Conway's Game of Life variant with local rules
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Demonstration of a Conway's Game of Life variant where the dynamics vary locally.

For an informal, blog-y treatment, see my post I want more Life.


The standard Game of Life plays out on a square grid where each cell is defined as adjacent to its 8-neighbors. Each cell has one of two possible states: alive or empty. The dynamics are synchronous and discrete. Living cells remain living if they have two or three living neighbors. Empty cells remain empty unless they have three living neighbors. Otherwise, the state flips.

Lif retains the synchronous discrete dynamics, grid topology, and binary state of standard Life. However, in Lif, each state has associated with it a stasis set, containing the number of living neighbors that will not cause it to flip state. (The dynamics of standard Life can be expressed like this with the stasis set for alive cells of {2,3} and that for empty cells of {0,1,2,4,5,6,7,8}.) These sets can vary individually for every cell. Biologically-inspired rules describe how the stasis set of a newly-settled living cell is derived from its possible parents. In the absence of parents, life may arise de novo. Other biologically-inspired rules describe how an existing alive cell's stasis set may be affected by those of its neighbors. Overcrowded alive cells may become empty. The stasis sets of empty cells also change over time, with a tendency to become increasingly habitable.

These processes are denoted settlement, birth, exchange, death, and gain of habitability (suggestions for a better name welcome) and are sketched out below. We can switch into thinking of the stasis set as a bit vector of length 9 when convenient.

  • Settlement or birth: If an empty cell has a number of living neighbors not in its stasis set, it becomes alive. If this number of neighbors is 0, it forms a new lineage having a stasis set with each bit on with independent probability alive_p. If the number of neighbors is not 0, one of the neighbors will become its parent with probability proportional to exp(-fit_cost * #(parent stasis set)). It will inherit this parent's stasis set with each bit flipped with independent probability mut_p.
  • Exchange: A living cell with living neighbors experiences this with probability exchange_r. A random neighbor is picked uniformly. If any of the neighbors is of the same lineage, the selection is only from among these conspecifics. The living cell takes as its new stasis set the consensus of its old stasis set and that of its neighbor, where disagreement is resolved with independent probability 0.5. This is followed by mutation, as in settlement.
  • Death: A new empty cell comes into existence with stasis set {0,1,2,3,4,5,6,7,8}.
  • Gain of habitability: An empty cell experiences this with probability goh_r. A chosen 1 in the stasis set is flipped to 0. Depending on the mode goh_m, the largest (max), the smallest (min), or a random element (random) is chosen.

In the absence of an infinite gird, I default to a toroidal (wrap-around) topology. This can be turned off using the toroidal parameter but the non-uniformity is annoying.


To install the program, navigate to a destination directory and then run the standard sequence of commands:

git clone
cd lif/
./ install

Once this is done, running the command in the console starts a visual demonstration of this model. All model parameters can be specified as command line arguments (see -h for details) or can be changed by editing the script wherever it has been installed.

The first two positional arguments set the grid width and height. If the terminal is too small to display the entire grid, only the top-left corner will be seen but the dynamics are not affected. Particularly for the max gain of habitability mode, a grid size of at least 80 × 80 is suggested to avoid triviality. Keep in mind, though, that the computation time for a generation scales roughly with the number of grid cells.


  • q: Quits program.
  • r: Restarts simulation.
  • space: Cycles through display modes for living cells.
    • "stasis": Length of cell's stasis set.
    • "parent": Cell's parent.
    • "max": Maximum value in cell's stasis set, "x" if stasis set is empty.
    • "min": Minimum value in cell's stasis set, "x" if stasis set is empty.
  • p: Toggle erasing alive sites after death. When turned off, the visualization effectively shows the territory held by each lineage.
  • left/right: Lower or raise exchange_prob.
  • down/up: Lower or raise fit_cost.
  • 1: Set gain of habitability mode to min.
  • 2: Set gain of habitability mode to random.
  • 3: Set gain of habitability mode to max.

Various statistics are computed and written to a file (lif_stats.csv by default) for later analysis.


  • A reasonably modern copy of Python (developed and confirmed to work in 2.7.9 and 3.4.3)
  • A terminal that can handle curses
  • Time that it will take for a horribly slow implementation to do its work