What splitting a transfer costs

1,000 USDT (TRC-20) into BTC

Measured 28 September 2026, and refreshed through the day. Costs are the shortfall against mid-market, so they are what the transfer gives up rather than a fee anybody itemises.

What the choice is worth

One exchange gives up 1.18% and scores 0 — a single edge from your wallet to your destination, which one company watched from end to end. 4 exchanges give up 2.16% and score 91, with no provider seeing more than its own leg.

The difference between those two numbers is the whole decision, and it is smaller than most people expect. Nobody else publishes it because pricing it requires asking every provider at every leg size, which is what this site does anyway.

Why more legs cost more

Providers charge a fixed amount as well as a percentage, so a quarter of a transfer costs more than a quarter of its price. How much more is a fact about the pair rather than about small transfers in general: measured on one day across the pairs here, thirty-odd dollars gave up 1.12% on Litecoin into Bitcoin and 23.48% on Ether into Tether.

A split also never reuses a provider — one seeing two legs can rejoin them — so the second leg takes the second-best rate and the third the third. That, not the leg size alone, is most of what the extra legs cost.

How linkability is scored

Out of 100, from the shape alone: how many legs, how many different providers, how unevenly the amounts are divided, and whether each pays into its own address. Spacing in time is reported where a plan has run rather than scored, because what a given spread is worth depends on how much other traffic shared the window — a fact about the chain, not about the plan. It reads nothing from the chain and makes no promise about anonymity — a provider that logs and correlates its own legs defeats the amount signal entirely.

Equal legs score zero of the twenty-two points for amount shape, because identical payouts arriving together reassemble into one transfer by inspection. Plans built here are deliberately unequal for that reason, and it turns out to cost nothing: the largest leg goes to the cheapest provider, which more than pays for the smallest leg taking a worse rate.

Method

Every provider is asked for the same pair at six sizes in one pass, and each leg is priced against the sample nearest its own size. That is an approximation worth naming: a leg between two samples is priced at whichever it is closer to, so a figure here is good to roughly the gap between them rather than to the basis point.

Mid-market comes from the median of several large exchanges' public ticker endpoints, never one venue's word. FixedFloat returns an authorisation error on this deployment and is absent from every count above, which is why seven providers quote rather than eight — and why eight legs cannot be run.