Metric vs metric

Do two metrics move together? These views plot one metric against another over a uniform random sample of ~50,000 seeds (the full 2.1B joint distribution can’t be shipped to a browser, so we sample). The sample is exact per seed — it’s the coverage that’s sampled, not the values.

Most metric pairs are independent (each is its own set of PRNG draws), so expect a shapeless blob. A pair that shows a tilt or curve shares an underlying cause — that’s the find.

Scatterplot matrix

A brushable SPLOM over a handful of metrics: every cell is one pair; drag to brush a region and the selection highlights across all cells. Choose the metrics to include.

Data: a uniform random sample of seeds (target ) over the full archive. Correlations and shapes are sample estimates; independence shows as a round cloud. axisSum.* sums all three parallel worlds; axis.{-1,0,1}.* are single-world (0 = Main).