Best Swap Amount: Where Fixed Fees Stop Eating Orders

Two quotes, same pair, four minutes apart. At 14:02 UTC on 25 October 2026 I asked our own quote endpoint for USDT-TRC20 into XMR at sixteen different sizes. The smallest legal order, 19.81 USDT, paid out 0.01443418 XMR. The 15,000 USDT quote paid out 24.372964 XMR. Run both through the rate the large quote implies, 615.42 USDT per XMR, and the small order kept 44.8% of what its own money should buy. The large one kept 100%. The price had not changed. What changed was the size band, and its edges are cliffs, not slopes. (XMR to USDT exit)

How to see the band yourself. For a fixed pair a quote behaves almost exactly like `out = (amount − F) × m`, where `m` is the marginal rate at large size and `F` a flat deduction per order. Two quotes an order of magnitude apart give you `m`; any third quote gives you `F`. Here `m` came out at 0.0016249023 XMR per USDT, that is 615.42 USDT per XMR, from the 5,000 and 15,000 quotes. Every figure below is `F = amount − out ÷ m` with retention `= out ÷ (amount × m)`, all measured between 14:02 and 14:06 UTC so that drift stays smaller than the differences being discussed. (XMR to USDT exit)

The bottom edge: a flat 11 USDT. The minimum for this corridor is 19.805975 USDT, and it is not a round number because it is derived from the payout side rather than set by hand: by 14:03 the same endpoint reported 19.809883 as the price moved. An order sitting exactly on that floor loses 10.93 USDT, which is 55.2% of the order. `F` then stays remarkably constant as you scale up: 10.93 at 19.81, 10.96 at 30, 11.02 at 50, 11.18 at 100, 11.33 at 120. Retention climbs 44.8% then 63.5%, 77.9%, 88.8%, 90.6%. This is one fixed charge spread over more money, and it is the largest single cost in a small swap. (XMR to USDT exit)

The step at 130 USDT. At 120 USDT the endpoint returned one route and no advantage. At 130 it returned two routes and a 929 bps advantage, and `F` collapsed from 11.33 USDT to 0.344. Effective cost fell from 679.57 USDT per XMR to 617.06. Going from 100 to 130 USDT is a 30% increase in size that buys an 11% better rate; adding 30% anywhere else on this curve buys nearly nothing. If you were about to send 100, send 130 instead. It is the cheapest upgrade on the page. (XMR to USDT exit)

The band: 130 to 15,000 USDT. Across that range the quote is effectively flat. Deduction was 0.344 USDT at 130, 0.351 at 1,000 and 0.351 at 15,000; effective cost wandered between 617.06 and 615.44 USDT per XMR, a 0.26% spread that is mostly rounding on the payout side. This is the band where an order pays for the route and nothing else. The pool that entered at 130 is still filling every size you can send, so there is no liquidity penalty to pay yet, and the fixed charge has stopped being a percentage worth thinking about below roughly 500 USDT. (XMR to USDT exit)

The top edge: a 0.51% step at 15,100. One hundred USDT above the ceiling the second route disappears: 15,100 returned one route, zero advantage and 24.41171341 XMR, which is 618.56 USDT per XMR against 615.44 at 15,000. From there the curve is flat again: 50,000 gave 617.98, 175,000 gave 618.69, and the corridor maximum was 175,594.45. Worth stating plainly, because most people assume large orders bleed on depth. Here the penalty from 15,100 to 175,000, eleven times the size, is 0.02%. The 0.51% was paid once, at the crossing, because one pool stopped being able to serve the order. (XMR to USDT exit)

When splitting actually pays. Above the ceiling, splitting becomes arithmetic. One 25,000 USDT order paid 40.45379864 XMR. Two 12,500 USDT orders, both safely under the ceiling, paid 20.336526 each, or 40.673052 together, which is 0.54% more. Part of that gap is drift between the two calls, so treat the structural gain as about 0.42%, the distance between the plateau rate and the single-route rate. Then subtract one extra deposit, 6.4285 TRX measured on a USDT transfer on 24 October, roughly 2.12 dollars at TRX 0.32901, plus a second 30-minute window. Net gain near 100 dollars on 25,000, and nothing at all below the ceiling: two 7,500 orders land on the same rate as one 15,000 order. (XMR to USDT exit)

Do not generalise this to other corridors. The flat deduction belongs to the route serving your pair, not to the site. BTC into XMR today had a minimum of 0.00018113 BTC, which is 14.53 dollars at 80,288, and a deduction of 0.0000751 BTC, about 6.03 dollars, so a 0.0002 BTC order kept 62.5% while a 0.0016 BTC order kept 99.7% once the second route appeared. XMR into USDT-TRC20 has almost no floor at all: 0.01 XMR, about 6.07 dollars, kept 94.2% with a deduction near 0.35 dollars. Note what these numbers are not. A TRC20 transfer cost about 2.12 dollars that week and a 141 vB Bitcoin transaction about 0.11. The 11 USDT and the 6.03 dollars are billing minimums baked into the quote, not miner fees passed through. (XMR to USDT exit)

The five-minute version. Quote three sizes for your exact pair: the minimum, 500 units, and ten times a size you would actually trade. Compute `F` on each. If `F` is more than 3% of the amount, a charge that does not scale is eating the order, so raise the size or accept it knowingly. Then watch the route count as you scale up: the moment it drops from two to one, the rate steps down, and that ceiling is where an order belongs in two pieces. Three habits keep the arithmetic honest. Measure inside a few minutes, because quotes drifted 0.2% to 0.37% over ten minutes in an earlier sample. Never compare a quote from this morning with one from tonight. And remember that a bigger order buys a better rate while carrying more price risk across the 30-minute deposit window. The band is a cost curve, not a reason to trade money you cannot afford to lose.