The Illusion of Control in Modern Markets ft. Yoav Git | Systematic Investor | Ep.393

Episode Overview

• Explores the foundational mechanics of machine learning classification through the Expectation-Maximization (EM) algorithm and its role in interpreting incomplete data.
• Examines the vulnerabilities and structural challenges facing Commodity Trading Advisors (CTAs) and trend-following strategies during inflationary periods and market crises.
• Investigates the "Excess Volatility Puzzle," revealing how market noise and the mechanical interactions between different types of traders drive price movements far beyond fundamental news.
• Highlights the unpredictable nature of macroeconomic narratives, showing how physical commodity constraints and local political shifts defy traditional forecasting.

Key Concepts

• The Expectation-Maximization (EM) Algorithm: An iterative mathematical and machine learning process used to estimate parameters and classify data when initial labels are hidden or missing. It alternates between calculating expected values and updating parameters to maximize data likelihood.
• CTA Blind Spots During Crises: Systematic trend followers rigidly reduce risk across all positions (both long and short) during market shocks. This can be highly expensive due to widening spreads in illiquid markets, whereas discretionary managers efficiently hedge using highly liquid proxies like US Treasuries.
• The Excess Volatility Puzzle & Market Pendulum: A significant portion of market volatility is "noise" rather than reactions to fundamental news. Markets operate like a pendulum: noise traders provide the initial push, trend followers exacerbate the momentum, and value traders eventually act as gravity to pull prices back to fundamental value.
• Macroeconomic Regime Shifts: Inflation fundamentally alters market behavior, notably turning the traditionally negative correlation between bonds and equities into a positive one, which severely disrupts rules-based trading models.
• Physical Limits of Commodities: Oil volatility is uniquely complex because it directly and indirectly drives inflation, and physical supply chain shortages cannot be resolved simply by deploying more capital.
• The Illusion of Diversification: Adding smaller, highly correlated alternative markets to a CTA portfolio offers diminishing diversification benefits while drastically increasing execution costs, slippage, and the risk of a fund moving the market against its own positions.

Quotes

• At 0:03:04 - "EM stands for expectation and maximization." - Clarifies the meaning of the EM algorithm.
• At 0:03:45 - "Given that you've estimated this distribution you look at each point and say which of these two distribution is it likely... and based on that you recalibrate." - Details the iterative process of the EM algorithm.
• At 0:06:12 - "It ended up being animal welfare and clean drinking water... that were the main two narratives that actually pretty much defined the last two weeks of all the debates." - Highlights the unexpected focus of local political narratives over global events.
• At 0:08:45 - "Positioning initial positioning is so important and you can be lucky you can be unlucky with how you enter a situation like that." - Emphasizes the critical role of timing in market outcomes.
• At 0:13:50 - "In times of inflation we are going to see correlation between bonds and equity being positive rather than negative." - Explains a key shift in market dynamics during inflationary periods.
• At 0:26:27 - "Oil features energy features about 6% into the inflation, directly as a component into inflation... there is a secondary effect on food prices and the rest of manufacturing." - Explains why oil is critical to inflation modeling beyond direct energy costs.
• At 0:27:02 - "In the case of commodities, it's not just about money... for the love of money you might not be able to get the oil that you need." - Highlights that supply shocks in physical commodities are immune to purely financial interventions.
• At 0:30:26 - "When there is a spike in correlations, actually a lot of these assets becomes good substitutes of each other. So hedging with the TY with the US treasuries very quickly is actually as effective as you would do then just shrinking your entire portfolios both the longs and the shorts." - Describes a discretionary risk management tactic pointing out inefficiencies in systematic CTA models during crises.
• At 0:35:46 - "The outcome of this paper... is that most of the market is noise. There is a reason why option pricing and everything works very nicely is because the vast majority of volatility is sort of noise." - Challenges the idea that all volatility is fundamentally driven.
• At 0:39:13 - "The amount of risk that we like to put into each market is limited precisely for the same discussion that we had that we don't want to trade against ourselves." - Explains why CTAs limit exposure in smaller markets to avoid market impact.

Takeaways

• Recognize that traditional asset correlations break down during inflationary regimes; adjust automated trading models to account for bonds and equities moving together.
• During acute market crises with spiking correlations, hedge portfolio risk using highly liquid proxy assets instead of systematically liquidating expensive, illiquid positions.
• Prioritize market liquidity and execution costs over superficial diversification when expanding the number of markets traded in a quantitative strategy.
• Treat physical commodity shortages differently from financial shocks in your forecasting, recognizing that supply chain constraints cannot be bypassed with capital alone.
• Factor structural "noise" into your trading strategies by recognizing that short-term price movements are often mechanically driven by trend followers rather than fundamental value shifts.
• Utilize the Expectation-Maximization framework when dealing with incomplete datasets to iteratively recalibrate your models and improve classification accuracy.