Estimate collision probability for a chosen UUIDv4 count or invert a target probability using the standard birthday approximation.
Compare identifier entropy, expected collision pairs, one-in-X risk and generation-rate horizons using stable birthday-problem math.
The calculator uses the 122 random bits left after UUIDv4 version/variant bits and evaluates large counts in the log domain to avoid NaN/Infinity overflow. Probability, expected pairs, one-in-X, target probability count, and the ~50% threshold are shown together.
UUIDs are 128-bit values. Version meaning is interpreted only inside the appropriate variant, and representation conversion never changes the underlying 16 bytes.
Collision math depends on the effective random-bit space, not just the printed UUID length. A UUID with deterministic fields or a weak generator can have less entropy than the nominal format suggests. Enter the number of independent random bits that actually vary, then use the birthday approximation only within that model.
A total-count estimate answers how many identifiers exist at once, while a generation-rate estimate answers how quickly that population grows. For services that create IDs continuously, convert the rate into the expected population over the retention window. This makes the probability easier to compare with an operational time horizon.
The calculator can produce extremely small numbers that look reassuring, but acceptable risk depends on the cost of a duplicate. A cache key, payment identifier, database primary key, and analytics event ID do not have the same consequence. Record the collision threshold your system is willing to tolerate before comparing alternatives.
Use the displayed entropy, population, and probability together rather than copying only the final percentage. For critical estimates, recompute with an independent formula or simulation at a smaller scale. Agreement across methods is a useful verification that exponent notation, units, or an order-of-magnitude input was not misread.
WebToolArc