Behind
the score.

What goes into a prediction

Instant analysis needs at least five valid digits from a healthy result feed for the current mode and period. The app examines colour frequency, short streaks, recent transitions, alternation and the presence of special digits. Pro details let you inspect the saved sample rather than a chart that silently changes later.

History analysis uses the digits entered for that analysis. Its source is different from a live-feed snapshot, so check the source label before comparing two insights.

How the score is formed

The engine turns those observations into colour weights. It selects a colour using weighted selection, with a repeatable seed for the Instant period and mode. It does not simply choose the most frequent colour every time.

The displayed score combines the selected colour’s weight, pattern strength, available context and sample size. It is bounded by the analysis mode. These are product rules, not a statistical model calibrated against future winning rates.

How Big/Small is selected

The size rule looks at the recent sample. For a strong concentration of Big or Small results, it can lean toward the other size; otherwise it uses the sample’s average digit. This is a heuristic choice, not proof that a reversal is due.

Colour and size are separate selections. See the colour Win/Loss example for a round where colour matches but size does not.

What happens after a loss

A user’s past losses do not change the Instant colour weights. The app does not learn a guaranteed recovery sequence from your account, and a losing streak does not make the opposite of the next prediction reliable.

Use Stats to review what actually happened. Feed-verified colour outcomes need a matched actual digit; pending predictions are not settled wins or losses. The current prediction history is stored locally, so a device’s history is not a complete account-wide performance record.

Read with context