Waiting buys nothing, until it buys everything

Ten minutes is the design target, not an observation. Measured against the live chain this morning:

Close to target, and close to the 10.84 measured the previous evening. Everything below uses the sixty-block reading, because ten blocks is far too few to estimate an average that varies this much — a one-hour sample taken at the same time read 5.2 minutes, which is noise rather than news.

Block arrivals are near enough a Poisson process that a gap of length t contains at least one block with probability 1 − e^{−t/interval}. That is not a curve anyone chose. It follows from blocks arriving independently at a known average rate:

A two-minute pause separates your transfers about one time in six. Five times in six you waited and got the same result as not waiting. That is the part the advice leaves out.

While measuring the above, a fifteen-minute window of the chain was queried and came back empty. Zero transactions, zero blocks.

Bitcoin was not idle for fifteen minutes. No block had closed inside that window, so there was nothing confirmed to count. At a 10.36 minute interval that happens 23.5% of the time — and it is the same event as two transfers sent fifteen minutes apart landing in the same block anyway.

Roughly one wait in four buys nothing, no matter how carefully you timed it. The odds are good at fifteen minutes. They are not a guarantee, and nothing you do makes them one.

The obvious conclusion is to wait longer. The arithmetic disagrees, because the curve flattens. Here is what a band of random gaps buys across a five-leg split, and what it costs:

Twenty to forty minutes separates almost everything and takes two hours to do it, earning 0.031 blocks per minute of waiting. Four to eight earns more than twice that rate. Every additional minute is worth less than the one before it, because the probability of a block having already arrived keeps rising.

So the question is not “how separate can I get” but “how much am I willing to pay for the last ten percent”. On this site the answer is four to eight minutes between legs, which is why.

The cost is not only patience. A non-custodial transfer is signed in the browser, so the key is in the tab for as long as the plan runs. Every extra minute of spacing is another minute of a private key held in memory with a page open — a different risk, traded against a smaller one.

An eight-leg plan at four to eight minutes waits 42 minutes and reaches about four distinct blocks. The same plan at twenty to forty minutes waits three and a half hours. A threat model that ignores the second number is not a threat model.

It is not a privacy score, and this site does not turn it into one. Spacing is reported where it can be measured and never converted into points, because what a gap is worth depends on how much other traffic shared the window — a fact about the chain rather than about your transfer, unmeasurable for most chains and unmeasurable by anyone for Monero, whose design exists to prevent that enumeration.

Nor does separating two transfers into different blocks make them unlinkable. It removes one signal, the one that says these happened at the same instant. Amounts, destinations and the providers in between are all still there, which is what the cost of splitting is actually about.

A regular interval is its own signature, too. Transfers spaced exactly eleven minutes apart reassemble as easily as transfers of exactly equal size — which is why the gap is drawn at random from a band rather than fixed, the same reasoning behind splitting unequally.

Block counts and spans were read from the live chain on 1 October 2026 at 05:45 UTC: 32 blocks across 337 minutes over six hours, 60 blocks across 622 minutes over twelve. The probabilities are computed from the 10.36 minute figure, not observed directly — the model is the standard one for Poisson arrivals, and it is the model rather than a measurement of your particular transfer.

Intervals drift. Difficulty adjusts every two weeks to pull the average back toward ten minutes, so these numbers are a reading of one morning rather than a constant. The shape of the argument does not depend on the exact figure: whatever the interval is, gaps much shorter than it buy very little, and gaps much longer than it buy progressively less per minute.