, no emotional baggage so it won’t be mesmerized by previous experiences or propaganda. Designers may attempt to imitate doctrinal fallacies but it won’t throw away what advantages or capabilities it might have.
Therefore, players should not expect a replay of history. Dumb as AIs are, they won’t be paralyzed like the Austrians or Gamelin. Players will win eventually, but they will do so only by doing something new with historical parameters, assuming the game has those parameters.
Given all this, what should gamers expect? When playing on defense, the AI should expect the human player to try the historical strategy first. Players have a mania about doing better than Rommel or Napoleon using their own approach. The AI should see this coming and prepare counter-offensives, concentrating in better positions, moving on flanks and so forth. When players get inventive, the AI should try one of two extremes: either a surprise attack early on or a precipitate retreat to gain time to evaluate the human’s intention. In the latter case, the AI should then decide which level to try for and then allocate resources accordingly.
AI offense is another matter entirely. Using a historical approach against a human with the most meager knowledge of the era is suicide. The player will see “A” and react in a different way than the historical defender then. The AI should think out of the box. “AIs don’t think” some exclaim. Why not? Half the key to winning in real life or games is psychological. Programmers should understand that nobody reacts well to surprises. AIs could either be programmed to make a very powerful attack somewhere to inspire “shock and awe”. Conversely, the AI can “ice” the human like a defending football team calling a timeout right before the field goal kicker signals for the snap. Wait a few turns before implementing the main plan, launch a few feints, reconnoiter a while. The human will get nervous, become tentative and question his dispositions or even try a premature counterattack.
These concepts would make games more suspenseful and exciting. I’m no programmer but I would be surprised if game developers couldn’t implement these ideas. Anything would be better than the present obvious AI options, giving the AI hidden advantages or--most unworthy of all--surrender writing good AI to PBEM or online play.