[Back to Intro]

VIDEOS

These videos complement some of the examples discussed in my books and papers.

AUTHORS YEAR TITLE ABSTRACT
Lopez de Prado, Marcos 2023 Causal Factor Investing This seminar reviews the current state of causal confusion in financial econometrics, and proposes solutions with the potential to transform factor investing into a truly scientific discipline.
Roumi, Alex 2021 Asset Allocation - Hierarchical Risk Parity MathWorks' video-tutorial on Hierarchical Risk Parity (HRP).
Fabozzi, Frank 2020 A Conversation with Marcos Lopez de Prado Frank Fabozzi, the Editor of The Journal of Portfolio Management, interviews Marcos Lopez de Prado.
Cesa, Mauro 2020 Lipton and Lopez de Prado on Covid-19 and optimal trading strategies Risk Magazine interviewed Alex Lipton and Marcos Lopez de Prado regarding their extension of the SEIR epidemiological model.
Lopez de Prado, Marcos 2020 Three ML Solutions to the Bias-Variance Dilemma In this Thalesians webinar, I present a framework to justify ML's use of cross-validation, regularization, and ensembles.
Lopez de Prado, Marcos 2020 Three Quant Lessons from COVID-19 In this presentation, I explain three important quant lessons that we can learn from COVID-19.
Himani, Oussama 2020 Interview: The Machine Learning Revolution in Finance A conversation between Oussama Himani and Marcos Lopez de Prado.
Lopez de Prado, Marcos 2020 Bloomberg Quantitative Seminars: Machine Learning Portfolio Construction Machine learning can help overcome the two sources of instability in the estimation of the efficient frontier: (1) noise-induce instability, and (2) signal-induced instability.
Piputri, Rani 2020 CFA Institute: Machine Learning Portfolio Construction Machine learning approaches for the robust estimation of the efficient frontier.
Lopez de Prado, Marcos 2020 Machine Learning for Asset Managers This seminar explains the pitfalls of standard portfolio optimization, and how machine learning can overcome those pitfalls.
Lopez de Prado, Marcos 2020 Cornell University (ORIE 5256): Advances in Financial Machine Learning The current crisis in financial research has been caused by the overfitting of classical methods. By joining us in this course, we hope that you will become part of the solution, and that you will help us modernize finance.
Lopez de Prado, Marcos 2019 Ten Financial Applications of Machine Learning Keynote presentation at the EXDC2019 Conference in New York.
Bloomberg TV 2019 Cornell's Lopez de Prado on How AI is Transforming Finance The financial industry is one of the least automated industries in the United States, says Marcos Lopez de Prado, Professor at Cornell University and CIO at True Positive Technologies. He spoke to Bloomberg's Shery Ahn and Taylor Riggs on "Bloomberg Technology: Global Link" about what needs to be done.
Bloomberg News 2019 Robots in Finance Could Wipe Out Some of Its Highest-Paying Jobs Bloomberg's report on Prof. Lopez de Prado's testimony before Congress on December 6, 2019.
Mathematical Investor 2019 The Impact of AI on Jobs in the Financial Sector Remarks by Prof. Lopez de Prado before the Financial Services Committee of the U.S. House of Representative (December 6, 2019).
U.S. House of Representatives 2019 Robots on Wall Street: The Impact of AI on Capital Markets and Jobs in the Financial Services Industry December 6, 2019 hearing before the Financial Services Committee of the U.S. House of Representatives.
Lopez de Prado, Marcos 2019 DG-FinTech Podcast with Marcos Lopez de Prado A conversation of how machine learning is transforming finance, and how to become part of this revolution.
Lopez de Prado, Marcos 2019 Marcos Lopez de Prado on the Democratization of Alpha Marcos Lopez de Prado, co-founder and CIO of True Positive Technologies, discusses changes in quantitative investing processes, expected changes moving forward, and whether or not alpha will exist in the future.
Lopez de Prado, Marcos 2019 A Conversation at Numerai's 2019 Erasure Conference Numerai's founder Richard Craib speaks with Marcos Lopez de Prado about machine learning and the future of finance.
Lopez de Prado, Marcos 2019  Robust Cross-Section Studies in the Presence of Outliers Cross-sectional studies are particularly sensitive to the presence of outliers. In this experiment, we compare the performance of two regression methods: i) Ordinary Least Squares (OLS): The standard regression method in the industry and in academia ii) Random Sample Consensus (RANSAC): A machine learning method popular in computer vision.
M-A-F-F-I-A 2018 The Reason Most Quantitative Investment Funds Fail An intuitive explanation of how selection bias leads to investment losses.
Lopez de Prado, Marcos 2018 Uncovering Behavioural Biases with Machine Learning Three examples of how machine learning is transforming research in behavioral finance.
Lopez de Prado, Marcos 2018 Ten Financial Applications of Machine Learning A webinar on multiple machine learning applications that are changing the financial industry.
Lopez de Prado, Marcos 2017 The 7 reasons most machine learning funds fail A presentation at CornellTech on common errors made by financial data scientists. A similar presentation at QuantCon 2018.
Lopez de Prado, Marcos 2017 Solutions to the Crisis in Financial Research A presentation at the CFA Institute on why most financial theories are wrong, and what can we do about it.
Lopez de Prado, Marcos 2016 Financial Quantum Computing A presentation at Exponential Finance 2016.
