Academic Signal #6

1. Train Your Models on P&L, Not Prediction Accuracy

Traditional loss functions like Mean Squared Error (MSE) penalize models equally for all prediction errors. But in trading, a small directional error that loses you 0.5% matters infinitely more than a large magnitude error that still gets the direction right and makes you 3%. Yet we keep training models to minimize prediction errors instead of maximizing trading profits.

The Mean Absolute Directional Loss (MADL) Solution

The researchers created a loss function that directly optimizes for actual trading P&L:

MADL = (1/N) × Σ[(-1) × sign(Actual_Return × Predicted_Return) × |Actual_Return|]

Translation: When your prediction direction is correct, the function rewards you with the actual profit magnitude. When wrong, it penalizes you with the actual loss. The model learns to maximize real trading returns, not minimize prediction errors.

Results: Testing across six assets over 8+ years of out-of-sample data using rigorous walk-forward validation, the MADL function outperformed, especially on a risk-adjusted basis (when accounting for volatility, maximum drawdown, and maximum drawdown duration).

Immediate Action: Test MADL with your existing ML trading models. The mathematical framework is straightforward to implement and could improve performance without architectural changes.

Risk Management: The superior Information Ratios indicate better downside protection – models trained on MADL naturally learn to avoid large losses while capturing profitable moves.

Key Takeaway: Your loss function is your model's objective function. If you're optimizing for forecast accuracy instead of trading profits, you're teaching your model the wrong goal.

Hit “reply” and let us know if you’d like us to dig deeper into this paper or have specific questions.

2. Companies getting the best loan terms are actually more likely to go bankrupt!

New Cornell research analyzing 38 years of syndicated loan data reveals that since 2010, distressed firms secure increasingly favorable amendments – rate cuts, maturity extensions, and looser covenants. Yet these same companies are significantly less likely to recover and more prone to bankruptcy than their pre-crisis counterparts.

What's Really Happening

Post-crisis banking regulations pushed traditional banks to the sidelines, replaced by yield-hungry institutional investors – CLOs, insurance companies, and asset managers. These lenders proved far more generous: institutional investors became 2.15 times more likely to grant borrower-favorable terms after 2010.

Explaining The Paradox

The paradox isn't that good loan terms cause bankruptcy. Rather, it's about what those terms signal in today's market versus historically.

Before 2010: When a distressed company got favorable amendment terms (rate cuts, maturity extensions), it typically meant:

• Banks believed the company had good recovery prospects

• The company was fundamentally sound enough to warrant lender support

• Favorable terms actually predicted higher recovery rates

After 2010: When a distressed company gets favorable terms, it often means:

• Institutional lenders are reaching for yield and being more lenient

• They're giving good terms to fundamentally weaker borrowers than banks used to

• The favorable terms no longer predict recovery—that relationship disappeared

The Quality Problem

The paper shows that post-2010 distressed firms are objectively worse: 90% larger, more leveraged, and with significantly lower credit ratings. Yet these weaker firms are getting better loan terms than stronger firms got before the crisis.

So when you see favorable amendment terms today, you might actually be looking at a weaker borrower than you would have seen getting similar terms historically.

The Investment Insight

It's like grade inflation in schools. An "A" today might represent the same knowledge as a "B" from 20 years ago. Similarly, favorable loan terms today might indicate the same fundamental weakness that would have received unfavorable terms before 2010.

The research suggests investors can't rely on amendment favorability as a positive signal the way they could historically – institutional lenders have essentially lowered their standards while being more generous with terms.

Hit “reply” and let us know if you’d like us to dig deeper into this paper or have specific questions.

3. Price-weighted diversification beats market-cap weighted?

New research from University of Sussex reveals that the price-weighted DJIA achieves superior diversification efficiency compared to the market-cap weighted S&P 500 – challenging fundamental assumptions behind passive investing.

This isn't about absolute diversification (the S&P 500 obviously wins with 16x more stocks), but about how effectively each index uses its stocks to achieve risk reduction.

The paper's simulations show that price-weighting translates into better risk metrics: price-weighted portfolios consistently delivered lower volatility and better Value-at-Risk measures compared to market-cap-weighted portfolios of the same size.

The researchers conclude that since "price-weighted indices tend to be more diversified than market cap-weighted indices… they tend to be less risky." The core insight is that diversification quality (how effectively you use each position) matters more than diversification quantity (total number of holdings).

The Core Problem with Market Cap Weighting

Market-cap-weighted indexes suffer from concentration bias. The top 10 S&P 500 stocks represent about one-third of the index, meaning most of those 500 companies barely contribute to diversification. You're buying a concentrated bet on mega-cap stocks, not true diversification.

Price-weighting creates more balanced exposure by weighting stocks based on price rather than market capitalization.

Crisis Resilience: Price-weighted portfolios showed superior performance during the 2008 financial crisis and COVID-19 – when diversification matters most.

Implementation

For Asset Allocators: Replace market-cap-weighted equity exposure with price-weighted alternatives for efficiency gains that compound over time.

For Risk Management: Use EDP to evaluate existing portfolios. Low diversification efficiency indicates unnecessary concentration risk.

Key Takeaway: Stop confusing diversification with stock count. Price-weighted and equal-weighted strategies may deliver better risk-adjusted returns than the market-cap-weighted products dominating passive investing.

Are you paying for diversification but receiving concentration?

Hit “reply” and let us know if you’d like us to dig deeper into this paper or have specific questions.