Crypto Execution Cost Study, Version 1.0 — citable findings
Frozen release. 50 sweeps of 24 scenarios, 73,001 observations, collected to . SHA-256 verified. CC BY 4.0. These numbers do not change. The study continues live at /research/cheapest-crypto-exchange, where figures move every six hours — cite this page, not that one.
No exchange is cheapest everywhere
No exchange was cheapest across all 24 scenarios. The most frequent winner, MEXC, led 11 of them (46%); 4 different venues won at least one.
Method: Per cell, the cheapest sufficient-depth venue in each sweep; scenario winner = modal winner over the window. Scenario: 24 buy-side cells (asset × market × order size), full window. n = 50; window to . Coverage version 3. (FeeFloor Crypto Execution Cost Study v1.0.)
The lowest headline fee usually is not the cheapest trade
Across 1200 comparable measurements (≥2 venues able to fill), the venue with the lowest advertised fee was also cheapest all-in only 65% of the time.
Method: Per observation: compare the lowest-advertised-fee venue to the lowest fee+spread+slippage venue; count agreement. Scenario: Buy-side cells where ≥2 venues could fill the size. n = 1,200; window to . Coverage version 3. (FeeFloor Crypto Execution Cost Study v1.0.)
The cheapest venue changes with trade size
In 4 of 6 asset-market lines, the cheapest venue at the smallest size was NOT the cheapest at the largest — fees are flat but depth is not.
Method: Compare the modal winner of the smallest-size cell to the largest-size cell in each asset-market line. Scenario: Per asset × market: winner at the smallest vs largest order size. n = 6; window to . Coverage version 3. (FeeFloor Crypto Execution Cost Study v1.0.)
The advertised fee is a fraction of the real cost
On a $100K BTC perp buy, across order-book venues the advertised fee was ~75% of total cost; spread ~6% and slippage ~19% made up the rest.
Method: Mean component share (fee / spread / slippage) of total all-in cost across sufficient-depth order-book venues. Scenario: BTC perpetual, $100K, order-book venues (oracle venues excluded), latest sweep. n = 40; window to . Coverage version 3. (FeeFloor Crypto Execution Cost Study v1.0.)
Buying and selling do not always share the cheapest venue (perp)
On perp cells, the cheapest venue to BUY differed from the cheapest to SELL in 1 of 12 paired cells (8%); median buy/sell cost gap 0.07 bps.
Method: Per cell, compare the modal buy winner (ask walk) to the modal sell winner (bid walk); gap = |sell winner total − buy winner total|. Scenario: 12 paired perp cells measured on both sides. n = 43; window to . Coverage version 3. (FeeFloor Crypto Execution Cost Study v1.0.)
Buying and selling do not always share the cheapest venue (spot)
On spot cells, the cheapest venue to BUY differed from the cheapest to SELL in 0 of 12 paired cells (0%); median buy/sell cost gap 0.23 bps.
Method: Per cell, compare the modal buy winner (ask walk) to the modal sell winner (bid walk); gap = |sell winner total − buy winner total|. Scenario: 12 paired spot cells measured on both sides. n = 43; window to . Coverage version 3. (FeeFloor Crypto Execution Cost Study v1.0.)
How to cite
FeeFloor (2026). Crypto Execution Cost Study, Version 1.0 [data set]. feefloor.com. https://feefloor.com/research/snapshots/v1 — CC BY 4.0. Data cutoff 24 July 2026.
dataset.csv SHA-256 42aa8b12082fb85dc41b16f066d6cc41576e20767915c4e30965880b290154ee