In this session, Roy presents his perspectives on key technologies and trends that are creating both challenges and opportunities in computational finance. He highlights trends in computational resources, quantitative analysis, system integration, and production deployment, and identifies how developments in MATLAB enable quantitative researchers and developers to stay ahead of their peers.
Dr. Roy Lurie, MathWorks
Most valuation and risk management models for derivatives are minor tweaks of Black-Scholes. Black-Scholes is a great model, whereas some of the tweaks are a step in the wrong direction. However, there are a few models based on Black-Scholes that are closer to how derivatives are used in practice and that capture essential features previously ignored. These are great models, better than all the more classical tweaks. So far so good. The main problem stopping these models from becoming more popular is that they are just too difficult to solve numerically when you have a typical portfolio consisting of many underlyings and many different types of derivatives. Can this number-crunching problem be solved?
Paul Wilmott, Wilmott Associates
The derivatives market has experienced significant changes since the recent financial crisis. A lot of efforts have been made to improve its safety, transparency, and efficiency. In this session, we look at an often overlooked but vital component of the derivatives market: its initial economic purposes, as tools for risk mitigation and alternative investment. Many financial decisions are based on the same ingredients: defining a subjective view, exposure, scenario generations, risk/utility criterion, and possible trades/products. We present a unified framework that incorporates all these ingredients and enables us to efficiently analyze, compare, and optimize investment and hedging strategies using derivatives. The main objective is to achieve better alignment of product to need/view. In this presentation, we demonstrate a MATLAB prototype we have recently developed that encompasses many aspects of derivatives, from payoff definition and pricing/calibration to risk analysis, back testing, and optimization, together with many innovative and flexible interactive and visual tools based on MATLAB GUIs.
Bryan Liang, Bloomberg LP
11:50 a.m.–12:30 p.m.
Deploying internally created financial applications intrafirm can be challenging enough, but as a boutique software and consulting provider to investment banks and institutional financial advisors, Intuitive Analytics develops software using MATLAB that simply must work across platforms and environments. Peter Orr, CFA, President of Intuitive Analytics, describes some of the unique MATLAB components his firm uses in its SmartModels Excel add-in software to manage everything from memory to intellectual property.
Peter Orr, Intuitive Analytics
In this session, we discuss with few words and many visualizations several advanced quantitative techniques for the buy side.
We also discuss multivariate Bayesian statistics, random matrix theory, robust estimation, and projection of risk to arbitrary horizons (with pitfalls).
Attilio Meucci, Kepos Capital and SYMMYS
Standard & Poor’s makes extensive use of quantitative models in many ways, including analyzing market conditions, forecasting stress scenarios, supporting credit analysis, and evaluating pricing for complex debt instruments. One important aspect of ensuring model quality is rigorous ongoing validation, particularly for models that are recalibrated daily. This presentation provides an overview of a framework deployed at S&P that incorporates MATLAB to produce and analyze daily validation reports. The credit default swap–based Market Derived Signals model is introduced briefly, and the corresponding automated framework for daily validation of model output is discussed. Additional reports that allow for time series analysis of model behavior over a specified time period are also considered. We briefly describe a dashboard currently under development that automatically evaluates the validation output and provides a compact display of the information. Finally, we provide an overview of the technical infrastructure that was built to support the engine that drives the validation framework.
Fair pricing in financial markets requires agents to treat prices like probability forecasts and to update their beliefs using Bayes’ rule. When risks are stable, rational learning is relatively smooth and prices behave as standard theory says they should. When risks change abruptly enough, rational learning creates market turbulence, analogous to the turbulence in fast-moving fluids. This session shows how MATLAB models can help explore rational turbulence.
Kent Osband, RiskTick LLC
Key to developing successful investment strategies is ongoing research, and this depends on the pace of investigating and understanding new ideas. Two variables can influence this speed of research progress: the organization of the research team and the tools they have available. In this session, Ben demonstrates how FFTW, the global fixed-income investment manager within BNP Paribas, used MATLAB to build such tools. They were built using standard functionality from MATLAB products, user-written toolboxes, community content from MATLAB Central File Exchange, and more advanced integration with native Java™. These tools have both allowed greater team collaboration and increased research bandwidth, together enabling faster research progress.
Ben Steiner, Fischer Francis Trees & Watts
This session focuses on elements of PnL volatility arising from dynamically hedging VA (GMxB) contracts with a particular focus on risk-managed subaccounts. Historically, variable annuity contracts have been criticized as being underpriced in the marketplace given that the benefits and guarantees they offer are very rich. In this session, we show that the cost of hedging can be controlled if variable annuity contracts are written on risk-managed subaccounts.
Eoin Elliffe, Lincoln Financial Group
MATLAB has traditionally been used for analyzing data offline, presenting analytic recommendations that were then acted upon manually. However, MATLAB supports a direct interface with data feeds and online brokers, as well as the ability to present sophisticated graphics and user interfaces—all in real time.
