At TokenAnalyst we've always believed that there is fundamental value in blockchain data. Metrics such as active addresses, block rewards/hashrate, circulating supply, coin volumes and transaction counts are strong indicators of the health of a network. We often get asked – Are these metrics valuable for trading? Can on-chain data be used to predict short/medium term price movements? We believe the answer is YES, but not with the existing set of metrics.
New metrics such as flows into/out of centralized exchanges can be used for short/medium term trading strategies and to gain an edge while market making. We think exchange flows are important because centralized exchanges are where price discovery happens, and crypto has to flow into an exchange before it can be traded!
To limit the scope of our research, we focus our attention on Ethereum on-chain flows (from the TokenAnalyst Exchange API) to and from Binance, Bitfinex, Bittrex, Kraken, Kucoin and Poloniex and find the correlations among flows and Ether price.
We first look at the number of transactions to exchanges. We call transactions into exchanges inflows, while the transactions leaving the exchanges are called outflows. Using the TokenAnalyst API, we get all inflows of the last two years into the six above mentioned exchanges. We calculate the number of daily inflow transactions and plot them against the Ether price.
During the boom of 2017, we see that the number of inflow transactions (red) follows the pattern of the ETH price (blue). Interestingly, the curve for the inflows seems to precede the price, i.e. the inflows can be used as an indicator for the development of the Ethererum price.
In order to quantitatively check the correlation between the inflow and price curves, we calculate the Spearman correlation coefficient, which measures monotonic relationships. We calculate the correlation between each exchange's inflow and also the sum of the inflows of all exchanges (last line in the following table) against the ETH price.
A correlation coefficient of 1 would mean a perfect correlation between two metrics. We see a strong correlation for all exchanges, except for Kucoin, which has many outliers in the inflows and thus cannot accurately calculate correlation values.
Due to the extent of the 2017 boom, it is hard to visually confirm the correlation in the above graph for the time after the boom. Therefore, we also plot the number of inflows against the price for the past few months below.
It can be observed from the curves that while the number of inflows is volatile, the correlation between the two metrics is still strong. In fact, our analysis has shown that the correlation is strongest for times of high price volatility and the price volatility has strongly increased in the past few months, reaching levels similar to the beginning of the boom in 2017.
Using the knowledge that the number of inflows is strongly correlated with the price, one could set up a trading strategy based on the inflows. The number of inflows can be used as signals by taking today's value and comparing it to the average of the previous week's values. To quantify the strength of the signal, all values should be standardized by dividing them by the standard deviation. A trading strategy can then be set up that buys or sells depending on the sign and the strength of the signal. Using historical data, the trading strategy can be backtested to validate its profitability, before implementing it for future trades.
In this post, however, we will focus on extracting more metrics about exchange flows. We will give more details on setting up trading strategies in a later post.
After having analyzed the number of inflows into exchanges, we are going to go into more detail and compute the average values of inflows. Large average values signify that large players such as hedge funds are active in the market, while small average values point to retail traders. We suspect that the presence of large traders will have an impact on the price development. In particular, a high inflow amount could lead to larger price movements. Therefore, we will calculate the price volatility per day and compare it against the average inflow amounts.
The two curves seem to be strongly correlated and the inflow curve again precedes the ETH volatility curve. We can again calculate the Spearman correlation coefficient.
We observe a strong correlation between the average transaction size and price volatility. This means that when large players become active, the price volatility is expected to be high. This makes sense since large players have the capital to move the markets. However, the correlation could also be interpreted the other way around: times of high volatility are interesting for large players, such as market makers, that use volatility to make gains.
For our third analysis, we will again use the average size of inflows into exchanges (similar to the above). However, now we will compare it against the Ether price, instead of its volatility.
We can again see a strong correlation in the graph. The correlation coefficients confirm this impression.
The strong correlation means that at times where the ETH price is high, larger inflows are observed into exchanges on average. This might have three causes.
The three above analyses have shown that flows into exchanges are strongly correlated to the price and volatility and that they might be a driving factor. A logical next step and an interesting technical analysis is now to detect metrics of how exchange flows can be matched to price development which can be ultimately used for trading strategies.
Therefore, instead of analyzing high-level metrics such as the number of transactions or the average transaction size, we will now have a look at the amount of flow itself. More specifically, we calculate the daily net flow to exchanges, i.e.
net flow = sum(inflows)-sum(outflows). We will compare the net flow against the ETH price and look for correlation patterns. The following graph shows the two metrics for roughly the past year (we avoid plotting the 2017 boom to see smaller changes in the metrics).
At first sight, we see that this graph is not nearly as correlated as the previous ones. However, there seem to be some events that give an insight into the long-term development of the ETH price. In particular, positive outliers in the net flow are often followed by a price decrease (Nov 2018) or the end of a price hike (May 2019). Similarly, negative outliers in the net flow are often followed by the end of a price dip (Dec 2018) or proceeds a price hike (May 2019).
There is a logic behind this observation. At times of large positive net flow, there is a lot more inflow into exchanges than outflow. When traders send their ETH into an exchange, it is usually to sell it. Therefore, a large series of inflows will lead to an increased supply in ETH, which will lead to falling prices. On the other hand, a large negative net flow means that a lot of outflow is happening. This means that traders have bought ETH, therefore reducing the supply and thus inducing a price hike.
When observing the above plot in more detail, it becomes clear that this is not a certain signal but can be rather seen as a hint towards buying or selling sentiments at the market. An interesting next step would be to use this insight for setting up a trading strategy and backtesting on our historical data.
We have used TokenAnalyst's data to extract metrics regarding on-chain exchange flows and have gained the following insights.
By Lorenzo von Ritter