arbiter

We sense a soul in search of answers.

If someone buys a bunch of Monero with Bitcoin on one exchange, how long does it take for this impact to spread to other exchanges? Due to the prevalence of automatic trading, probably not long.

Let’s look and see anyway!

First, we need to select a few exchanges. Unfortunately, many exchanges have been accused of fabricating, falsifying, and manipulating trade volume in order increase their rank. We want to look at mostly real data, so let’s go with (from largest to smallest): Binance, Poloniex, and BitTrex. These Know-Your-Credential markets offer USD pairings and are subject to tax law. Some consider them to be more reliable.

What do we expect to see? Well, Binance is one of the largest cryptocurrency exchanges in the world. One theory could be that Binance “leads the market.” BitTrex, on the other hand, is a much smaller exchange with less activity. Maybe we will see BitTrex play a game of catch up with Binance? However, if arbitrage occurs in under five minutes, all three exchanges may look identical here.

2020-07-12-arbitrage.jpg

Let’s zoom in on the sudden increase on 07-09-2020.

2020-07-12-arbitrage2.jpg

Looks pretty instantaneous to me.

It would be interesting to see the inclusion of additional exchanges and other market pairings, such as USD-XMR. We would like to create plots of historical data back going back to 2014, and live charts, too. Many exchanges report additional information such as the number of trades and book volume for each candle period, which may be valuable to see. Bisq, the free and open source decentralized exchange with autonomous governance, does not save user data at all. This would require us to set up a daemon monitoring and recording all recent trades. While we’re setting up daemons, we could also record a history of unfilled orders on an exchange’s order book… we…

… we better stop here for now.


Notes on data collection: Exchanges cannot stop users’ manipulation of the market. Brief Google searches will lead us to services that inflate a specific market’s volume for a given time period, for a price. More simply, an individual user can repeatedly buy his own sell order, inflating volume by paying the exchange’s minimal trade fee. These actions and many others that we are not clever enough to think of will be included in our data. The data used here was collected directly through each individual exchange’s API, which can be verified by any interested individual.

In order to turn the OHLCV candles for each exchange into line charts that can be overlain, we take the mean, $\frac{open + close}{2}$, and move the time stamp forward by half the time step. We chose to request data at five minute intervals because that’s the smallest interval shared by every exchange.