We are building a trading brain that will 'learn' by use of Evolutionary Algorithms (EA). Skeleton strategies will be traded both with real money and hypothetically, and then put through a process of 'evolution' which can happen rapidly (thousands of cycles per minute) resulting in mutated, child strategies which are 1 in 1 Million. This process will be repeated in real time, ensuring that it is constantly adapting, constantly evolving. Trades will be executed automatically on client accounts starting at Bittrex which is our testing ground. At first we will use Crypto majors on single pairs and eventually a multi-dimensional strategy will be built. The beauty of using EA is that we can disclose this methodology to you because if you copy it, the results you get would certainly be different, because you would be using a different pool of base strategies, different GA methodology, different data, and other factors. That means this is non-competitive we could all be profitable using intelligent trading systems. And also for this reason, there is a compelling reason for open-source style collaboration, not to divulge our code, but to share in results of what methods work and what don't. For this reason we are being open about how we are developing this brain, and are open to feedback and suggestions from fellow quants.
Machine Learning with Genetic Algorithms
Core concepts: Artificial Intelligence, Machine Learning, Genetic Algorithms, Infinity and Limits, Combinatorics, Quantitative Finance
Goal: To develop an intelligent trading system for the crypto currency markets based on Machine Learning. Our target method is Genetic Algorithms, if this proves unworkable we will also explore Neural Networks and Fuzzy Systems.
Origin of iCATS
The creators have been working on algorithmic trading systems for some time (15 years). During that time, we noted that many very robust systems will work for a few months, or even years, and will eventually breakdown. The reasons varied; the market changed, the behavior of the price action changed, a political crisis caused the currency to shift in one direction. But one thing was common amoung all - they failed to adapt to market changes. A truly intelligent system which will evolve with the markets must self-adapt, or learn- what is the best new system. This simple aritifical intelligence we believe is the future of trading systems and the future of investing because:
- Markets become more complex over time
- Computing power grows over time
- As machines replace many automated processes, so they will too in markets
There are many funds already using AI to make trading decisions - this is not the point. Neural Networks are generally a means of optimization of parameters. A really intelligent system that self-adjusts in real time is our final goal and this can represent real intelligence. We believe the singularity has occured and computing power exists to assemble this system, and iCATS is the basis for this strategy. Crypto provides a good testing ground because it's less connected to 'real world' events, it trades all weekend, and has a low barrier to entry.