Automated Bitcoin Trading: A platform for innovation?

Automation is something that fascinates me, no more so than in its application to financial trading; I find the notion that (in theory) I could go to sleep and wake up the next morning richer, thanks to a piece of intelligent software trading on a foreign exchange, thrilling. My recent (and, arguably, obvious) discovery that  autonomous cryptocurrency trading is a thing, was therefore one of glee, and I rushed to discover what the implications of this were to conventional automated financial trading. 

It is of no surprise that the impact of cryptocurrency on trading is the same as its impact on currency in general: the removal of financial middlemen and institutions. For those wishing to trade using autonomous bots, this means that they can connect directly to crypto-exchanges, rather than bearing the costs of accessing expensive APIs (Application Programming Interfaces), or being forced to seek the backing of a traditional broker. For me this is exciting, because I believe that the more open a system is, the more it is open to innovation; this is a logical conclusion if we consider the positive impact of techniques such as crowdsourcing. And, although brokers may grumble at being circumvented (particularly if the popularity and use of cryptocurrency continues to rise), it is them who will ultimately benefit from the creation (and subsequent dissemination) of new trading techniques from the Internet `hive mind’. What may these new techniques be? Well, in the same way that taking a technical project open-source tends to open up fundamental practices to review, the introduction of `open-source’ trading, it likely to challenge traditional views, and introduce unconventional techniques. For example, there are typical beliefs about the limitations of trading bots, voiced in the very article I reference for this piece: `It’s difficult to program a computer to react to fundamental market conditions such as, say, rumours about the Chinese government taking a new stance on Bitcoin, or the latest Bitcoin-based black market trading site shutting down.' Is it? Why? Semantic approaches to Artificial Intelligence (AI) (which is what these two quoted pieces of information are; semantic data), are rapidly becoming a reality, and I believe that open environments, such as Bitcoin exchanges, are the best place to implement and test these approaches, and breed similar ideas and techniques.

It is also of no surprise that the arrival of cryptocurrency provided a money-making opportunity for those with existing algorithmic tools available, particularly those tools tuned to process of arbitrage. These bots fared especially well with the introduction of crypto-exchanges, not only because of the limited number of exchanges (making their task simple), but because one particular exchange, Mt. Gox, dominated the market, allowing for its coins to be bought cheap on this exchage, and sold for profit on smaller ones. As with most things, this `free ride' did not last long for these traders, with Mt. Gox ultimately paying for its size by being the target of a mass Bitcoin theft. This theft, it would appear, serves to show the negative impact of open exchanges, as it is speculated that the thieves themselves used bot strategies to carry out their theft. 

Beyond the use of arbitrage, successful due to its exploitation of a young currency, what other autonomous strategies suit the trading of cryptocurrencies more than trading in conventional currencies? It would appear that data-intensive strategies, or bots that utilise mathematical techniques predicated on the presence of vast corpuses of data, fare well once again due to the open nature of crypto-exchanges. For example,'s vast dataset of over 200 points, collected entirely over a 6-month period this year, has allowed very simple probabilistic algorithms to correctly predict the movement of price in similar markets. For me, this demonstrates more than ever that there is no magic formula to automated trading, as there is no magic formal to its parent discipline, AI. In reality, intelligence can be simulated with straight forward probability, backed by huge amounts of data. And open-exchanges are providing the opportunity for this.





Martin Chapman - 2014