Splitting unequally costs nothing

A thousand USDT into Bitcoin, four exchanges, four different providers, priced in one pass. The only difference between the two rows is how the amount was divided:

Twenty points better on linkability, and 0.097 points cheaper. Not a trade — both directions at once, from the same four exchanges.

A split never sends two legs through the same provider, because a provider handling two legs can rejoin them. So the legs are handed out best-rate-first: the cheapest provider takes the first, the second-cheapest the second, and so on down.

Equal legs waste that ordering. The cheapest provider is given exactly the same amount as the most expensive one. Descending legs line the two orderings up — the best rate carries the most money — and that gain is larger than what the smallest leg loses:

The 112.5 leg does pay 2.41%, the worst rate in the table. It is also the smallest amount, so it carries the least weight — while the 0.87% rate applies to 387.5 rather than 250. That is the whole mechanism.

The first measurement gave 1.550% equal against 1.446% unequal. A day later, on quotes taken fresh, it gave 1.509% against 1.413%. The absolute numbers moved — they always do — and the gap did not: unequal came out cheaper by about a tenth of a point both times.

That second pass matters more than it sounds. Another figure measured the same day did not survive: the cost of a thirty-dollar Litecoin-to-Bitcoin transfer read 1.12% one day and 8.90% the next, with two fewer providers quoting it. A single day's reading of a live market is an anecdote. This one was worth publishing because it repeated.

Unequal amounts remove one signal of several. They do nothing about a provider that logs its own legs and correlates them, they do nothing if the legs fire back to back, and they do nothing if every leg pays into the same address. Timing and separate destinations are scored separately for that reason, and neither is free in the way this is.

It is also not anonymity, and nothing here claims to be. Providers see the amounts and addresses they handle, the chains are public, and this deployment records what the privacy page says it records.

Every provider was asked for the same pair at each leg size in a single pass, so the quotes are comparable to one another. Mid-market is the median of four large exchanges' public ticker endpoints, read in the same session. Cost is the shortfall against that benchmark, weighted by leg size — a flat average would let a small expensive leg count as heavily as the large cheap one it was split from.

The linkability scores come from the same heuristics the app applies to any plan: leg count, how many distinct providers, how uneven the amounts are, timing, and whether each leg pays into its own address. Equal legs score zero of the twenty points for amount shape. FixedFloat returns an authorisation error on this deployment and is absent throughout, which is why seven providers quote rather than eight.

You can see the same comparison drawn across every split size on the splitting page, and why small swaps cost more covers the size effect the smallest leg runs into.