Behind the Curtain: Wrangling the AI
by Charles Watkins

The Artificial Intelligence routines of the Heroes game engine have been a major reason for the success of the series. Having gone through several major releases and who knows how many minor adjustments, the AI driving Heroes IV computer players represents the main opponent for solo play and a big part of the multi user game.

Most every experienced Heroes III player has noted some inferior qualities of the Heroes IV AI. In fact, this is probably the one factor that has led some veteran players to shun the game. (Lack of multi user support was another, but many players hung on long enough to see it finally arrive.) We’ve seen some improvements in the more recent versions, but in Heroes IV, the computer players just don’t seem to play as well as they did in Heroes III. To be sure, Heroes IV is a more complicated game than Heroes III, but even so, there are some obvious shortcomings.

I think there is also a familiarity factor involved. As I recall, in the early days of Heroes III we also complained a lot about the AI. It did not develop castles well and often made stupid mistakes in recruiting. I remember the struggle my friend Corribus had in getting a hero to go through a Lith. But eventually mapmakers learned to live with the problems and found ways to get the AI to behave.

I have no doubt that if they had been given more time to produce the game, the developers would have addressed most of these. But as it is, we have to live with the flaws—and by ‘we’ I mean those who have found that on balance the new features in Heroes IV make it worth playing, despite the weakness of the AI.

We are only now arriving at that point in Heroes IV. Mapmakers need to accept the strengths and weaknesses of the AI and take them into account in their designs. When we get to the point where mapmakers are building maps where the AI’s idiosyncrasies support the theme or frame conflicts, then we will have arrived at where Heroes III left off.

In this short series, I’d like to take a look at the Heroes IV AI and suggest some ways of making it play better. I call this wrangling. (After all, I’m a native Texan.)

In Part 1, you’ll hear about what I’ve observed about the Heroes IV and maybe account for some its problems. I should note that I have no knowledge of the game’s internals—only some observations and experiments, plus vague recollections of what others have said.

In Part 2, you’ll get my suggestions for ways to get the most out of the AI’s capabilities. This starts with the realization that the AI will never play like a good human opponent--and that it is only one tool available for creating a challenging game.