Lopez de Prado, Marcos 2016 Covariance Matrix HRP-Clustering Clustering a covariance matrix allows us to recognize hierarchical structures present in the data. Once clustered, that covariance matrix can be used to derive robust HRP portfolios that significantly outperform mean-variance (Markowitz) style or risk-parity solutions. This video shows how a large, numerically ill-conditioned covariance matrix of changes in yields for 1000+ bonds, becomes quasi-diagonal as the HRP-clustering proceeds.
University of Newcastle (Australia) 2016 The Tenure Maker Simulator This online tools overfits an econometric investment strategy within the parameter ranges specified by the user. It applies the findings published in this paper.
Risk Magazine / Quant Congress USA 2015 Interview with Marcos Lopez de Prado Interview for Risk Magazine, in June of 2015.
Global Derivatives Conference 2015 Interview with Marcos Lopez de Prado Interview for the annual Global Derivatives Conference, in May of 2015.
Wharton, University of Pennsylvania 2015 Illegitimate Science Presentation at the Annual Conference of Wharton's Jacob Levy Equity Management Center for Quantitative Financial Research.
Institutional Investor Journals 2014 Why most published investment strategies are likely to be wrong Institutional Investors invited me to explain the findings published in our recent study, forthcoming in the 40th Anniversary special issue of the Journal of Portfolio Management.
Berkeley Lab 2014 Backtest Overfitting Simulator This online tools overfits a seasonal investment strategy within the parameter ranges specified by the user. It applies the findings published in this paper. Special thanks to Stephanie Ger, who set up this web application and prepared this research poster.
Lopez de Prado, Marcos 2014 How easy is to Overfit a Backtest? We present four examples of overfit strategies. Although the backtests exhibit high performance, the underlying series are unpredictable. This illustrates the fact, unless researchers report the number of trials involved in computing a backtest, its representativeness cannot be determined. The "hold-out" method also fails to prevent overfitting, because that method does not take into account the number of trials carried out. Computing the Probability of Backtest Overfitting (PBO) prevents this phenomenon.
Lopez de Prado, Marcos 2014 Macro Financial Flows Stochastic Flow Diagrams (SFDs) are topological representations of complex dynamic systems. We construct a network of financial instruments and show how SFDs allow researchers to monitor the flow of capital across the financial system. Because our approach is dynamic, it models how and for how long a financial shock propagates through the system. Practical applications include stress-testing of investment portfolios under user-defined scenarios, and the discovery of Macro trading opportunities.
Lopez de Prado, Marcos 2014 Fast Fourier Transform Overfitting In this example, we form a signal as a composite of 5 sinusoidal functions with different frequencies, to which we add white noise. Then we apply a Fast Fourier Transformation (FFT) to recover the signal. At frequency 0, FFT gives as the average value of the time series. As we incorporate additional frequencies, the FFT approaches the signal, and as a result the Ljung-Box statistic (computed on the residuals) drops. At around 55 frequencies, the Ljung-Box statistic reaches a minimum of 504.86. Beyond that point FFT fits more noise than signal, even though FFT continues to converge towards the time series.
Lopez de Prado, Marcos 2013 Rolling WTI Crude Oil Futures This animation illustrates the process of rolling WTI Crude Oil Futures. The blue line represents prices over twelve active futures expirations. The red lines displays their corresponding open interest. The open interest on a futures expiration indicates the number of active wagers at a given date. Contracts nearing expiration (first from the left), as well as June ("M") and December ("Z") contracts are typically the ones with highest open interest. As the expiration date for the front (leftmost) contract approaches, the open interest is transferred to the nearby expirations, in a monthly process called "the roll". This recurrent process generates the predictable harmonic pattern shown in the video.
Lopez de Prado, Marcos 2013 Rolling Henry Hub Natural Gas Futures This animation illustrates the process of rolling Henry Hub Natural Gas Futures (see explanation for the video above).
Lopez de Prado, Marcos 2013 Trading dynamics: WTI Crude Oil vs. RBOB Gasoline A popular misconception is to equate "Big Data" with large data sets. Big Data's defining characteristic is not size but complexity, like hard to interrelate data. A small data set can generate a Big Data problem if it includes variables with hard-to-visualize connections. Here we present a simple example: We display the dynamics of the calendar spreads for two energy commodities, WTI Crude Oil (CL) and RBOB Gasoline (XB) Futures. The series itself is relatively small, a few tens of thousands of data points. However, it combines non-numeric variables (contract symbols), integer variables (days to expiration), real variables (prices) and dynamic systems (term-structure). Standard statistical approaches would not be able to grasp the interrelations that exist between these heterogeneous sources of information.

