This document discusses challenges with measuring liquidity and detecting bad behaviors in crypto exchanges. Traditionally, order books and web activity are analyzed but are easy to manipulate. The document proposes analyzing exchanges at the blockchain level by using machine learning to classify exchange wallet addresses and transactions. This could provide more robust liquidity measures like inflows/outflows correlating with order books, exchange inventory levels, and liquidity provider interactions. Five new ideas for blockchain-based analysis are described to tackle exchange liquidity in a way that is difficult for bad actors to manipulate.
2. Agenda
● Liquidity and bad behaviors in crypto
exchanges
● The challenges with the traditional
method
● The blockchain angle
● Some new ideas for measuring liquidity
in detecting bad behaviors in crypto
exchanges
4. Crypto
Exchange
Liquidity
Crypto exchanges regularly manipulate key
market indicators
Liquidity is a key indicator of the viability of
an exchange
Key indicator to detect bad behaviors
Fundamental to trading strategies
5. Fake volumes in exchanges are
one of the biggest challenges in
the crypto market
6. All Sorts of Bad Behaviors
Wash
Trading
Spoofing Front
Running
Tape
Painting
7. Wash Trading
• Simulate buy or sell
activity by trading with
the same person or entity
• Often used to generate
fake volumes or trigger
commission fees
8. Spoofing
• Fake large orders to
create the illusion of buy
or sell pressure
• Its used to manipulate
market sentiment
9. Front Running
• Used by exchanges to
take advantage of buy or
sell orders before its
customers can
• Very common in high
frequency trading
scenarios
10. Painting the Tape
• Similar to wash trading
but with multiple
participants
• Multiple players create
the illusion of market
activity by trading among
themselves
19. Order books analysis is effective to
detect patterns and anomalous
activity but an inefficient metric of
liquidity…
20. It becomes a race. For any new
liquidity methodologies, bad actors
can create new manipulations that
adapt to it...
21. The exchange liquidity challenge can’t be solved
with a single score and it rather requires a
combination of solutions each one tackling
specific dimensions of the problem….
22. Let’s look at a different dimension
of the problem….
23. Real centralized crypto exchanges need
to interact with blockchains to transfer
funds in and out of its accounts…
25. The challenge is that centralized
exchanges remain black boxes that
result incredibly hard to analyze…
26. But what if we could understand
crypto exchanges at the
blockchain level?….
27. Used to
transfer
funds out
of the
exchange
Used to
transfer
funds into
an exchange
Used to secure
the storage of
crypto-assets.
The main
interaction point
between
external parties
and an
exchange.
Hot
Wallets
Cold
Wallets
Withdrawal
Addresses
Deposit
Addresses
28. Using Machine Learning Classification Methods we can
Get a Solid Understanding of the Architecture of a
Centralized Exchange
35. Idea #2
Exchange netflows should maintains
a regular correlation with the
corresponding orderbooks.
Variations of that rule could be signs
of anomalous activity…
36.
37. Idea #3
The inventory held in exchange’s
deposit addresses, hot wallets and
cold wallets is a relevant measure of
liquidity…
43. 43
● Exchange liquidity and fake volumes remain real challenges in the crypto
space
● Addressing this challenge will require a compendium of different solutions
instead of a magic metric
● Machine learning classification methods enable the deanonymization of a
relevant percentage of the blockchain records related to crypto exchanges
● Deanonymized blockchain datasets offer a unique angle for analyzing
liquidity and fake volumes
Summary