---
title: "red-blue-buttons"
summary: "Simulation, data, and figures behind https://aesv.io/feed/which-button"
external_url: "https://github.com/4esv/red-blue-buttons"
links:
  - "https://github.com/4esv/red-blue-buttons"
  - "https://www.kaggle.com/code/axelst/which-button"
  - "https://www.kaggle.com/datasets/axelst/which-button-world-trust"
  - "https://raw.githubusercontent.com/4esv/red-blue-buttons/main/figures/death-curve.svg"
  - "https://raw.githubusercontent.com/4esv/red-blue-buttons/main/figures/toll-bars.svg"
---

- Stars: 0
- Language: Jupyter Notebook
- Updated: 2026-10-09
- Topics: data-visualization, game-theory, notebook, simulation

[GitHub](https://github.com/4esv/red-blue-buttons)

Simulation behind the essay **[Which button?](https://www.kaggle.com/code/axelst/which-button)**.

Everyone privately presses red or blue. If more than half of everyone presses blue, everyone lives; otherwise only the red-pressers live.

## Kaggle

The essay + model live on Kaggle as a runnable notebook and dataset:

- notebook: [kaggle.com/code/axelst/which-button](https://www.kaggle.com/code/axelst/which-button)
- dataset: [kaggle.com/datasets/axelst/which-button-world-trust](https://www.kaggle.com/datasets/axelst/which-button-world-trust)

`python -m scripts.export_kaggle` regenerates `kaggle/dataset/`; push with `kaggle datasets version -p kaggle/dataset -m "<msg>"` and `kaggle kernels push -p kaggle/notebook`.

## Run

```bash
pip install -r requirements.txt
python -m scripts.model              # results -> data/canonical_results.json
python -m scripts.model --confirm    # ~7e9-agent confirmation (slow)
python -m scripts.plot_death_curve   # -> figures/death-curve.svg
python -m scripts.plot_toll_bars     # -> figures/toll-bars.svg
python -m unittest tests.test_model
```

## Results

| metric | value |
|---|---|
| world death rate | **5.1%** |
| does blue ever win | **no**, in no country or the world |
| highest | Denmark, **42%** |
| lowest | Colombia, **~1,700** people |
| China share of world toll | **89%** |
| world death rate without China | **0.7%** (below the US) |
| confirmed at | **6,991,900,000** agents (matches the closed form to 0.0004 pp) |

Rule: press blue iff `trust × (1 + altruism) ≥ 1`. Altruism `a = 0.28`, Beta concentration 7, no misclick. Global: blue wins iff `>50%` of everyone presses blue.

## Figures

![Death rate vs trust](https://raw.githubusercontent.com/4esv/red-blue-buttons/main/figures/death-curve.svg)

![Per-country death toll](https://raw.githubusercontent.com/4esv/red-blue-buttons/main/figures/toll-bars.svg)

## Data

`data/countries.json`: trust (WVS Wave 7 / EVS 2017), altruism (Engel 2011), population (UN 2022); sources in `_meta`. China's 64% trust is contested, and the world figure leans on that number.