05 Dec 2014
Maker-taker interaction and bitcoin price formation
The volatility of bitcoin prices is often mentioned as one of the obstacles slowing down public adoption. A contributing factor to high volatility is low liquidity. Although liquidity is a diverse concept as discussed in our previous post, it is characterised by the ability of the market to absorb large buy or sell orders.
The price formation of bitcoin mainly happens on exchanges which bring together buyers and sellers of digital currencies. Traders on exchanges can be categorised into market makers and takers: market makers quote limit orders and signal their willingness to buy or sell bitcoin at two distinct prices, the bid and ask price. They then collect the price difference, or spread, as compensation for not being in control of when their orders are executed. Effectively they take the risk of buying (selling) bitcoin just before the price drops (rises).
Market takers in contrast pay the bid-ask spread in exchange for immediate execution of their market orders. Thus, they remove liquidity from the order book. Merchants who dispose of their bitcoin holdings fall into this category just like most profit-seeking traders.
If there are not enough quotes by market makers on the order book, large market orders will be able to shift the mid price significantly as they eat their way through the sparse limit quotes. This in turn manifests in high levels of volatility and can lead to the huge price swings the industry has become accustomed to.
To stabilise the price of bitcoin in the longer term, it is thus crucial that market makers add liquidity to the order books and help to absorb the price impact of huge market orders.
An example of how traders drive the bitcoin price
Taking liquidity by executing market orders is favourable if there are structural inefficiencies and mis-pricing. In the following, we provide an illustrative example of how these inefficiencies can be traded and outline why low liquidity can foster volatility.
The first figure below shows the average hourly traded BTC volume of Bitstamp (BTCUSD, UTC timezone, data obtained from bitcoincharts.com). Since Bitstamp is one of the most prominent European exchanges that facilitates exchange of bitcoin and USD, the traded volume is largest in the overlapping US morning / European afternoon sessions, while dropping off significantly overnight.
Interestingly, we observe heavy activity early in the European morning that does not fit into the general night-day rhythm of activity. Around 6-9am UTC time, consistently increased trading volumes were recorded over the past 5 month period.
Obviously there is a plethora of possible explanations of this phenomenon, however, we suggest the possibility that it might be due to regular merchant activity.
To illustrate how market takers can benefit from such a periodic pattern we develop a systematic trading strategy that follows the regular flows in the European morning. Our strategy thus benefits from the price impact of large market orders and could even reinforce them.
Figure 2 displays back-tested cumulative percentage returns of our market taker strategy, which shows strong performance over the previous five months. The fact that the model works can be attributed to the lasting price impact of large market orders and benefits from price volatility.
We believe that increased market making activity can reduce volatility by absorbing the price impact of large orders, thus removing the observed inefficiencies and stabilising the ecosystem.
03 Dec 2014
Bitcoin is known for its high volatility, which is seen as one of the largest hurdles on the way to widespread adoption. Consequently, a large number of industry professionals are working on solutions which aim to reduce the impact of volatility.
For example, some wallet providers like Coinapult or Bitreserve allow to "lock in" your bitcoin, similar to hedging with forward contracts, in order to guarantee a stable US dollar equivalent value. This can be an effective way for consumers to alleviate the effects of price volatility. However, why is bitcoin so volatile in the first place?
We believe one major reason is the lack of liquidity on exchanges. After all, the price of bitcoin is determined on exchanges and not on the block chain. Most exchanges also noticed this and are in a fierce battle for market share, as their income partly depends on users' trading volume. To make their exchange more attractive some offer better commissions to high volume traders, or provide features such as futures and margin trading. Some exchanges advertise themselves as "highly liquid", however, how is that liquidity actually defined? Liquidity is not simply money on the order book or volume traded. It is more subtle than that and in this post we will try to make the concept more accessible.
Defining and measuring liquidity is not straightforward as it cannot be summarised in one number. Liquidity is also not a concept that is either present ("we are highly liquid") or not, but rather changes continuously. We attempt to define liquidity as follows
Liquidity is the quality of executions received when trading for a period of time on a market place.
Here, quality of executions refers (but is not limited) to
Trading time: the ability to execute immediately at a given price. The number of trades per minute or the average inter-trade time is a relevant measure.
Tightness: the ability to buy and sell at about the same price. The difference between bid and ask price, or spread, is the defining factor of tightness. The spread is also a direct measure of the cost of a market order.
Depth: the ability to buy or sell a certain amount at any time. The sum of all limit orders at a given level of the orderbook make up the depth at that price.
Resilience: the ability to buy or sell a certain amount with minimal influence on the current price. A resilient orderbook is one which reverts back to its previous state after a large market order traded through a couple of levels of the orderbook.
For each of the properties above there is a number of measures one can calculate to assess that particular aspect of liqudity. Some look at orderbook snapshots in time, others look at the evolution of the orderbook over time. Even without measuring the observables above some qualitative conclusions are intuitive: An order book for instance with a thousand bitcoin on the first level of bid and ask but with only two trades a day is clearly not liquid.
Real liquidity requires a reasonable combination of the above characteristics. Note that it might be possible to define liquidity as shown above without even mentioning traded volume - an often evoked measure of liquidity across the industry.
To the interested reader, sites like bitcoinwisdom.com or cryptowat.ch give a good overview of the state of the order books on various bitcoin exchanges.
In the figure above we show a snapshot reporting spread, depth, and cumulative volume on the book taken from cryptowat.ch.
We could go into a lot of detail what measures are useful, perhaps we will do that in a future post if there is interest. For now we only want to demonstrate the usual lack of liquidity by most measures on bitcoin exchanges and its effect on the price during a particularly volatile period in December. We show evidence that individual traders are able to move the market with a small number of large trades due to a lack of depth and resilience requirements.
Lack of liquidity
The next graph shows the BTCCNY price on OKCoin China in the first three days of December overlaid with the cumulative order flow (cumulative sum of all market order transactions in BTC) on the right hand side vertical axis. Both time series appear highly correlated. This is indicative of rather illiquid markets with mostly profit-seeking players, where large aligned order flows drive the price and vice versa.
Low liquidity and reactive trading leads to volatile markets where the impact of large trades is not absorbed well but instead amplified. The sharp price drop in the morning of 3 December provides just one example.
The following figure takes a closer look at individual market orders that morning.
Up until 8:45 it is all quiet when one sell trade of slightly more than 200 BTC is executed. From that point on volatility increases and the price drops. If the trader of the first trade wanted to sell another 200 BTC 10 minutes later he would have received a price 2-3% worse than the initial one.
There's no immediate cure to the illiquidity and volatility of bitcoin markets apart from a gradual maturing of the ecosystem. However, we believe that market makers like ourselves can contribute to making this process as smooth and swift as possible.