REAL Ranking

The REAL score — Reserves, Execution, Activity, Liquidity — for every exchange, and how each pillar is built.

Reserves · Execution · Activity · Liquidity

One REAL score for every exchange, built only on what we can measure or independently verify: the reserves behind the venue, execution quality from tick-level fills, volume and open interest that survive verification, and order-book depth we snapshot ourselves. It is not a judgement of conduct or solvency; a low pillar means weak measurable evidence, and where our measurement is the limit the venue is not penalized.

Why this design: the usual exchange ranking is a volume table, and volume is the one figure a venue reports about itself and can inflate at no cost. Each pillar here is something a venue cannot print. Reserves must show up on chain or in filed accounts, execution is measured from the fills that actually happened, activity only counts volume and open interest that our own order-book snapshots could support, and liquidity is resting depth we observed. The weights put half the score on those two hardest-to-fake pillars.

RReserves30 EExecution30 AActivity20 LLiquidity20points out of 100 · weights redistribute over measured pillars

S85+A70+B55+C40+Dbelow 40S also requires measured execution.

AllCEXDEX

59 exchanges scored · computed 2026-09-11

#	Exchange	Tier	Score	Reserves 30	Execution 30	Activity 20	Liquidity 20
1	Binancecex	S	86	92	59	98	99
2	Coinbasecex	A	79	100	n/a	72	75
3	OKXcex	A	77	80	56	88	88
4	Hyperliquiddex	A	74	82	50	84	86
5	Bybitcex	A	72	75	47	86	87
6	Krakencex	A	72	77	42	71	84
7	Gatecex	A	71	66	52	83	91
8	Bitgetcex	B	69	66	52	77	91
9	MEXCcex	B	68	61	57	76	85
10	Upbitcex	B	67	71	n/a	70	59
11	Bitfinexcex	B	67	70	39	62	67
12	Deribitcex	B	67	66	72	59	67
13	Asterdex	B	66	61	n/a	70	70
14	Lighterdex	B	66	62	62	65	77
15	HTXcex	B	65	64	65	69	61
16	Crypto.comcex	B	61	59	46	70	64
17	KuCoincex	B	60	59	51	68	61
18	Variationaldex	B	59	50	n/a	67	66
19	Bitstampcex	B	59	56	n/a	66	57
20	Geminicex	B	58	66	n/a	57	46
21	Bithumbcex	B	57	58	n/a	67	48
22	edgeXdex	B	57	42	n/a	66	69
23	Bitvavocex	C	55	48	n/a	63	56
24	Extendeddex	C	54	48	n/a	49	67
25	Phemexcex	C	54	36	47	72	63
26	GRVTdex	C	52	39	n/a	60	65
27	Bitruecex	C	51	26	n/a	75	66
28	QFEXdex	C	51	50	n/a	46	57
29	Ondo Perpsdex	C	50	50	n/a	44	55
30	Backpackcex	C	48	45	n/a	48	53
31	Nadodex	C	48	41	n/a	47	58
32	SoDEXdex	C	46	43	n/a	38	59
33	Pacificadex	C	45	35	n/a	50	54
34	SunXdex	C	45	50	n/a	34	46
35	CoinExcex	C	44	37	n/a	49	48
36	ApeX Omnidex	C	44	29	n/a	55	54
37	StandXdex	C	42	37	n/a	47	45
38	Derive V2dex	C	41	51	n/a	39	28
39	Arcusdex	C	40	34	n/a	35	57
40	BingXcex	D	39	41	n/a	76	0
41	Katana Perpsdex	D	39	50	n/a	32	30
42	Hibachidex	D	39	50	n/a	29	33
43	Antarcticdex	D	39	25	n/a	52	47
44	Injectivedex	D	39	50	n/a	25	36
45	HashKeycex	D	38	57	n/a	29	18
46	dYdXdex	D	35	43	0	38	71
47	RiseXdex	D	34	30	n/a	31	43
48	Phoenixdex	D	33	18	n/a	33	55
49	GUMdex	D	31	50	n/a	18	17
50	Paradexdex	D	31	33	n/a	27	31
51	Perpldex	D	30	32	n/a	32	26
52	XT.COMcex	D	24	0	n/a	54	31
53	WhiteBITcex	D	23	0	n/a	42	39
54	01 Exchangedex	D	22	24	n/a	22	20
55	Bybit EUcex	D	19	0	n/a	40	27
56	Delta Exchangecex	D	19	0	36	13	26
57	Binance.UScex	D	17	0	n/a	40	21
58	OKX EUcex	D	16	0	n/a	32	25
59	Gate EUcex	D	12	0	n/a	22	20

How the REAL score is built

REAL is a weighted mean of its four 0 to 100 pillars. Within each pillar, size is log-scaled: the largest venue scores 100 and a venue a tenth its size still scores about 70, so a mid-sized honest venue is not crushed by the giants. Every credibility check lives inside the pillar whose data it questions and lowers that pillar; there is no separate penalty layer.

An input we do not have is not counted as zero: its weight redistributes over the pillars we did measure. An input a venue does not publish where its peers do scores zero. Execution and open interest describe the perp business, so in a venue-wide score their weight scales with the venue's share of verified volume that is perps; a spot-first venue is judged on its spot business.

