← All tasks
T26 · Systematic Allocation

Shrinking the Cross-Section

Serhiy Kozak et al. (3 authors), 2020 · Journal of Financial Economics

5%
pass@1 across 16 systems
3/16
systems with ≥1 successful attempt
2/5
2/5 attempts passed for the best system
80 attempts · 16 systems × 5
Claude Code + Opus 5
40%
Claude Code + Opus 4.8
0%
Claude Code + Opus 4.7
0%
Codex + GPT-5.6-sol
20%
Codex + GPT-5.5
0%
OpenCode + Kimi-K3
20%
OpenCode + Opus 4.7
0%
OpenCode + GPT-5.5
0%
OpenCode + Sonnet 4.6
0%
OpenCode + Qwen3.7-Max
0%
OpenCode + GLM-5.2
0%
OpenCode + DeepSeek V4-Pro
0%
OpenCode + GLM-5.1
0%
OpenCode + Qwen 3.6 Plus
0%
OpenCode + MiniMax-M3
0%
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.

Estimate an eigenvalue-shrunk stochastic discount factor on 50 anomaly-managed portfolio returns, then hold the resulting SDF-implied maximum-Sharpe portfolio out of sample with market-volatility targeting.

Type
fundamental
Implementation
pandas
Rebalance
monthly
Long-short
yes
Declared task setup
Data
Bundled public 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

17 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 (7)
  • A2
    no global normalize
    stage · feature eng
    72/72
  • A3
    no future data access
    stage · model fit
    70/72
  • A4
    causal consistency
    stage · model fit
    71/72
  • A6
    train test temporal order
    stage · model fit
    71/72
  • A7
    data frequency alignment
    stage · data loading
    72/72
  • A9
    signal delay test
    stage · signal gen
    71/72
  • A10
    e2e metric consistency
    7/72
TYPE B · PAPER-SPECIFIC (10)
  • B1
    universe size 50
    §§4.2.1 “We now turn to our primary data set of 50 portfolios based on anomaly characteristics.”
    provenance · paper-explicit
    71/72
  • B2
    train cutoff 2004 12
    §§5 “we conduct the entire estimation, including the choice of penalty, based on data until the end of 2004.”
    provenance · paper-explicit
    68/72
  • B3
    objective ridge in hj metric
    §§3.1.2 Eq. (27) “b̂ = arg min_b {(μ̄ − Σb)' Σ^{-1} (μ̄ − Σb) + γ b'b}”
    provenance · paper-explicit
    51/72
  • B4
    cv folds k equals 3
    §§3.3 / §5 “we choose penalties by three-fold cross-validation (two-fold if interactions are included) but with shorter blocks because we only use the pre-2005 sample here.”
    provenance · paper-explicit
    68/72
  • B5
    cv objective cross sectional oos r2
    §§3.3 “yielding the cross-validated R²_{oos}. Finally, we choose γ (or γ_1, γ_2) to maximize the cross-validated R²_{oos}.”
    provenance · paper-explicit
    50/72
  • B6
    mve portfolio p t formula
    §§5 “we construct the time series of the implied MVE portfolio P_t = b̂' F_t in the 2005-2017 OOS period.”
    provenance · paper-explicit
    71/72
  • B7
    oos volatility match market
    §§5, Table 4 caption “MVE portfolio returns are normalized to have the same standard deviation as the aggregate market.”
    provenance · paper-explicit
    70/72
  • B8
    gamma log spaced grid
    §§4.2.1 Fig. 3, §4.2.1 text “the OOS R² is maximized for κ ≈ 0.30.”
    provenance · paper-explicit
    70/72
  • B9
    cv r2 exact hj metric formulation
    §§3.3 Eq. (30) “R²_{oos} = 1 − (μ̄_2 − Σ_2 b̂_1)' Σ_2^{-1} (μ̄_2 − Σ_2 b̂_1) / ( μ̄_2' Σ_2^{-1} μ̄_2 )”
    provenance · paper-explicit
    43/72
  • B10
    gamma from kappa via tau squared scaling
    §§3.1 Eq. (19)-(20) & §4.2.1 “κ² ≡ tr( p' p ) / τ², ... γ = τ²/T · (1/κ²) ... yields the OOS R² maximized for κ ≈ 0.30.”
    provenance · paper-explicit
    70/72
What this task reveals

The lowest-passing observed requirement is e2e metric consistency: 7 of 72 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
Next in Systematic Allocation →
T27 · Out-of-Sample Equity Premium Prediction: Combination Forecasts and Links to the Real Economy