← All tasks
T21 · Event-Driven Trading

Asset pricing: A tale of two days

Pavel Savor and Mungo Wilson, 2014 · Journal of Financial Economics

10%
pass@1 across 16 systems
5/16
systems with ≥1 successful attempt
2/5
2/5 attempts passed for the best system
80 attempts · 16 systems × 5
Claude Code + Opus 5
20%
Claude Code + Opus 4.8
40%
Claude Code + Opus 4.7
40%
Codex + GPT-5.6-sol
0%
Codex + GPT-5.5
0%
OpenCode + Kimi-K3
40%
OpenCode + Opus 4.7
20%
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.

On scheduled macro announcement days (a-days: CPI, PPI, EMPLOYMENT, FOMC), high-beta stocks earn substantial premia over low-beta stocks, while non-announcement (n-) days deliver no compensation. Form value- weighted beta deciles from a 252-day rolling CAPM beta (assigned at prior month-end), go LONG D10 / SHORT D1, and HOLD only on a-days (zero-position on n-days).

Type
technical
Implementation
pandas
Rebalance
monthly
Long-short
yes
Declared task setup
Data
One-command fetcher for WRDS-sourced task data
Operational scope
Use a pre-specified executable calendar covering scheduled inflation, employment, and FOMC announcements over the benchmark period.
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
    24/73
  • A2
    no global normalize
    stage · feature eng
    69/73
  • A3
    no future data access
    stage · feature eng
    69/73
  • A4
    causal consistency
    stage · signal gen
    22/73
  • A5
    no survivorship bias
    stage · portfolio
    73/73
  • A7
    data frequency alignment
    stage · data loading
    22/73
  • A8
    transaction timing
    stage · execution
    22/73
  • A9
    signal delay test
    stage · signal gen
    64/73
  • A10
    e2e metric consistency
    29/73
TYPE B · PAPER-SPECIFIC (10)
  • B1
    a day definition three events
    §§1 / Fig. 1 caption “announcement days or a-days (days on which inflation, employment, or Federal Open Market Committee interest rate decisions are scheduled to be announced) and non-announcement days or n-days (all other days).”
    provenance · paper-explicit
    stage · data loading
    73/73
  • B2
    fomc scheduled only no intermeeting
    §§2.1 “The dates for the FOMC scheduled interest rate announcement dates are available from the Federal Reserve website from 1978. Unscheduled FOMC meetings are not included in the sample.”
    provenance · paper-explicit
    stage · data loading
    73/73
  • B3
    beta window 12 months daily
    §§2.1 “in the tables, we compute time-varying betas over rolling one-year windows using daily returns.”
    provenance · paper-explicit
    stage · feature eng
    33/73
  • B4
    decile count ten
    §Table 1 caption “Panels A and B show results for ten portfolios sorted by stock market beta and rebalanced monthly, value-weighted and equal-weighted, respectively.”
    provenance · paper-explicit
    stage · signal gen
    68/73
  • B5
    decile rebalance monthly
    §§2.2 “Table 1 reports results for portfolios sorted on stock market beta, which are rebalanced each month.”
    provenance · paper-explicit
    stage · portfolio
    67/73
  • B6
    within decile weighting value weighted primary
    §Table 1 “Panel A: Ten beta-sorted portfolios (value-weighted)”
    provenance · paper-explicit
    stage · portfolio
    64/73
  • B7
    long top short bottom decile
    §Fig. 1 / §1 “On such days, stock market beta is strongly related to average returns.”
    provenance · paper-inferred
    stage · signal gen
    71/73
  • B8
    hold only on a days
    §Fig. 1 / §1 “stock market beta is economically and statistically significantly related to returns on [announcement days]”
    provenance · paper-inferred
    stage · signal gen
    73/73
  • B9
    universe nyse amex nasdaq common stocks
    §Fig. 1 caption “ten beta-sorted portfolios of all NYSE, Amex, and Nasdaq stocks”
    provenance · paper-explicit
    stage · data loading
    68/73
  • B10
    beta at month end decides next month holdings
    §§2.2 “Table 1 reports results for portfolios sorted on stock market beta, which are rebalanced each month. We estimate betas for each individual stock using one year of daily returns, sort stocks into deciles according to this beta ...”
    provenance · paper-explicit
    stage · portfolio
    64/73
What this task reveals

The lowest-passing observed requirement is causal consistency: 22 of 73 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 Event-Driven Trading →
T22 · The Pre-FOMC Announcement Drift