The checks, and the pillar each one lowers

Open interest is credited at most 3× clean reserves. Positions the venue margins are bounded by what it custodies, so turnover the balance sheet cannot afford is discounted.
Depth is capped at clean reserves, and a CEX publishing none takes a ×0.6 haircut: nothing reconciles its book.
A book that never moves is decoration. Near-zero 30-day depth variation discounts the Liquidity pillar by up to 30%.
Where our probes established a book is substantially decorative across the whole ladder, depth credit is withheld rather than capped — whatever survives a cap is the cap, not measured liquidity.
Execution cannot out-credit measurement: a venue with no measured Execution pillar is capped at 66 on it, so missing data never beats being measured.

CEX and DEX are ranked within their own type, because the composite renormalizes reserves and regulatory inputs away for DEXes and the raw scores stop being comparable across types. Nothing self-reported that we cannot check enters the score — not reputation, age, brand, listing count or following. A venue's score moves when its measured behaviour moves.

RReserves

30 points

The one number a venue cannot print. Two thirds is magnitude: clean reserves by evidence tier (live on-chain wallets full credit, regulator-filed audited financials ×0.95, third-party attestations ×0.9, a figure older than a year drops a tier), log-scaled so only the largest balance sheet reaches full marks and a tenth of it still scores well. One third is consistency: whether the resting book, reported volume and open interest are sizes that balance sheet could actually back, plus custody quality (clean share, own-token concentration). DEX collateral is read on-chain from protocol TVL, credited at 5× as a floor on margin held elsewhere.

Magnitude 20 of 30: Clean reserves by evidence tier (live on-chain wallets ×1.0, regulator-filed audited financials ×0.95, third-party attestations ×0.9; a figure older than a year drops a tier), log-scaled from $100M to the largest balance sheet in the group. Only the leader reaches 100; a venue a tenth its size still scores about 70.
Consistency 10 of 30: Whether the venue's resting book, daily volume and open interest are sizes this balance sheet could back (full credit at or under 1×, 2.5× and 3× reserves, ramping to zero at 5×, 8× and 10×), plus custody quality: clean share and own-token concentration.

When it is missing: Never N/A for a CEX: publishing nothing where peers publish scores 0. A DEX with no mapped protocol takes the measured-DEX median.

Best

Coinbase100Typical

GUM50Lowest

Gate EU0

EExecution

30 points

Measured execution quality on the shared BTC/ETH/SOL/HYPE/BNB perp basket, each venue compared with peers on the same pair: markouts (adverse selection), all-in cost of a $10k order at the base fee tier, and actual slippage at fixed order sizes. The result is then discounted by up to 12% where fills came in materially worse than the venue's own quoted book promised at that size: an expensive quote honestly filled is not penalised; an advertised tight quote that does not deliver is. Feed speed is displayed but not scored: emission policy is not engine latency. In a venue-wide score the pillar's weight scales with the venue's verified perp share, since the basket measures perp execution.

Markouts 30: Where the price goes 1s and 1m after fills. Persistent moves against takers mean fills systematically happen at bad moments.
Cost to trade 20: All-in cost of a $10k market order: base-tier taker fee plus half the effective spread.
Slippage at size 20: What fills in the $10k and $100k size buckets actually paid beyond mid, size-controlled so venues are compared on the same trade.

When it is missing: No tick coverage for the venue, or less than 72 hours on every basket pair. Spot-only venues are not covered yet.

Best

Deribit72Typical

KuCoin51Lowest

dYdX0

AActivity

20 points

Verified volume (only what survives normalization against the books we snapshot, discounted where volume-reporting markets show no book), open interest (credited where we observe the perp book, at most 3× clean reserves, discounted when volume turns positions over implausibly fast unless the balance sheet behind the venue makes that turnover affordable), and how steadily depth, volume and positions hold across the window. Log-scaled. In a venue-wide score the open-interest weight scales with the venue's verified perp share.

When it is missing: None of the reported volume sits in markets we snapshot, so nothing could be verified; a perp venue whose OI we have not ingested is our gap.

Best

Binance98Typical

Antarctic52Lowest

Delta Exchange13

LLiquidity

20 points

Resting ±2% book depth over 30 days plus breadth (markets holding ≥$100k of depth), log-scaled. Depth credit is capped at clean reserves and discounted for a CEX publishing no custody evidence or running a book that never moves.

When it is missing: A scope with no book we ever observed is not ranked at all rather than scored low.

Best

Binance99Typical

Bitvavo56Lowest

BingX0

Tiers

DCBAS

Fixed bands on the 0 to 100 score, not quantiles, so a new listing does not reshuffle everyone. S also requires a measured execution pillar: the top grade cannot be earned from displayed data alone. A venue with no observed order book is not ranked rather than scored low.

REAL scores are recomputed after listings or material data changes, reviewed, and published; there is no unreviewed automatic re-ranking. Every venue's pillar breakdown is shown on its profile so the arithmetic behind every tier is inspectable. The ranking, its pillars and their weights are subject to change as coverage improves.