Title Information Doesn't Add Aggression — It Removes Calibration Risk

The updated test results show that real information significantly reduces calibration risk in decision-making. Initially, the "safe" doctrine of marching cautiously yielded no wins out of 80 runs. However, when a tuned fixed clock was employed, 16 out of 80 runs were won. Most impressively, when full scripting with real perception was used, all 80 runs were won.

This indicates that having accurate knowledge is crucial for making correct decisions, rather than just guessing. The addition of fog of war, implemented as a projection layer, and a perception function that filters each side's knowledge at the end of every turn have been introduced. The engine remains omniscient, and combat, movement, and the decapitation mechanic remain unchanged.

The engine also has public capture zones, and routs are designed to be loud to provide valuable information. The implementation of conditional orders has shown a 17 out of 80 win rate, which is a significant improvement over the previous 16. This demonstrates that while a single conditional order can react to certain conditions, it cannot choose between different pursuit paths, implying that pre-commitment comes with a trade-off.

The current state of the project has 133 tests passing, with fog of war enabled by default. The roadmap includes adding attrition and supply lines, allowing for the exploration of the impact of prolonged battles on an army's effectiveness.