← All tasks
T17 · Volatility Management

Betting Against Beta

Andrea Frazzini and Lasse Heje Pedersen, 2014 · Journal of Financial Economics

43%
pass@1 across 16 systems
13/16
systems with ≥1 successful attempt
5/5
5/5 attempts passed for the best system
80 attempts · 16 systems × 5
Claude Code + Opus 5
80%
Claude Code + Opus 4.8
20%
Claude Code + Opus 4.7
40%
Codex + GPT-5.6-sol
40%
Codex + GPT-5.5
100%
OpenCode + Kimi-K3
100%
OpenCode + Opus 4.7
80%
OpenCode + GPT-5.5
80%
OpenCode + Sonnet 4.6
20%
OpenCode + Qwen3.7-Max
40%
OpenCode + GLM-5.2
20%
OpenCode + DeepSeek V4-Pro
0%
OpenCode + GLM-5.1
0%
OpenCode + Qwen 3.6 Plus
40%
OpenCode + MiniMax-M3
20%
OpenCode + MiniMax M2.7
0%
all checkers pass Type-A clean, Type-B fail fail checking timeout no result / incomplete
Strategy

What the paper asks for.

Leverage-constrained investors over-pay for high-beta assets, so a rank-weighted long-low-beta / short-high-beta cross-sectional portfolio that levers each leg to ex-ante beta=1 (BAB factor) earns a positive risk-adjusted return.

Type
technical
Implementation
pandas
Rebalance
monthly
Long-short
yes
Declared task setup
Data
One-command fetcher for WRDS-sourced task data
Operational scope
The paper-defined methodological core is retained; the fixed instruction instantiates only the data mappings, interfaces, and evaluation conventions needed for execution.
Six-stage quantitative strategy pipeline
  1. 01 Data loading
  2. 02 Feature engineering
  3. 03 Model fitting
  4. 04 Signal generation
  5. 05 Portfolio construction
  6. 06 Execution
Checker suite

19 properties · 80 attempts.

Each row is one requirement the strategy must satisfy. The bar reports the share of observed checker verdicts that pass; missing checker output is not converted into a property verdict.

TYPE A · DOMAIN-LEVEL (9)
  • A1
    no backfill
    stage · data loading
    69/71
  • A2
    no global normalize
    stage · signal gen
    66/71
  • A3
    no future data access
    stage · feature eng
    64/71
  • A4
    causal consistency
    stage · signal gen
    62/71
  • A5
    no survivorship bias
    stage · data loading
    59/71
  • A7
    data frequency alignment
    stage · data loading
    63/71
  • A8
    transaction timing
    stage · execution
    62/71
  • A9
    signal delay test
    stage · signal gen
    63/71
  • A10
    e2e metric consistency
    stage · evaluation
    71/71
TYPE B · PAPER-SPECIFIC (10)
  • B1
    volatility window one year daily log returns
    §§3.2 “We compute rolling estimates of volatilities ... on a one-year rolling window”
    provenance · paper-explicit
    stage · feature eng
    62/71
  • B2
    correlation window five year overlapping three day log returns
    §§3.2 “five-year rolling window ... of three-day overlapping log returns”
    provenance · paper-explicit
    stage · feature eng
    48/71
  • B3
    beta decomposition rho sigma i over sigma m
    §§3.2 “β_TS = ρ · σ_i / σ_m”
    provenance · paper-explicit
    stage · feature eng
    70/71
  • B4
    beta shrinkage w zero point six
    §§3.2 “β̂ = w · β_TS + (1 − w) · β_XS, where w = 0.6 ... we use the prior β_XS = 1”
    provenance · paper-explicit
    stage · feature eng
    71/71
  • B5
    rank weighted portfolio w eq k z minus zbar
    §§3.2 “z is the cross-sectional rank ... z̄ is its cross-sectional average ... k = (z − z̄)”
    provenance · paper-explicit
    stage · signal gen
    67/71
  • B6
    bab return formula levered excess
    §§3.2, Eq. (16) “r_{BAB,t+1} = (1 / β_L) (r_{L,t+1} − r_f) − (1 / β_H) (r_{H,t+1} − r_f)”
    provenance · paper-explicit
    stage · execution
    63/71
  • B7
    monthly rebalance frequency
    §§3.2 “rebalanced every calendar month”
    provenance · paper-explicit
    stage · execution
    63/71
  • B8
    excess returns used for legs
    §§3.2, Eq. (16) “(r_L − r_f) ... (r_H − r_f)”
    provenance · paper-explicit
    stage · execution
    62/71
  • B9
    weights normalized each leg sums one
    §§3.2 “the portfolio weights of stocks in the low- and high-beta portfolios are scaled such that w_L'1 = 1 and w_H'1 = 1”
    provenance · paper-explicit
    stage · signal gen
    69/71
  • B10
    no concurrent market return in beta
    §§3.2 “rolling window of three-day overlapping log returns”
    provenance · paper-explicit
    stage · feature eng
    64/71
What this task reveals

The lowest-passing observed requirement is correlation window five year overlapping three day log returns: 48 of 71 observed checker verdicts pass. The full property list above shows whether errors concentrate in domain-level validity or the paper-specific strategy logic.

Open paper → All tasks
← Previous in Volatility Management
T16 · Volatility-Managed Portfolios
Next in Volatility Management →
T18 · Conditional Volatility Targeting