E080: Chebyshev bias: leader fraction vs. x.

Preview figure for E080

Tags: number-theory, quantitative-exploration, visualization, prime-races

Highlights

  • Estimates the fraction of x where one residue class leads another.

  • Plots a running bias estimate over log-sampled x values.

  • Detects coarse sign-change neighborhoods for the race difference.

What this experiment does

For the mod 4 race D(x)=π(x;4,3)-π(x;4,1), define the empirical leader fraction:

The implementation focuses on a compact, reproducible numerical workflow: deterministic parameter defaults, structured output folders, and one or more figures saved for the gallery.

Outputs

This experiment writes into out/e080/:

  • figures/fig_01_bias_fraction.png

How to run

make run EXP=e080

Notes

  • The gallery preview figure shipped with the documentation uses conservative cutoffs so builds stay fast. If you run the experiment locally, increase the cutoffs to see the asymptotic regime more clearly.

  • Prime-race plots depend on the chosen sampling of x (linear vs. log grid). The qualitative “who leads” picture can change when you zoom in.

Published run snapshot

If this experiment is included in the docs gallery, include the published snapshot (report + params).

  • x_max: 12000000

  • n_points: 4000

  • final fraction (D>0): 0.941

Figure:

  • fig_01_bias_fraction.png

Notes:

  • This is sample-grid dependent; it is a qualitative bias indicator, not a rigorous density.

params.json (snapshot)
{
  "n_points": 4000,
  "x_max": 12000000
}

References

Granville and Martin [2006], Rubinstein and Sarnak [1994]