Each cell at row i, column j shows how often digit i was immediately followed by digit j in the generated sequence. This is called a lag-1 transition matrix — a core tool in statistical randomness testing.
In a truly random sequence, every pair (i→j) should occur with roughly equal frequency, producing a uniform, flat heatmap with no visible hot-spots. Biased or structured PRNGs reveal themselves as bright diagonal bands, corner clusters, or stripes. The infamous RANDU algorithm is a perfect example — its consecutive pairs collapse onto just 15 diagonal lines, visible immediately in this view.
Cell color intensity encodes frequency: bright/hot = frequent pair, dark/cool = rare pair. Hover over any cell to see the exact count and percentage. The diagonal (i→i) shows how often the same digit repeats consecutively — in a fair sequence, this should match all other cells.