Algorithmic Finance

This paper examines a popular stock message board and finds slight daily predictability using supervised learning algorithms when combining daily sentiment with historical price information. Additionally, with the profit potential in trading stocks, it is of no surprise that a number of popular financial websites are attempting to capture investor sentiment by providing an aggregate of this negative and positive online emotion. We question if the existence of dishonest posters are capitalizing on the popularity of the boards by writing sentiment in line with their trading goals as a means of influencing others, and therefore undermining the purpose of the boards. We exclude these posters to determine if predictability increases, but find no discernible difference.

We study the problem of finding sparse, mean reverting portfolios based on multivariate historical time series. After mapping the optimal portfolio selection problem into a generalized eigenvalue problem, we propose a new optimization approach based on the use of simulated annealing. This new method ensures that the cardinality constraint is automatically satisfied in each step of the optimization by embedding the constraint into the iterative neighbor selection function. We empirically demonstrate that the method produces better mean reversion coefficients than other heuristic methods, but also show that this does not necessarily result in higher profits during convergence trading. This implies that more complex objective functions should be developed for the problem, which can also be optimized under cardinality constraints using the proposed approach.

This work's main purpose is to understand the price dynamics in a generic limit order market, and illustrate a dynamical trading mechanism that can be applied to explore its market microstructure. First and foremost, we capture the iterative nature of the limit order market, and quantitatively identify its capacities as a means to develop switching schemes for the appearances of different sorts of traders. After formally introducing a dynamical trading system to replace the complex limit order market, we then study trading processes in that trading system from both deterministic and stochastic perspectives, in the purpose of recognizing conditions of general instability and stochastic stability in the trading system. In the final part of this work, the dynamics of the spread and mid-price in a controlled trading system will be investigated, which fairly serves to verify the robustness of stochastic stability appearing in an uncontrolled trading system.

This paper analyses the co-movements between the US stock market and several commodity futures between 1998 and 2011. It computes dynamic conditional correlations at (i) 1-hour, (ii) 5-minute, (iii) 10-second, and (iv) 1-second frequencies and documents a synchronized structural break, characterized by correlations that have significantly departed from zero to positive territories, since late September 2008. Our results support the idea that high frequency trading and algorithmic strategies have an effect on the behaviour of commodity prices.

Understanding the microstructure of the financial market requires the processing of a vast amount of data related to individual trades, and sometimes even multiple levels of quotes. This requires computing resources that are not easily available to financial academics and regulators. Fortunately, data-intensive scientific research has developed a series of tools and techniques for working with a large amount of data. In this work, we demonstrate that these techniques are effective for market data analysis by computing an early warning indicator called Volume-synchronized Probability of Informed trading (VPIN) on a massive set of futures trading records. The test data contains five and a half year's worth of trading data for about 100 most liquid futures contracts, includes about 3 billion trades, and takes 140GB as text files. By using (1) a more efficient file format for storing the trading records, (2) more effective data structures and algorithms, and (3) parallelizing the computations, we are able to explore 16,000 different parameter combinations for computing VPIN in less than 20 hours on a 32-core IBM DataPlex machine. On average, computing VPIN of one futures contract over 5.5 years takes around 1.5 seconds on one core, which demonstrates that a modest computer is sufficient to monitor a vast number of trading activities in real-time – an ability that could be valuable to regulators.

By examining a large number of parameter combinations, we are also able to identify the parameter settings that improves the prediction accuracy from 80% to 93%.