Chris Stewart IS 201: Power Computing Spring 2012 -- Professor Brown Draft No. 2 – February 17, 2012 Logic Model for Statistical Model/Analysis of Historic S&P 500 Price Movements Problem Statement: Recent research has explored the use of trading algorithms (in the form of pre-programmed buy/sell orders resulting from different changes in stock price or other circumstances) as a means of generating profit through various stock/commodity trades. Building on these concepts, I would like to see if it is possible to conduct a statistical analysis of certain historic stock/commodity prices (starting with the S&P 500) in order to: 1. Identify key trends in historic prices to develop a (hopefully) profitable automatic trading strategy 2. Implement this strategy on a going-forward basis in a given data set to determine if it does in fact generate profits As such, this project has the following goals: To develop a statistical model to test the profitability of different trading algorithms based on historic stock/commodity prices To test these trading algorithms through a computer program which executes the trading strategy on a given data set, which ideally would be imported into the HiPerCiC framework and analyzed through the R statistical software program To develop a computer program which generates ideal trading parameters (i.e., buy/sell price and volume) based on a historic analysis of stock/commodity prices [The remainder of this page has been intentionally left blank.] Resources: Chris Stewart (primary student) Other Power Computing students interested in the underlying topic (statistical analysis of historic stock prices) or the design and implementation of the corresponding software Statistics students in the Center for Inter-Disciplinary Research Professors Brown, Goedde, and possibly another professor specializing in statistics The experience and time that the dedicated student(s) and faculty members invest in the project HiPerCiC Project Assistants (Mary and Andy) HiPerCiC resources (equipment and software) Data on historic stock/commodity prices Prior research into algorithmic trading models, statistical analyses of stock/commodity prices, etc. Activities: Collecting data for analysis (e.g., historical stock prices of S&P 500, DJIA, NASDAQ, and individual stocks and/or commodities as needed) Conducting research to formulate a statistical model and/or trading strategy (e.g., appropriate time period for "look-backs", other variables to consider, etc.) Developing statistical model and/or trading strategy (e.g., P = f(P5, P20, P60), P = f(P5Avg, P20Avg, P60Avg), etc.) Generating new data set(s) for statistical analysis (e.g., to include P5 = Price 5 days ago, P5Avg = Average price over past 5 days, etc.) Running regression(s) on the new data set(s) to test the model/strategy Revising the model/strategy as needed (based on the results of prior regressions, additional research, new ideas from project members, professors, etc.) Creating a computer program to test the model/strategy (e.g., buy/sell X% of stock based on Y% price dec/inc over t5, t20, t60, etc.) Creating a computer program to formulate a profitable trading model/strategy (e.g., multiple regression analyses to determine X% of stock to buy/sell based on historic price trends) Testing and revising the computer program as needed to incorporate new trading models/strategies or fix any bugs in the code Using the HiPerCiC framework to run (i) the R statistical software program and/or (ii) multiple statistical analyses of the underlying data using R Ongoing meetings with the dedicated student(s), faculty members, and others to discuss status of research, results of statistical analyses, formulation of trading model(s), development of computer program, and any additional tasks that need to be completed Outputs: We anticipate two types of outputs from this project: deliverables; and the people we reach. Deliverables: Formal research paper surveying prior literature, formulating a trading theory/model, and analyzing results A statistical analysis of the trading model/strategy A computer program generating the results of the trading model/strategy given historic stock/commodity prices People Reached: Students in IS 201: Power Computing and relevant St. Olaf faculty (e.g., Professors Brown, Goedde, and possibly others) Additional students and academics in the fields of statistical modeling, economics, finance, and/or computer science (or anyone interested in those fields or the underlying topic) Financial analysts, investors, stock/commodity traders, corporate executives, etc. Outcomes: At present, we see the following short-, mid-, and long-term outcomes associated with the project. Short-Term: Greater knowledge of different algorithmic trading models and the impact of different variables on current stock/commodity price Increased skill in generating and testing statistical models to test the relationship between current and historic prices Enhanced awareness of C++ and/or other programming languages to generate an appropriate computer program to test possible models/theories Greater awareness of the advances that power computing can make in this field Mid-Term: Possible changes in trading behavior based on the results of the model Increased profits (assuming a successful model) Greater awareness of the trading model (assuming results are successful and circulated or published), which would likely generate further critique(s) of such a strategy Possible revisions to existing trading models Long-Term: More profitable stock/commodity trades Improved economic conditions of those engaging in such trades Reduction of risky or poor/emotional trading decisions