The Impact of On Chain Exchange Activity On The Price of Ethereum

The internal plumbing of Cryptoassets are transparent

In prior articles, we have analyzed the long term (6+ months) insights gained from viewing on chain behavioral activity. Now let’s dive into the short term (24 hours) insights from monitoring on chain activity. As a reminder, we have collected on chain activity by exchanges, whales, retail, and miners. The short term trends are heavily influenced by the activity of the exchange accounts.

Before we show you the data, you should understand the flow at exchanges for buying and selling ETH. To buy ETH, you usually have to go to a centralized exchange, send them cash or other crypto, purchase the ETH from the exchange, then investors often take their ETH off the exchange to store in their private wallet. To sell ETH, you need to send your ETH from your private wallet to the exchange wallet, sell it, and take your cash or crypto off the exchange. Anytime you move from a private wallet to an exchange and back we get to see that activity on chain.

Hypothesis

Given that flow, I would expect the following to occur:

1) If I withdraw ETH from an exchange I have high likelihood of holding that ETH, causing the price of ETH to rise

2) If I deposit ETH at an exchange I have a high likelihood of selling ETH for something else, causing the price of ETH to drop.

Testing
Every day investors deposit ETH at exchanges and withdraw ETH from exchanges. If we sum up all the transactions for a given day and plot it on a histogram we get this:

In white is what happens two thirds of the time. In blue and green is what happens in the other third, the days where 50k+ ETH moves in our out of exchanges. This is the unusual activity that could cause the price of ETH to move over the next 24 hours.

Testing that hypothesis here are the results: When investors are withdrawing more ETH than usual from exchanges the price of ETH goes up over the next 24 hours 62–64% of the time.

The above table shows how ETH performs over the next 24 hours after hitting different zones since Jan 1, 2019. The columns are UTC times that mark the start of a simulated trade, there are slight differences in outcome depending on what time of day you trade.

First, we present the entire dataset for that period: Average Daily ETH Return is the average return of ETH over the next 24 hours, and % of Days ETH Goes Up is the % of 24 hour periods that ETH went up.

The Green Zone Average ETH Return is the average return of ETH over 24 hours after a Green Zone is hit. Green Zone % of Days ETH Goes Up is the % of the time that ETH goes up over the next 24 hours after a Green Zone is hit. The Green Zone P-Value is for you statisticians, a value closer to 0 means the ETH price behaves differently in the Green Zone vs. all other periods. If you recall from high school science, a p-value of 0.05 is statistically significant. We got close, but probably wouldn’t pass the test of getting published in a scientific journal.

The Blue Zone did not behave as we expected, when exchanges receive unusual amounts of ETH, it is not predictive of the price going down. To find evidence for the second part of the hypothesis for ETH deposits leading to a drop in price we need to focus on the behavior of the more sophisticated investor: Whales. Looking at the distribution of whale activity with exchanges we see a similar pattern, though skewed more heavily to Blue Zone transactions: ETH deposits at exchanges.

In testing the second part of the hypothesis we notice the following: When Whales send unusual amounts of ETH to exchanges the price of ETH goes down within 24 hours 54–57% of the time (i.e. it goes up 43–46% of the time). The p-values are not as impressive as the aggregate data but still signal we are onto something.

Conclusion

Monitoring exchange activity on a short term basis works. The signal will fire one third of the time, and when fired you have 54–62% hit rate in guessing the direction of ETH in the next 24 hours. In stock picking, those hits rate can build a very solid portfolio over time.

The short term is interesting, but the long term view is much easier to trade. I encourage you to check out our prior articles to see our long term view of ETH and BTC.

Follow us to see our next article, where we publish our ideal portfolio.

Notes on this analysis

• The price change over 24 hours assumes that it takes me one hour to compile and process the data before I trade.
• I am showing data as of Jan 1, 2019. The trend is even better (i.e. p-values of zero) if I include 2018 and 2017. However, I did not think it was fair because I tagged the list of exchanges in the fall of 2018, so it would include behavior of exchanges that in my simulation would not have actually been tagged. This also suggests other investors are running a similar strategy of tracking exchange activity.
• Similar trends exist for BTC but they are not as strong, you need to go deeper into Green and Blue zones to find shifts in the price of BTC.
• The time range of 16–21 UTC is noon to 5pm EST. I limited my selection to between 9am-5pm EST time. I then chose the afternoon trading times because the p-values were better during the afternoon, perhaps investors perform their ETH actions with exchange in the afternoon to trade the next morning.
• Why are we sharing this? Tagging activity by behavior is really hard and monitoring it on a short term basis gives you insight on only two assets (ETH and BTC). We prefer the long term viewpoint, contact us if you want the data.