Our CTO & Co- Founder, Jesus Rodriguez unveils a series of new and unique analytics for order book datasets from cryptocurrency exchanges.
We highlight some unique insights that arise when combining exchange order books and blockchain datasets.
Finally, we show how order book metrics can be used
4. Why should we pay attention
to the behavior of order books?
5. 5
Some Reasons to Consider
There is an average of over 12x more volume traded in exchanges than what is
reported on-chain
Binance Bittrex
Reported - On Chain
8. 841
Nothing…
But…
Cross exchange analysis
The combination of exchange activity
and order book analysis provides a
more complete version of the
behavior of a crypto asset
Detecting and understanding the
“non-conventional” patterns in the
order books of crypto-exchange(fake
volumes, wash trading, etc)
13. MORE SOPHISTICATED
ORDER BOOK ANALYSIS
SIGNALS THAT COMBINE
BLOCKCHAIN DATASETS AND
ORDER BOOK FEES
SIGNALS TAILORED TO THE
“VERY PARTICULAR”
BEHAVIORS OF CRYPTO
EXCHANGE ORDER BOOKS
18. 18
The Reasoning
● There are decades of quantitative finance research based on order book
patterns
● Some highly used quant signals apply seamlessly to crypto assets
19. 19
Trades per Side
Measures the number (or volume) of trades where the buyers “crossed the spread” and
bought at the Ask price vs the number (or volume) of trades where sellers “crossed the
spread” and sold at the Bid price, per minute.
23. 23
The Reasoning
● Correlations between order book metrics and on-chain volume represents
relevant indicators in areas such as liquidity and risk.
● Flows of funds in and out of exchanges directly impact the behavior of
order books.
● Relationships between order book and on-chain inventories help
understand anomalies in crypto exchanges.
24. 24
To Do This We
Need to
Understand the On-
Chain Behavior of
Crypto Exchanges
25. 25
Exchange-On-Chain Market Depth
Shows the number of open buy and sell orders at different prices (bottom graph) and the
on-chain positions at those price ranges
31. 31
The Reasoning
● Machine learning and deep learning models can output results beyond
traditional statistical models
● Machine learning models can shade some light over the strange behavior
of crypto exchanges
32. Problem
Exchange order book
information is plagued
with bad data and wash
trading records
Solution
A machine learning
classifier can identify
potential fake trades and a
percentage of fake
volume.
Scenario: Fake Volume and Wash Trading Detection
35. Problem
Order books indicators are
some of the strongest
predictors of SHORT-TERM
price movements in an
asset, but most statistical
models fail to capture those
relationships
Solution
New generation deep
neural networks have
shown strong potential for
creating short term
prediction models across
different asset classes
Scenario: Price Prediction Direction
36. We are working on it!!
Next Webinar
Crypto asset predictions,
challenges and crazy
ideas using deep
learning
38. 38
Summary
● Order book datasets provide an incredibly rich source of information about
crypto exchanges
● Order book records and a natural complement to on-chain datasets
● To obtain meaningful intelligence we need to look past the traditional
order book signals
● The combination of on-chain and order book dataset reveals fascinating
patterns about the behavior of crypto assets