This week bitcoin traded between $560/BTC and $580/BTC before tumbling to $440/BTC-$460/BTC amid rumors of another crack down by the People's Bank of China. Although the rumors have not been confirmed, certain news sources claim to have seen a PBoC document circulating which effectively bans bitcoin transactions in China. Some argue that the rumors could very well be true while others say this is just more of the cyclical ban/no-ban FUD from China. There are yet others who say that this is a ploy by government and media insiders to manipulate the bitcoin price for their own portfolios.
There was another rumor earlier this week that gox have, in their control, 670k of their "lost" BTC. If that is the case, it would put further downward pressure on the markets as unexpected supply gets reintroduced to the markets. This kind of market news, whether it be that coins are lost or found, always has two countervailing forces. First, a supply shock directly affects the price up/down depending on whether coins are lost/found respectively. Second there is a loss/gain of confidence in the bitcoin ecosystem when coins are lost/found which drives the price down/up. While the first effect is symmetric between positive and negative supply shocks, the second effect is not. The gain in confidence when coins are found is much smaller in effect than the loss of confidence when coins are lost. This suggests that the markets are overly optimistic about exchange counterparty risk likely because people are used to the ecosystem and infrastructure of mature exchanges and project those expectations onto this relatively fledgling industry.
Today I'd like to talk a little bit about the sort of information a person could glean from looking at the structure of the limit orderbook, particularly the dispersion of the limit order sizes at the same price level.
- First measure: Given a total size of 100 bid at some price level, it is a stronger "up" signal to see two bids than one bid e.g. (50,50) is better than (100). In general it is stronger to see multiple bids than a single bid given that the total size is the same e.g. (33,33,34) is better than (50,50).
- Second measure: Given a number of bids at the same price level, it is a stronger signal to see equally dispersed sizes rather than lopsided sizes e.g. (50,50) is stronger than (99,1). A simple mapping function that captures this relationship quantitatively and gives an ordinal ranking to size structures with the same number of elements is the geometric mean (as opposed to the arithmetic mean). For example f(50,50)=sqrt(50*50)=50 and f(75,25)=sqrt(75*25)~=43.3 and f(99,1)=sqrt(99*1)~=9.95 hence f(50,50)>f(75,25)>f(99,1).
- How do we create a tradeoff between a size structure which is strong by the first measure but weak by the second or vice versa. For example compare (10,10,80) with (50,50). (10,10,80) has more elements than (50,50) (first measure) but is also more disperse in its sizes (second measure).
- How does queue ordering affect short-term expectation? How does a size structure of (25,75) compare to (75,25) assuming the first number represents a bid/ask higher up the queue at the same price level than the second number?
- How about size structures across multiple exchanges? How does (50,50) on a single exchange compare to (50) on one exchange an (50) on another exchange?