◈ market integrity · rolling 30-day median · 14 venues
Real vs Reported
Exchange Volume
Every CEX claims big numbers. We score five fingerprints from the actual trade tape — without orderbook context, just the prints — that betray inflated or recycled volume. Rank #1 is the cleanest tape.
Soli's reading
OKX sings clean. KuCoin sings off-key. Trust the prints, not the press release.
OKX sings clean. KuCoin sings off-key. Trust the prints, not the press release.
Reported · 24h
$658.06B
sum of pair medians across venues
Honest · est.
$440.38B
reported × (1 − wash score)
Aggregate honesty
67%
real ÷ reported, all venues
Pairs · venues
8343 · 14
30d rolling median
Worst venue
KuCoin
27/100 · 68% honest
◈ compare two venues head-to-head →
◈ venues · clean → suspicious
#
Venue
Reported 24h
Honest est.
Honesty
Net flow 30d
Score
Verdict
#1
OKX
futures
662 pairs · worst BTC-USD_UM-260731 @ 55/100
$488.94B
claimed
$313.58B
estimate
64%
−$999.41B
buy 49%
12
clean
→
#2
Binance Futures
futures
749 pairs · worst POPMARTUSDT @ 53/100
$37.77B
claimed
$27.53B
estimate
73%
−$6.86B
buy 48%
14
clean
→
#3
Binance
spot
515 pairs · worst USDPUSDT @ 66/100
$7.31B
claimed
$6.46B
estimate
88%
−$933.2M
buy 49%
16
clean
→
#4
Bybit
futures
700 pairs · worst BTCUSDT-15MAY26 @ 53/100
$9.71B
claimed
$6.47B
estimate
67%
−$1.42B
buy 49%
17
clean
→
#5
OKX Spot
spot
810 pairs · worst VELODROME-USDC @ 66/100
$1.59B
claimed
$1.47B
estimate
92%
−$294.3M
buy 49%
17
clean
→
#6
Coinbase
spot
441 pairs · worst AVAX-USDT @ 83/100
$1.53B
claimed
$1.41B
estimate
92%
−$942.4M
buy 49%
18
clean
→
#7
Hyperliquid
futures
234 pairs · worst PROMPT-USDC @ 81/100
$5.46B
claimed
$5.08B
estimate
93%
−$130.5M
buy 49%
20
fishy
→
#8
Bybit Spot
spot
556 pairs · worst A8USDT @ 69/100
$1.49B
claimed
$1.34B
estimate
90%
−$324.9M
buy 50%
20
fishy
→
#9
Bitget
futures
800 pairs · worst COSTUSDT @ 55/100
$7.32B
claimed
$6.38B
estimate
87%
−$1.30B
buy 49%
20
fishy
→
#10
Exchange24
futures
266 pairs · worst H-USDT @ 54/100
$1.32B
claimed
$1.17B
estimate
89%
+$373.9M
buy 35%
21
fishy
→
#11
Exchange51
futures
224 pairs · worst MYXUSDT @ 48/100
$1.68B
claimed
$1.48B
estimate
88%
+$603.9M
buy 50%
22
fishy
→
#12
Bitunix
futures
729 pairs · worst FOGOUSDT @ 53/100
$59.43B
claimed
$41.90B
estimate
71%
−$46.16B
buy 49%
22
fishy
→
#13
Gate Futures
futures
922 pairs · worst AIOT_USDT @ 60/100
$10.74B
claimed
$9.85B
estimate
92%
−$467.7M
buy 50%
22
fishy
→
#14
KuCoin
futures
735 pairs · worst WCTUSDTM @ 64/100
$23.78B
claimed
$16.27B
estimate
68%
+$3.74B
buy 50%
27
fishy
→
◈ methodology · five fingerprints from the trade tape
Timing regularity
Real flow arrives in Poisson bursts (high coefficient of variation, >1). Bot grids tick on metronomic intervals — CV drops near 0. We penalise the metronome.
Benford divergence
Genuine prices obey Benford's Law (more 1s than 9s as first digit). Round-number wash patterns flatten the distribution. We compute KL divergence vs Benford on notional > $10.
Self-trade pairing
Wash trading mints buy/sell pairs of identical size within milliseconds. We count pairs with size-fraction tolerance 0.1% inside a 50ms window. Clean tape: <1%. Wash: >25%.
Price impact
Real flow moves price (Kyle's lambda > 0). Wash trades cycle without impact. We compare against a per-base Binance baseline λ and penalise venues whose impact is near zero relative to it.
Size entropy
Organic flow uses many size buckets (Shannon entropy near 1). Wash bots reuse the same notional sizes — entropy collapses. Normalised over log10-bucketed notionals.
Per-pair total = 0.10·timing + 0.20·round + 0.35·pair + 0.15·impact + 0.20·entropy. Per-day score over each 24h tape; the table shows the 30-day median. Median-of-medians at the venue level → robust to one or two outlier pairs.