← All tasks
T07 · Mean Reversion

A new approach to statistical arbitrage: Strategies based on dynamic factor models of prices and their performance

Sergio M. Focardi et al. (3 authors), 2016 · Journal of Banking & Finance

0%
pass@1 across 16 systems
0/16
systems with ≥1 successful attempt
0/5
no system produced a successful attempt
80 attempts · 16 systems × 5
Claude Code + Opus 5
0%
Claude Code + Opus 4.8
0%
Claude Code + Opus 4.7
0%
Codex + GPT-5.6-sol
0%
Codex + GPT-5.5
0%
OpenCode + Kimi-K3
0%
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.

Price-based long-short statistical arbitrage on S&P 500 constituents: extract 15 principal components from log-prices over a rolling 1000-day window, forecast each factor with an independent AR(1), reconstruct forecasted log-prices, and trade the resulting 5-day forecasted-return cross-section by going long the top 10% and short the bottom 10%, rebalancing every 5 days.

Type
statistical_arbitrage
Implementation
pandas
Rebalance
every_5_days
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 (8)
  • A2
    no global normalize
    56/73
  • A3
    no future data access
    69/73
  • A4
    causal consistency
    73/73
  • A5
    no survivorship bias
    66/73
  • A6
    train test temporal order
    67/73
  • A8
    transaction timing
    61/73
  • A9
    signal delay test
    67/73
  • A10
    e2e metric consistency
    1/73
TYPE B · PAPER-SPECIFIC (11)
  • B1
    log price transformation
    §§4.2 Factor models based on prices “Suppose there are N stocks and that p_i(t) = log(P_i(t)) is the logprice of the ith stock at time t.”
    provenance · paper-explicit
    63/73
  • B2
    number of factors
    §§7.1 Number of stationary factors “Anticipating the conclusions of Section 7.1, we will adopt models with a total of 15 factors.”
    provenance · paper-explicit
    73/73
  • B3
    forecast horizon
    §§5.1 Long-short strategies “we use ew = 1000 and fh = 5; that is, every five days we estimate our models on those stock processes that existed in the prior 1000 trading days and forecast five days ahead.”
    provenance · paper-explicit
    67/73
  • B4
    forecasted return formula
    §§4.2 Factor models based on prices “Given forecasted logprices, we obtain forecasted returns: R̂_i(t) = exp(p̂_i(t + fh)) − P_i(t) / P_i(t)”
    provenance · paper-explicit
    72/73
  • B5
    portfolio percentile threshold
    §§5.1 Long-short strategies “We use k = 10%; that is, 'PF_LS_10_10' and 'PF_LS_10_10' strategies are investigated in detail.”
    provenance · paper-explicit
    73/73
  • B6
    equal weight portfolio
    §§5.1 Long-short strategies “Those two strategies allocate equal weights among the forecasted winners and among the forecasted losers. We call them naïve strategies.”
    provenance · paper-explicit
    73/73
  • B7
    long short normalization
    §§5.1 Long-short strategies “To this end, we normalize a portfolio's weights so that the sum of the weights of the long portfolio is equal to 1 and the sum of those of the short portfolio is equal to –1.”
    provenance · paper-explicit
    73/73
  • B8
    pca method for factor extraction
    §§4.2 Factor models based on prices “Using the methodology proposed by Stock and Watson (1988), and analyzed by Escribano and Peña (1994) and Harris (1997), we can represent the observed logprices... If we estimate cointegrating relationships using PCA”
    provenance · paper-explicit
    71/73
  • B9
    factor reconstruction method
    §§4.2 Factor models based on prices “Next, we regress prices on factors in-sample and obtain the coefficients a_i1,...,a_iK. Note that these coefficients can also be obtained from the matrix of the eigenvectors of the covariance matrix of logprices. The final step is to forecast logprices as follows: p̂_i(t + fh) = a_i1*F_1(t + fh) + ... + a_iK*F_K(t + fh).”
    provenance · paper-explicit
    70/73
  • B10
    mean restoration in reconstruction
    §Sections 4.2--4.3 “PCA centering and the subsequent log-price reconstruction.”
    provenance · paper-inferred
    stage · feature eng
    68/73
  • B11
    sharpe calc convention
    §Section 10 “All models were estimated using daily prices. We compute the annualized Sharpe ratio.”
    provenance · paper-inferred
    72/73
What this task reveals

The lowest-passing observed requirement is e2e metric consistency: 1 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
← Previous in Mean Reversion
T06 · The Profitability of Pairs Trading Strategies: Distance, Cointegration and Copula Methods
Next in Mean Reversion →
T08 · Empirical Investigation of an Equity Pairs Trading Strategy