The protocol aims for one block every ten minutes, and on average it hits the mark. But the average is a costume. The real gaps between blocks are exponentially distributed — the fingerprint of a process with no memory. A block that's twenty minutes late is no closer to arriving than one a second old. And the gap you happen to be waiting in is, on average, nearly twice the headline ten. Live from mempool.space.
Distribution of inter-block gaps · — blocks
CV = —
measured gaps
exponential fit — memoryless
the metronome — the FAULT
How this was made. The timestamps of the most recent
~1000 blocks,
pulled from
mempool.space, differenced to get the
gap between consecutive blocks. Miner-set timestamps drift slightly, so non-positive and >3 h gaps
are dropped before fitting. The
exponential fit uses the single parameter λ = 1/mean — a Poisson
process, which Bitcoin's difficulty adjustment is engineered to approximate. The
FAULT is the
metronome: the intuition that "every ten minutes" means a regular tick, which would pile every gap at
600 s with coefficient of variation 0. Truth-checks:
GREEN
the measured CV lands at ≈1.0, the exponential signature;
GREEN
survival past any waiting time stays flat at e⁻¹ ≈ 0.37 — memorylessness;
FAULT the metronome's CV=0 and its "you wait ≤10 min"
both collapse against the spread. The inspection-paradox gap is E[g²]/E[g], the length-biased mean.
Snapshot baked on wake; the chain keeps mining.