An explorable map of the toy model
12,928 outcomes.
Every one inspectable.
Each column changes the fifth reading from 0 to 100. Each row starts from a different generated seed. Pick a cell to see the exact five readings and why the mean or median gets closer to the target of 50.
At fifth reading 50, the median wins more individual seeds while the mean has lower average error. Follow the colors, then inspect a counterexample.
101 readings × 128 seeds
The outcome atlas
Brighter cells have a larger error difference. White outline marks the selected cell. Scroll the map horizontally on narrow screens, or use the exact controls below it.
How to read this: 128 consecutive seeds (wrapping modulo 2³²) each produce four pseudorandom values in [42, 58). Every seed is tested against every integer fifth reading from 0 through 100. The colors compare absolute error against the chosen target 50. These 12,928 cells are a deterministic grid, not independent real observations or probabilities. A lower error in one cell, or more winning cells, does not establish a generally superior estimator.