Spring 2024 Recap
Dear SIF Community,
The SIF Quant Team is proud to complete the 2023-2024 school year with progress made on many meaningful projects alongside a series of enriching lectures. We are excited to share these updates with you regarding our work, education, and future.
Projects
The Spring semester brought along the progress of several ongoing projects as well as the start of new projects by our first-year members. Working both independently and collaboratively, members emerged with significant findings across a variety of topics.
Sumit, Ravi, and James explored the applications of reinforcement learning (RL) for online equity trading strategies. After studying the literature on trading reward functions, RL learning algorithms, transaction cost accounting, and regularized policy structure, they conducted experiments on various RL systems using price and alternative data sources, including news headlines and SEC filings processed with sentiment and topical analysis. Their custom strategy beat the S&P500 and common trading benchmarks out-of-sample.
Richard looked at mechanics of economic dispatch in operations of power grids along with the relationship with power prices. He built a toy economic dispatch model using pyomo and investigated future extensions to cover unit commitment problems.
Kaushik, Julia, and Eileen focused on wrapping up the development and deployment of SIFSearch, the internal search engine of SIF. They were able to significantly enhance the application’s general user experience and prepare it for a smooth deployment via Digital Ocean.
Rohan created a fully-featured alpha and forward testing platform for the cryptocurrency markets. Cryptocurrency exchanges provide the unique asset of openness, where real-time, high-fidelity data is widely available. Rohan’s project establishes the ability to define a crypto trading strategy in code and test it against real-time market data on the Binance US Exchange.
Darius, Vijay, Eckart, Krisha, and Phillip finalized a modular alpha backtesting infrastructure, implemented a data ingestion pipeline on AWS, and chartered research into orderbook generation and live-trading integration. Next semester, the team will continue efforts in live-trading on Alpaca and spearhead a subproject towards building an interactive market simulation with dynamic price movements induced by agentic or generative approaches.
Rajat, Parth, and Raymond constructed an orderbook for cryptocurrencies that provides a new indicator for calculating the price of a coin which may differ from a simple midpoint of the spread. Cryptocurrencies are still a relatively new investment that allows retail traders to interact directly with an orderbook. Their research aimed to create buy signals that capitalize on any potential indicator of momentum in the difference between a volume-weighted average price and the midpoint price.
Ocarina developed an alpha based on the Discounted Cash Flow model, a fundamental valuation model for companies. The DCF strategy estimates the present value of an investment using its expected future cash flows. The alpha incorporates the steps of summing the company’s free cash flow for each year, calculating the net present value based on the discount rate, and determining the company’s terminal value using the Gordon Growth Method.
Ankit created an LSTM (Long-Short Term Memory) model which predicted the change in stock prices based on previous data. Different implementation strategies were tested (inspired by previous research), including the use of multiple LSTM models together and using an additional attention parameter, which yielded 35% cumulative returns on average over multiple time periods.
Spring Lecture Series
The Quant Team participated in a series of engaging presentations this semester led by fellow members of the team and reputable professors at the UMD Smith School of Business.
Member Lectures
Several senior members volunteered to prepare lectures relating to topics of their choice, giving members deeper academic viewpoints in the broader field of quantitative finance. We were impressed by the high level of knowledge presented by these senior members and their ability to spark inspiration across the team.
Sumit Nawathe presented options theory across two lectures. In his first lecture, he illustrated the binomial model through examples that proved several key principles regarding Martingale and Markov properties, which provided an overview of using risk-neutral expectations to price options. In the second lecture, Sumit introduced Geometric Brownian Motion, Ito's Lemma, and walked through a stochastic calculus derivation of the Black-Scholes formula, concluding with some comments on the option greeks.
Sanjeev Devarajan discussed the aspects of building a low-latency trading system and optimizing the hot-path. His presentation covered a high level overview of the different components in a system as well as a detailed explanation of the computer systems topics needed to optimize these components effectively.
Richard Zha introduced the basics of probability theory using random variables and their properties in his Statistics lecture. He also presented how probability is used in hypothesis testing and regression analysis.
Phillip Guo presented ideas of how prediction markets can aggregate information in the form of measurable outcomes. His market making lecture walked through a few benefits of them and an example with a prediction market.
Rohan Uttamsingh gave an introductory lecture on practical machine learning based on his prior experience conducting deep learning research on campus. This lecture included an introduction to machine learning from first principles with an overview of the required theory and math and a deeper dive into the inductive biases of different model architectures to understand their strengths and limitations.
