Paper: arXiv 2609.37108
Authors: Thomas Schmelzer
Abstract
In a recent paper, Schmelzer and Hastie argue that Markowitz’s Critical Line Algorithm and the LASSO path trace the same curve. Here we use that identity to compute efficient frontiers with a stock LASSO solver, \texttt{lars_path} from \texttt{scikit-learn}. It handles long–short portfolios under a leverage cap, fixed leverage with varying risk appetite, and the classical long-only, fully invested frontier. Called naively, the last path stops at the maximum-Sharpe portfolio. One shift of the response, by an amount computed in advance, lets a single call reach the minimum-variance portfolio.
Complexity vs Empirical Score
- Math Complexity: 7.5/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: This paper presents a novel and practical application of the LASSO-CLA identity for efficient frontier computation, demonstrating strong empirical results with clear methodology. The mathematical underpinnings are solid, and the code availability significantly enhances reproducibility.
Research Flowchart
flowchart TD
A["Research Goal: Compute Efficient Frontiers"] --> B{"Key Identity: CLA = LASSO Path"};
B --> C["Methodology: Use LASSO Solver (lars_path)"];
C --> D["Inputs: Stock Data, Leverage Constraints"];
D --> E["Computational Process: Call lars_path, Shift Response"];
E --> F["Outcomes: Efficient Frontiers (Long-Short, Fixed Leverage, Long-Only), Max-Sharpe, Min-Variance"];
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes using the identity between Markowitz's Critical Line Algorithm and the LASSO path to compute efficient frontiers with the stock LASSO solver lars_path from scikit-learn, handling long‑short portfolios under a leverage cap, fixed leverage with varying risk appetite, and the classic
Source sentence
Here we use that identity to compute efficient frontiers with a stock LASSO solver, \texttt{lars\_path} from \texttt{scikit-learn}. It handles long--short portfolios under a leverage cap, fixed leverage with varying risk appetite, and the classical long-only, fully invested frontier.
Assumes 3
- Portfolios are subject to a leverage cap when using long‑short positions. model
Source sentence
It handles long--short portfolios under a leverage cap, fixed leverage with varying risk appetite, and the classical long-only, fully invested frontier.
- A fixed leverage (e.g., c = 2) is imposed while varying risk appetite through a tilt λ > 0. model
Source sentence
3 6 Fixed leverage, varying risk appetite A mandate usually fixes the leverage, say c = 2, and varies risk appetite through a tilt λ > 0. Substituting w = λv in min 1 2w⊤Σw −λµ⊤w subject to ∥w∥1 ≤c gives (2) with cap c/λ.
- The classical long‑only fully invested frontier assumes the budget constraint 1ᵀw = 1. model
Source sentence
It handles long--short portfolios under a leverage cap, fixed leverage with varying risk appetite, and the classical long-only, fully invested frontier. Neither end is the minimum-variance portfolio, which needs the budget 1⊤w = 1 that (2) lacks.
Data
- 20‑asset test problem synthetic public
Source sentence
3 The test problem We use the seeded 20-asset example of Figure 1 in [16]: a five-factor covariance Σ = diag(d) + U diag(δ)U⊤, and µ the sample mean of 50 days of returns simulated from it.
- S&P 500 daily returns market-prices public
Source sentence
On daily returns of 494 S&P 500 stocks over 1213 days (July 2021 to May 2026, from Yahoo Finance; the script that fetches them is published with the code), with the centred returns as X, the last corner sits at νlast = −0.178, about fifty times maxj |µj| below zero. The stock returns of Section 8 come from Yahoo Finance and are not redistributed. fetch_sp500.py downloads them: the S&P 500 constituents listed on
Costs
Not applicable
Finds 5
- On the test problem the LASSO path has 21 columns: the empty portfolio, 19 corners each adding one asset, and the end point. theoretical
Source sentence
On the test problem the path has 21 columns: the empty portfolio at the start, 19 corners, each adding one asset, and the end.
- The end point of the path maximises µᵀw − ½wᵀΣw, which is the Kelly portfolio and also maximises the scale‑free Sharpe ratio, pointing towards the maximum‑Sharpe portfolio. theoretical
Source sentence
That end maximises µ⊤w −1 2w⊤Σw, the second-order approximation to expected log growth, so it is the Kelly portfolio [10]. Since Σ−1µ also maximises the scale-free Sharpe ratio, it points towards the maximum-Sharpe portfolio.
- Neither end of the path is the minimum‑variance portfolio because the budget constraint 1ᵀw = 1 is absent. theoretical
Source sentence
Neither end is the minimum-variance portfolio, which needs the budget 1⊤w = 1 that (2) lacks.
- Algorithm 1 returns all 94 corners for the S&P 500 data in 0.14 seconds, faster than a single quadratic program (≈0.12 s) and far quicker than a 100‑point grid (≈12 s) which still misses corners. performance figure from elsewhere in the paper
Source sentence
Algorithm 1 returns all 94 corners in 0.14 seconds, while one QP takes about 0.12 seconds, so a 100-point grid costs about 12 seconds and still misses the corners.
- On the stock data, choosing the margin δ = 1 ensures all corners are found with KKT residuals below 4 × 10⁻¹³. performance
Source sentence
On the stock data δ = 10−4 loses the last corner; from δ = 10−2 to 1 all 94 are found with KKT residuals below 4 × 10−13, and at δ = 100 the residual has grown to 2 × 10−11. The code takes δ = 1.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 18 related papers). Coverage is partial, so this is neither replication nor refutation.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.