Gaining Valuable Insights from Crypto Exchanges
Since Bitcoin’s early days, exchanges have played a key part in the ecosystem facilitating trading activity and global reach of cryptocurrencies. However, it hasn’t exactly been a smooth ride for all crypto exchanges throughout these years, with several hacks and continued controversy regarding fake volume constantly grabbing headlines. These issues, along with lack of transparency, have made it difficult to extract trustworthy information regarding cryptocurrency trading and exchange activity. Fortunately, though, blockchain’s public and immutable nature allows us to shed light onto what’s happening behind the scenes at exchanges and derive valuable actionable insights for cryptocurrencies.
Prior to diving into some of these insights, first it may be helpful to review some of the basics of the types of addresses operated by exchanges, as well as the processes we utilized at IntoTheBlock in order to identify and track them. There are four main types of addresses managed by exchanges to manage users’ and their own tokens. These are:
Hot wallets — these addresses serve as the main point through which cryptocurrencies are transferred within the exchange. Hot wallets are online, making them easy to send and receive transactions, but also vulnerable to potential hacks. For this reason, exchanges’ hot wallets tend to receive funds from deposit addresses, creating a network of addresses to increase security.
Cold wallets — as you may imagine, these are sort of the opposite of hot wallets. Thus, cold wallets are offline making it harder to transact, but much safer. Because of this, exchanges utilize cold wallets to securely store users’ funds. Exchanges deposit in their cold wallets, tokens not needed for their daily liquidity needs.
Deposit addresses — these act as the entry point for cryptocurrency to flow into the ecosystem of an exchange. Similar to how a bank has branches, exchanges create these one-way addresses, for the purpose of receiving tokens.
Withdrawal addresses — addresses that facilitate transactions for cryptocurrency to flow out of an exchange.
These four main components of exchanges interact with each other, creating a system where each address’s qualities serve a purpose. While it may be simple to identify one deposit address as a user sending tokens to an exchange, exchanges have tens of thousands of addresses some of which are used only for a few transactions until another one is created. Therefore, this makes it a very complex and time consuming process unsuitable for manual identification. Instead, at IntoTheBlock, our engineering team tackled this challenge using multiple machine learning models.
Through ensemble learning, a technique in which multiple learning algorithms try to master the same task and are combined in a single master model, we were able to classify the Ethereum addresses for several large cryptocurrency exchanges. After readjusting the model multiple times, the ensemble model was able to determine the function of each address and extract data that otherwise would not be publicly available in a cryptocurrency exchange. For a more detailed overview of this process feel free to check out the video of my recent webinar on this topic.
The results were incredible, though they may be hard to digest at first:
As can be seen in the image above, tokens enter Binance through the deposit addresses (blue dots), after which they are transferred to a hot wallet (green) prior to being redirected into a cold wallet (black) or withdrawal address (orange). Tokens from several addresses are collected and sent within multi-input transactions that allow exchanges to move cryptocurrency arriving from a myriad of sources into their hot wallets at a fraction of the cost and time of doing these individually.
At this point, you may be perplexed and wondering “Ok, this may be an interesting visualization, but how does it affect me as a regular crypto user/trader?”
In short, exchange activity has large implications in both liquidity and price performance of cryptocurrencies. Exchange flows, akin to traditional markets’ fund flows, paint a top-level picture of global trading activity regarding a cryptocurrency and the market sentiment overall.
In highly volatile periods, such as the second week of this past March, we see a strong influx of tokens entering exchanges ( from customers' private wallets ) likely as some users panic and rush to sell. Simultaneously, though, several users may have seen this as a buying opportunity and then proceeded to withdraw their tokens into a more secure address. Additionally, this period of high turbulence resulted in the bid-ask spread of one of the most liquid exchanges jumping, thus making it more costly for users to trade. These patterns are depicted in the image below:
Similarly, the exchange’s blockchain data is also interconnected with other onchain metrics. This day of uncertainty and volatility now referred to as Black Thursday in the crypto space led to the volume recorded on the blockchain to hit year-to-date highs. Volume for large transactions, those over $100,000 in value, reached over 2 million Bitcoin as exchange inflows and outflows grew as can be seen below:
At IntoTheBlock, we have been carefully reviewing these and many more patterns arising from the exchange’s blockchain activity as it provides valuable insights for both trading and research purposes. For example, we are crafting an indicator tracking large ($100k+) inflows and outflows into large exchanges real-time as a potential buying/selling signal. As well, by comparing onchain volumes for an exchange with the ones reported by them, we can gain an idea of fake volumes and detect anomalies.
We are excited to share these indicators with you on our website as we expand our robust ecosystem of blockchain intelligence. Keep an eye out for this release coming soon!