Basin Toy | Physics Lab

How it works: Each trial samples a random initial board, runs for N steps, and classifies the outcome (extinct, stable, oscillator, chaotic, growth). The table shows frequencies and probabilities across all trials.

Simulation Controls

Grid width Grid height Steps per trial Trials Density p Boundary Tail window (cycle detection) Oscillator max period Growth threshold (density) Activity epsilon RNG seed (optional)

Outcome Probabilities

Progress 0 / 0

Mean final density -

Mean tail activity -

Mean time to quiescence -

Probability bars update as trials complete.

Outcome	Count	Probability
extinct	0	0.000
stable	0	0.000
oscillator	0	0.000
chaotic_persistent	0	0.000
growth	0	0.000
long_cycle	0	0.000

Live Sample Run

Click “Preview run” to animate a single sample.

Step -

Live density -

Activity -

Sample outcome -

Quiescence Histogram

Distribution of step counts until quiescence (when activity hits zero).

Design document (reference)

Design document: Probabilistic “QM-style” prediction for Conway’s Game of Life
Purpose

Build a Python program that demonstrates this claim:

A deterministic system (Conway’s Game of Life) becomes effectively probabilistic for an embedded observer who does not know the exact microstate.
You can’t predict the exact end state after 1000 steps, but you can predict distributions over attractor outcomes and coarse observables.

This is prediction without description: we estimate outcomes statistically, the way quantum mechanics does operationally.

Scope
In scope

Fast Game of Life simulation on bounded grids.

Generate ensembles of initial states from controllable distributions (e.g., Bernoulli density).

Run many trials for N steps (e.g., 1000).

Classify outcomes into coarse “attractor-ish” categories.

Produce probability estimates, confidence intervals, and summary plots/CSVs.

Optional: “measurement” (coarse-grain / collapse) experiments.

Out of scope (for v1)

Infinite-plane Life or advanced pattern detection (guns, breeders) beyond simple heuristics.

Exact basin computation (infeasible at realistic sizes).

Formal derivation of Born rule analogues (this tool is empirical).

Conceptual model
World model

True state evolves deterministically:

St+1=F(St)

Observer model

Observer does not know the exact S0; they know a distribution P(S0) (e.g., i.i.d. density p).

What they can predict is a distribution over coarse outcomes:

P(Outcome|p,N,grid,boundary)

This is the “QM-style” move:

no trajectory prediction

only ensemble statistics