Real signals.
Clear outcomes.

Five calls. One Daily.

Choose one side in each of five matchups. You can play anonymously. Your predictions are saved in this browser when you lock all five.

The same starting line.

Hourly groups close predictions together. The next 24 hours form the outcome window. You see the exact closing and expected result times before submitting.

Growth, not just fame.

For count signals, we compare activity gained during the outcome window with activity gained during the preceding 24 hours. We add 20 to each activity count to soften tiny baselines, then combine source ratios with a weighted geometric mean.

The index starts at 100 for the window. If the combined activity ratio is 1.5, the ending index is 150: +50% momentum. Watchlist star counts are raw source observations, not this index.

Independent evidence.

A game trend needs at least two approved, independent source families. Matchups use the same source families and metric types. We freeze the source set and rules before predictions close.

Developer interest is not global or regional audience measurement. Geography stays unspecified unless the source genuinely measures it.

No signal, no invented winner.

Missing or stale observations get up to two hours of grace. If evidence remains insufficient, the pair is void. Counter corrections, frozen trends and abnormal changes can also void an outcome.

Boundary observations must be collected before the boundary and no more than 45 minutes old. We never turn unavailable data into zero or change source weights halfway through a round.

Your proof is earned.

The side with more momentum wins. Differences of one percentage point or less are ties. Tied and void pairs are excluded from your correct-call denominator.

We show actual results. We don’t invent users, percentile rankings or reputation scores. AI may help explain evidence later; it never decides outcomes.