Financial markets are governed by a combination of rational and irrational forces, statistical probability, and “animal spirits.” Understanding the market, much less winning in it, requires fluency in both. However, market participants, including asset traders, now frequently use machine learning techniques to generate predictions of future asset prices.
scholars such as Bo FuThe assistant professor of finance at George Mason University’s Costello College of Business is studying how these machine learning tools are changing the decision-making processes that drive markets, for better or worse.
The subject of President Hu’s recent paper is business administration This is a well-known machine learning technique called LASSO (Least Absolute Shrinkage and Selection Operator), which was introduced by statistician Robert Tibshirani in 1996 and has since been widely adopted by financial practitioners.
“If you look at it, Base paper“This describes an approach created by adding a regularization penalty to the least squares regression method,” says Hu. A LASSO type trading strategy includes an “inactive zone” for small-scale activities, where the trading strategy will do nothing. ”
The paper was co-authored by Wen Chen of Texas Tech University and Liang Yang of the University of Toronto.
Despite LASSO’s popularity and influence, the soundness of its economic rationale remains unclear. Traders are probably looking for every edge, no matter how small, in pursuit of outsized profits. How does it make sense to employ a system designed to relegate small signals to an “inactive zone” where they are ignored?
To address this question, researchers developed a theoretical framework that models financial markets where multiple agents (i.e., traders) use the price history of assets to predict returns and make trading decisions.
In the case of benchmarks, when traders understand the trading environment and do not face model uncertainty, they act according to an alternative to LASSO known as MSE (Mean Squared Error). “MSE is essentially a Bayesian learning approach based on economic rationality,” Hu says. “This means that rational agents use Bayesian learning to update their beliefs and design trading strategies. This is in stark contrast to LASSO estimation, which filters out weak signals.”
However, the researchers found that when traders faced significant ambiguity regarding the distribution of asset values, their trading calculations changed. Ambiguity-averse agents employ strong LASSO-like strategies and refrain from trading in response to weak or intermediate market signals. Because linear constraints were imposed on the allowed trading strategies, the equilibrium decision matched exactly the LASSO estimate.
As an equilibrium trading strategy, LASSO can improve total profits compared to edge-seeking Bayesian alternatives in the benchmark case because more conservative positions determined by “inactive zones” moderate competition among traders. In large markets, intense competition drives the total profit for traders using traditional MSE strategies close to zero. LASSO traders may be able to maintain positive total profits by trading less aggressively. This is a mechanism that researchers describe as “tacit collusion,” even though there is no communication or explicit coordination between traders.
However, Hu emphasizes that the usefulness of LASSO as a hedge against ambiguity depends on how closely traders’ beliefs match the market’s true risk distribution. Therefore, the profitability of a LASSO strategy depends on the balance between trader bias and strengthening market power (due to LASSO conservatism).
This balance is especially important when traders need to distinguish between passing fads and permanent trends. “Recently, the semiconductor index has experienced a record-setting rally. This is very strong momentum, but if you’re a contrarian, you might want to bet against the trend,” Hu said. “Traders must decide whether to follow the trend or take a contrarian position. LASSO’s inactive zone helps prevent overreaction to weak evidence, but it can also delay action if the emerging trend is real.”
LASSO strategies can also increase market volatility when combined with other objectives and constraints of market makers, including high-frequency traders. “These traders need to manage their inventory,” Hu said. “If they follow a LASSO-type strategy with inactive zones, they will respond to the market’s liquidity demands until their inventory reaches a certain threshold. At that point, they can start trading like momentum traders and cause liquidity depletion in financial markets. This dynamic is part of what fuels events like the 2010 ‘flash crash.'”