This presentation demonstrates an end-to-end demo trading system in MATLAB, highlighting its potential as a platform of choice. Interactive Brokers is used to demonstrate live market data feed and account/portfolio input, as well as for sending trading orders to the market. The system’s user interface showcases the hidden visualization and interactivity potential of MATLAB for tracking order executions and charting financial time-series in real time. Some best practices for improving real-time performance are also discussed.
Yair Altman, Undocumented MATLAB
This session demonstrates how to build a real-world asset allocation web application backed by MATLAB Production Server™. The application will be delivered over a thin-client browser on desktops, tablets, and phones with various modern web UI technologies as well as through a Microsoft® Excel® front end. Learn how MATLAB Production Server enables you to create such scalable, highly available server applications.
Ameya Deoras, MathWorks
11:50 a.m.–12:30 p.m.
Counterparty credit risk (CCR) presents a number of computational and architectural challenges. Different front- and back-office teams independently access data sets, request specific reports or quotes, and update, calibrate, and run models. Some of these tasks involve simulations, others involve pricing, and others involve reporting and sensitivity analyses. These are asynchronous tasks, some of which require significant computational power. In this presentation, we use an example to illustrate some of these workflows and to demonstrate how MATLAB Production Server can be a powerful tool in a CCR system.
Machine learning techniques are often used for financial analysis and decision-making tasks such as accurate forecasting, classification of risk, estimating probabilities of default, and data mining. However, implementing and comparing machine learning techniques to choose the best approach can be challenging.
In this session, you will learn about several machine learning techniques available in MATLAB. We illustrate how to quickly explore data, evaluate machine learning algoirthms, and apply the best algorithm for your problem. Highlights include unsupervised and supervised learning techniques such as:
Abhishek Gupta, MathWorks
Advances in parallel and distributed computing technologies and a reduction in the cost of computational hardware have opened up many opportunities for quants who rely on intensive computations for their research. Gone are the days when quants had to migrate applications written in high-level languages like MATLAB to low-level languages like C or C++ in pursuit of performance gains. In this talk, we discuss the seven technological innovations every quant should know. We demonstrate the opportunities and challenges and discuss how MATLAB applications can be scaled with parallel computing technologies, graphical processing units (GPUs), web technologies, and the Amazon EC2 cloud. We highlight the kinds of applications suited for these computing methodologies. We also make a case for how quants can leverage these new technologies and obtain performance gains without having to significantly alter their research process.
Sri Krishnamurthy, MathWorks
Learn how MATLAB can be used to perform advanced portfolio management and performance attribution. In this session, we explore a robust framework for portfolio optimization where we relax some typical simplifying assumptions. The session covers how built-in functionality allows you to easily deal with incomplete/messy data, multiple asset classes, incorporating transaction costs, and fat-tail return distributions.
Leveraging Financial Toolbox™, we characterize our results using descriptive statistics at the portfolio level as well as decompose returns using factor analysis. Using predictive modeling techniques, we can then back-test our strategy against a benchmark index to get a complete picture of performance. Finally, we demonstrate how an end user can leverage your model by generating a dynamic report capable of “what if” scenario analysis.
Discover why MATLAB is an ideal platform for quickly developing financial models and performing risk assessment, asset allocation, and hedge fund strategies.
Ian McKenna, MathWorks
With the explosion of market data volumes and venues, quantitative trading firms face increasing data management complexities in their quest for alpha. The issues that they have to address include data gathering and preparation, model development, back-testing and calibration, integration with existing systems, and reporting and visualization.
This presentation focuses on addressing these challenges with Thomson Reuters and MATLAB products and features a joint use case of MATLAB fed by Thomson Reuters News Analytics Data and Tick History. You will learn how to:
11:10 a.m.–12:30 p.m.
In this session, you will learn about the different tools available for optimization in MATLAB. We demonstrate how you can use Optimization Toolbox™ and Global Optimization Toolbox to solve a wide variety of optimization problems. You will learn best practices for setting up and solving optimization problems, as well as how to speed up optimizations with parallel computing.
Seth DeLand, MathWorks
In this master class, you will learn how teams can create an agile development infrastructure by using advanced modeling, analysis, and development tools in MATLAB to build quantitative applications quickly and then deploying and integrating them into enterprise architectures. The first half of this master class presents a case study on taking an asset allocation model from a prototype to a production web application. You will learn best practices for collaborative development, ensuring code quality, production deployment, and integration with a web server. The second half of this master class demonstrates how MATLAB analytics can be integrated into enterprise environments including databases, Hadoop, distributed computing frameworks, and messaging systems.
In this session, you will learn simple ways to reduce the execution time of computationally intensive MATLAB applications. We demonstrate how you can use Parallel Computing Toolbox™ and MATLAB Distributed Computing Server™ to speed up MATLAB applications by using the desktop and cluster computing hardware you already have. You will learn how minimal programming efforts can speed up your applications on widely available desktop systems equipped with multicore processors and GPUs, and how to continue scaling your speed up with a computer cluster.
We also cover how to automatically generate portable C source code from MATLAB algorithms to accelerate computationally intensive portions of your code.
Yi Wang, MathWorks