Thirty-five percent of what
12 November 2026 · figures measured 12 November 2026
Yesterday the standings said PegasusSwap answered 35% of the requests put to it, second from the bottom. The number was accurate. What it measures is less obvious, and it turned out to be three different things added together.
Three ways not to answer
When a provider returns no price, the reason is kept. Reading the recorded fan-outs back, every miss falls into one of three groups:
- It does not trade the pair. PegasusSwap says so with a 422 and the words “Pair unavailable”. SideShift says “estimate failed”, which means the same thing and does not say it.
- The amount is outside its range. Too small, or above a maximum.
- It refused the request. A timeout, a server error, a rate limit.
Only the third is about whether a provider can be relied on. The first is about the size of its catalogue, and the published figure was counting it as a failure.
Counting only the pairs it trades
The fix is to count a provider’s answers only on pairs it has answered at least once. Every answer is on such a pair by definition, so only the denominator changes:
| Answered | On pairs it trades | |
|---|---|---|
| FixedFloat | 99.9% | 99.9% |
| SideShift | 41.2% | 99.7% |
| CCECash | 65.3% | 99.3% |
| PegasusSwap | 35.1% | 89.8% |
| ChangeHero | 51.5% | 83.2% |
| Changelly | 71.4% | 71.5% |
| Exolix | 45.0% | 58.1% |
| ETZ-Swap | 29.1% | 29.2% |
SideShift was ranked sixteenth of eighteen for reliability. On what it actually trades, it misses seven requests in 2,523, and it now sits second. Across the whole record, PegasusSwap’s 3,810 “Pair unavailable” answers were most of its misses, so its real problem is coverage: 17 pairs, where the widest providers quote 50 to 60.
Correction, 13 November: most of those 3,810 were this site sending the wrong network names, not pairs PegasusSwap does not trade. See the spelling was ours.
Changelly hardly moves, because its misses are almost all amounts. That is a real limit for anyone sending a small sum, so it stays counted.
The one that did not move
ETZ-Swap is 29% either way, and its own words say why. Across the whole record, 2,118 of its misses read “Client banned”. That is not a pair or an amount. It is this site’s server being refused. From 30 October to 5 November it answered nothing at all, and since 20:19 UTC on 11 November it has again answered nothing.
Why is not something this record can say. It can say that the low figure is not ETZ-Swap’s catalogue, and that an average over those weeks describes the connection between two servers as much as it describes a provider.
What changed here
Both figures are now published side by side, on /providers, on each provider’s page, and as answeredOnItsPairs in the open data. The standings are now ordered by the second one. The first stays, because a reader choosing a pair still needs to know who will not quote it. It just no longer decides who looks reliable.
Every figure above can be checked against the providers yourself — no account, no sign-up, and nothing is held for you at any point.