Johannes Muhle-Karbe: The Hidden Mistakes Quants Make in Market Impact Models

Johannes Muhle-Karbe joins The Sophron® Network to discuss market impact, model misspecification, and why simple trading strategies often perform surprisingly close to complex nonlinear optima. The conversation spans three of his recent research areas: the P&L consequences of using incorrect impact models, Bayesian methods for estimating true out-of-sample Sharpe ratios, and the efficiency of linear approximations in markets with power-law impact.Johannes Muhle-Karbe is Head of Mathematical Finance at Imperial College London and Director of the CFM-Imperial Institute of Quantitative Finance. His research sits at the intersection of mathematical finance and real-world market practice, with work spanning market microstructure, liquidity, price impact, transaction costs, and optimal execution. Prior to joining Imperial, he held faculty positions at Carnegie Mellon University, the University of Michigan, and ETH Zürich. His research has made him a widely recognized figure in quantitative finance, particularly for his contributions to understanding how trading strategies perform in markets shaped by frictions, uncertainty, and complex dynamics.We examine why market impact is the dominant trading friction for large systematic funds and how misspecifying impact models can destroy profitability — with the key insight that overestimating liquidity is far more dangerous than underestimating it. Follow Johannes Muhle-Karbe on LinkedIn: linkedin.com/in/johannes-muhle-karbe-77428a105Core Timestamps04:25 – Introduction to Johannes Muhle-Karbe and his background06:17 – What is market impact and why it matters for systematic trading07:14 – Why impact models are always wrong and how much that matters09:32 – Why you can't just backtest without a model10:25 – Do firms use academic research on misspecification directly?13:05 – When does impact misspecification matter most?15:03 – The asymmetry of P&L: overestimating liquidity vs. underestimating it20:27 – Dimensionality challenges in fitting impact models22:14 – Bayesian estimation of Sharpe ratios and the complexity haircut23:25 – Why the 50% haircut rule is too simplistic25:51 – What happens when you add more predictors and complexity31:05 – How signal autocorrelation affects the framework35:12 – Linear approximations for power-law impact37:55 – Simple parametric strategies vs. machine learning optimization39:07 – Why statistics matter more than optimization42:02 – Is there an optimal execution algorithm?44:27 – How much is still unknown in market impact research47:15 – Current research: a unified theory connecting market regularities50:47 – Will AI and LLMs fundamentally change quantitative finance?Main Topics Covered• Market impact as the dominant friction for systematic trading strategies• The asymmetric P&L consequences of impact model misspecification• Why overestimating liquidity is far more dangerous than being too conservative• Bayesian Sharpe ratio estimation and the complexity haircut• How model complexity widens the gap between in-sample and out-of-sample performance• The surprising efficiency of simple linear strategies under nonlinear impact• Why statistical estimation matters more than optimization sophistication• Connecting square-root impact, rough volatility, and order flow regularities• The need for better industry–academia data sharing in market microstructure• The uncertain but transformative potential of AI in quantitative financeConnect With UsInstagram: / amsterdaminvestLinkedIn: / amsterdam-investment-clubX: https://x.com/amsterdaminvestSubscribe for more conversations at the intersection of markets, research, and technology.