Professor Lectures
Professor Albert “Pete” Kyle is the Charles E. Smith Chair Professor of Finance at the Smith School of Business. In April, members of the Quant team learned about the research he conducted at Wharton regarding market microstructure invariance. Professor Kyle presented ideas relating to how market liquidity can be developed as a “metamodel”, a set of reduced-form log-linear equations which define related liquidity variables.
Professor Seokwoo Lee is a visiting assistant professor of finance at the Smith School of Business, where he conducts research on financial institutions and contracting using machine learning methods. During his lecture to the Quant team in March, Professor Lee provided a comprehensive overview of his current research in constrained institutional investors and asset pricing anomalies.
Reading Group Presentations
This semester, we challenged first-year members to apply their knowledge through “Reading Group Presentations.” In small groups, members chose a research paper of their choice and conducted an analysis on the literature and code, presenting their findings with the supplement of their unique understanding and perspective. We believe that this gave new members a distinct opportunity to dive deeper into a topic of their interest in the field of quant and diversify the entire team’s knowledge base, and we are excited to continue this learning opportunity for future first-year members.
Ankit, Rajat, and Parth analyzed a paper that classified 30x30 images of candle data and as either a BUY, SELL, or HOLD signal. This data was then used to train a CNN that could take a given image and return a signal with the appropriate aligned action.
Raymond and Shrinav read Red Rock Capital’s paper on Sortino Ratios. They found that the Sortino Ratio focuses on downside volatility by modeling the Sharpe Ratio and compared the two by creating simple alphas combined with the alphas from the fall semester’s Alpha Competition. Although they found that the Sortino Ratio gave a more accurate representation of portfolio performance for skewed data, it is still a method of historic analysis that does not necessarily indicate future performance.
Anning and Eileen presented a paper on the bond yield curve convexity trading and investigated why the shape of the yield curve was concave. They constructed an arbitrage portfolio under the assumption that the yield curve moves in parallel, showing that the zero coupon bond yield curve and swap curves should be concavely shaped.
Eckart looked into Variational Quantum Eigensolver (VQE), a quantum algorithm that can be used for constrained optimization problems like Portfolio Optimization. From their analysis and implementation using Qiskit, they found that although VQE yields accurate answers, current classical approaches still outperform VQE in terms of both computational time and accuracy.
Competitions
Our members continually seek to be intellectually challenged by finding opportunities outside of SIF. These opportunities most often take the form of competitions relating to quant, trading, and math.
Estimathon is an annual, team-based math contest hosted by Jane Street. This year, it was held at UMD to test participants’ abilities of predicting the answer to a Fermi question and valuing their level of uncertainty. Sanjeev and Aaquib worked in a team of four to tackle about 10 problems, where they ended up winning the competition in First place.
Prosperity, presented by IMC Trading, challenges teams from across the globe to develop a prevailing trading strategy. This semester, James, Darius, Omkar, Rohan, and Richard formed a team and had the opportunity to learn more about market making, market taking, and high frequency trading.
Conclusion
The SIF Quant Team is proud to close out the 2023-2024 school year with continued progress on impressive projects, interesting lectures relating to a number of different topics in quant, and the incorporation of contributions made by our class of new members.
We would like to congratulate Rohan Uttamsingh, Kaushik Vejju, and Sumit Nawathe for graduating this semester.
Rohan made remarkable contributions to crypto research in SIF and served as a supportive mentor and friend to many. After serving as Co-President and Senior Leadership Advisor during his time in SIF, he is returning to Stripe in NYC as a software engineer later this year.
Kaushik was a pioneer in the SIF Search project, leading the platform from development to deployment and significantly establishing the club’s infrastructure. He recently started his role as a technology strategy engineer at Appian in McLean.
Sumit was a knowledgeable teacher to members of the club, leading several lectures in options, the Markowitz model, and statistics during his time in SIF. After serving as President this past year, he will be pursuing his M.S. in Computer Science through UMD’s Plus One Program.
We have been pursuing efforts to keep our SIF alum in the loop and hope to establish a wide network of these talented individuals. We will certainly miss Rohan, Kaushik, and Sumit and wish them the best in their bright futures.
Next semester, we have an exciting schedule planned with lectures led by reputable UMD professors, senior members, and first-year members. James Zhang and Ocarina Lin now serve as Co-Presidents with Sumit Nawathe as Senior Leadership Advisor.
Thank you once again to our members, alumni, and readers for your continued support of the Smith Investment Club. We look forward to the many exciting opportunities that await us in the future.
Best,
SIF Quant Executive Team