There is a lead-lag relationship between the WTI and RBOB curves, but it is obscured by the rolldown, seasonal effects, etc. Statistical tests have a hard time detecting this lead-lag, because they look at error-correction patterns between pairs of points. Once we visualize the dynamics between the two curves, it becomes apparent that the lead-lag exists. That is, a lead-lag between curves, rather than between points.

A Big Data solution is to visualize all of these interrelations. Sometimes visualizing a problem can be more insightful than applying fancy mathematical models.

Lopez de Prado, Marcos 2013 A Journey through the "Mathematical Underworld" of Portfolio Optimization A seminar on the Critical-Line Algorithm (CLA) for portfolio optimization. An open-source implementation of CLA in Python has been developed by David H. Bailey and Marcos López de Prado, and is available in the Software section. It is fully documented in this paper. The presentation material used in this seminar can be found here. An example of how Enthought used this class in their own developments is discussed here.
Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen

2011

What is high frequency trading?

Example of actual trades from a High Frequency strategy.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen

2011

The VPIN Flow Toxicity metric and liquidity crashes

Far from being an exception, liquidity crises (such as the 'flash crash') are becoming the norm. Hundreds of times over the last 12 months, financial products have experienced extreme moves as a result of the new market microstructure trends. On August 4, 2011, market makers supported E-mini S&P500 Futures prices for over 20 hours, despite of persistent flow toxicity. As soon as they gave in, an almost 5% selloff unfolded.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen

2011

The Microstructure of the Flash Crash

E-Mini S&P500 Futures intraday returns with readings of order flow toxicity during the 'flash crash' of May 6th 2010.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen
2011 Example of Flow Toxicity in Energy products

In early May 2011, the CFTC reported the largest long speculative position among crude traders in history, who thought that energy prices would rump up, fueled by the violence sweeping through North Africa and the Middle East. Some of these traders decided to take profits on May 5th 2011. The unwinding of such massive positions led them to seek liquidity from uninformed traders. But as these realized that the selling pressure was persistent, they started to withdraw, which in turn increased the concentration of toxic flow in the overall volume. This video shows that by 9.53am CDF(VPIN) crossed the 0.9 threshold, remaining there for the rest of the day. During those few hours, WTI Crude Futures lost over 8%, well in excess of the 'flash crash' loss one year minus a day earlier.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen
2011 Example of Flow Toxicity in Agricultural products

At the end of April 2011, an agricultural-information site owned by the China Meteorological Administration named XN121.com forecasted that rains across northern China would help the planting of corn by boosting soil moisture. About 10 millimeters to 50 millimeters were expected to fall in the Northeast's top corn-growing regions the following day. CDF(VPIN) crossed the 0.9 threshold on April 26th 2011 at 11.44am, and would remain in that area until April 28th 2011 at 12.23am. Corn Futures lost over 6% of its value as a result of that toxic flow.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen
2011 Example of Flow Toxicity in Precious metals

On October 8th 2010 market chatter speculated that further Fed stimulus would be needed. Buying pressure started at around 9am, and by 9.41am CDF(VPIN) had surpassed the 0.9 threshold, which would not touch down until 10 hours later. During that period, Silver Futures rallied 3.5%.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen
2011 Example of Flow Toxicity in Treasuries

Treasuries fell on February 2nd 2011, pushing the 10-year note yield to the highest in nine months, after the U.S. unemployment rate unexpectedly dropped to the lowest level since April 2009, fueling speculation the labor market would improve. U.S. 30-year bond yields also reached a nine-month high as Labor Department data showed the jobless rate fell to 9 percent. Another report showed gains in service industries. The Federal Reserve bought $7.27 billion of Treasuries, less than average, as part of a stimulus program. Consequently, this video shows that order flow toxicity levels reached the 0.9 CDF(VPIN) threshold at 12.42pm, and then remained at elevated values until February 7th 2011 at 8.30am. During that period, T-Notes Futures lost over 1.2% in value.

Easley, David;
Lopez de Prado, Marcos;
O'Hara, Maureen
2011 Example of Flow Toxicity in Currencies

May 20th 2010 is remembered for a sudden reduction in the number of wagers by hedge funds for a decline in the Euro. CFTC data showed speculators holding Euro bearish bets for 107,143 contracts more than those bullish, down from a record 113,890 a week earlier. The toxicity generated by the continuation of that flow can be appreciated in this video. CDF(VPIN) crossed the 0.9 threshold at 10.24am, and remained over that level for the rest of the session. Two and a half hours later began a massive rally that would take the Euro/Dollar Futures up to 3% higher.