IMG: How much time was spent planning MoOIII before actual coding began?
Rantz Hoseley: Well, MOO3 went through a lot of starts and stops... Originally it was started by Microprose, so we spent about 2-3 months trying to scope and scale the project. When MP was bought by Hasbro, the entire game went on 'hold' for about 2-3 months while Hasbro decided what to do with the properties and games they had just acquired. Then, we went through a similar phase when Hasbro was bought by Infogrames. So, there was a lot of 'blurring lines' as far as when any one item started, because elements were changing and being reworked both from a core code stance and from a scoping stance until IG took over.
IMG: Were any features cut from the game because of the scope or size of the project?
Rantz Hoseley: There were a lot of wish list items that we wanted to get in that got cut due to time and budget, like Space Monsters, and nomads and pirates attacking. Losing those really, in a word, sucked. (laughs) But, there were also a lot of elements that were cut, or seriously reworked in order to make the game less overwhelming. Not that it's a simple game now, but it was truly frightening in its complexity prior to us taking an 'editing pass' at the game, back in April of 2002.
Bill Fisher: It's my belief that a game isn't done when you've put in all the features; it's done when you've removed all of the features that you don't really need. This is a tough discipline, and nobody likes to do it. But you have to be willing to take an impartial look at every feature of the game and see if it really works. I think every one of us lost a pet feature or two in the process of winnowing down the design, and I'm glad that it happened. MOO3, by its very nature, invites you to just keep adding stuff. I think we did a great job, in the end, of facing up to some things that just didn't play as well as we'd hoped. What's left is still a very sophisticated, involving game, and is a great deal of fun to explore. And we have plenty of ideas that we can explore for future titles, if we want to.
IMG: The AI must process massive amounts of data each round. How did you go about scripting it, and how much more complex is it than the AI in MoOII?
Bill Fisher: We started work on the AI on the first day of the project. That was the only way to get it all done. I think we identified about 20 separate AI modules that needed to be written. It was huge. They ranged from strategic military AI to planetary viceroy AI to ground combat AI, to name just a few. We followed a "divide and conquer" method, separating them into functional units that were independent of one another, then scoping, writing and testing them individually. Fortunately, this sort of game lends itself to that approach. We had a large number of folks writing AI modules at various points in the game - as many as three at once, at times.
Another thing that saved us was that Greg Marsters (our lead programmer) and I had worked on Conquest of the New World together, so we already knew the pitfalls of trying to design an effective multiplayer turn-based game. As a result, Greg created from the first day a very clean messaging system that allowed us to coordinate the activities of all of the machines and manage both AI and human player actions. This let us make some late-game changes, such as shifting AIs between computers, that significantly improved performance without changing the play experience at all.
I can't really compare the AI to what was in MOO2, since we didn't write that. But I am certain that the AI in MOO3 is far more intricate. Just the code for scoping out borders and prioritizing planets to attack is worthy of an academic paper. In fact, one of our programmers is writing a paper about how he did it and publishing it in an AI book.