This README describes the basic concepts of Rezzo, including the definition of
the cellular automaton and basic ideas behind agents. The other READMEs are:

 * README.simple: Read this if you've never encountered cellular automata
   before, or otherwise want a more intuitive introduction to the underlying
   model behind Rezzo.

 * README.agents: Describes the protocol used by agents to communicate with the
   server. Read this if you want to write Rezzo bots.


Rezzo is a capture-the-flag game played by bots (agents) on a cellular
automaton based on Wireworld. The CA has the following states:

 * blank
 * conductor (wire)
 * electron
 * electron tail
 * photon
 * agent
 * flag
 * flag geyser
 * base

The agent, flag, flag geyser and base states are multiplied by the number of
players. That is, the state agent is really agent0, agent1, agent2, etc.

The game progresses with a CA state transition, followed by agent moves,
repeating forever. The CA state transitions are based on the 3x3 neighborhood
surrounding the given cell (eight neighbors). The transitions are:

 * blank, agent, flag geyser and base do not change
 * conductor:
   * if one or two neighbors are electrons, -> electron
   * else, -> conductor
 * electron:
   * if one or more neighbor is a flag or flag geyser and one or more neighbor
     is an electron tail, -> photon
   * else, -> electron tail
 * electron tail -> conductor
 * photon:
   * if one or more neighbor is a flag or flag geyser, and all neighboring
     flags or flag geysers have the same owner, -> flag with that owner
   * else, -> conductor
 * flag:
   * if one or more neighbor is a base, the owner of this flag loses (this is
     independent of the state transition)
   * if one or more neighbor is an electron, -> conductor
   * else, -> flag with the same owner

At the end of a CA transition in which a player loses, every cell owned by them
is transitioned into an unowned state as follows:

 * agent -> blank
 * flag -> conductor
 * flag geyser -> blank
 * base -> blank

Agents have a location (reflected in the CA) and a cardinal direction. Agents
are not told their absolute location or direction. During each agent's turn,
the agent is given a viewport (the 13x13 part of the world in front of them),
and allowed a short time period (typically 1/15th of a second) to make a move.
During an agent's turn, they may:

 * advance
 * turn left
 * turn right
 * build
 * hit

Advancing is legal when the cell in front of the agent is blank. It is replaced
with the agent, and the agent's cell is made blank.

Turning is always legal.

Building is legal in the same circumstances as advancing. The agent is
advanced, but the cell it was formerly in is replaced with a conductor, instead
of blank.

Hitting is legal when the cell in front of the agent is in the state conductor,
electron, electron tail or flag. Each cell of these states carries a damage
value, which persists with the cell even as its state changes. Each time the
cell is hit, its damage value increases. When the damage reaches four, it is
reset and the cell's state is updated to blank.