Programmed Opponent Learnings from Battle of Smoliani

I played out the Smoliani scenario using a programmed opponent for the Russian side. I ended up modifying the approach as the game progressed and I learnt what worked and what didn't. Here is a summary of those learnings.

The introduction of written orders, and the constraints they impose, was a key element of the programmed opponent implementation. The written command orders also gave a more realistic period 'feel' to the game, mitigating the player's 'god-like' overview of the battle.

The factors influencing what order options are available to each command increased substantially as the game progresses. It became clear that trying to set all these options up front wouldn't work very well.

The approach I eventually landed on was to limit command order changes to 1 per turn and randomly select from the 2 or 3 most likely options given the game situation. This appeared to work quite well, and gave enough uncertainty that I didn't feel like I was unneccessarily influencing the programmed opponent. In fact, on numerous occasions the programmed opponent chose an option that may have been viable but I wouldn't have chosen it myself. The outcome of that was actually better for the Russians and much worse for me!