Design
Combinatorially symmetric cross-validation across strategy configurations.
sounding open dataReading the waterlineLive sources are loading. Nothing is filled with placeholder market data.
Research validity · methodological
Bailey and coauthors argue that ordinary holdouts can be unreliable after many strategy trials and propose combinatorially symmetric cross-validation to estimate overfitting probability.
Bailey and coauthors argue that ordinary holdouts can be unreliable after many strategy trials and propose combinatorially symmetric cross-validation to estimate overfitting probability.
Bathymark uses this as a testing warning, not as proof that a particular strategy is overfit before its trial history and specification are inspected.
The paper presents a numerical framework for estimating the probability that a selected investment backtest is overfit.
This claim record depends on a reported relationship or method, not a single headline number. No summary metric is manufactured.
Combinatorially symmetric cross-validation across strategy configurations.
In-sample ranking, out-of-sample ranking, and probability of backtest overfitting.
The method requires the tested strategy variants and trial family, not only the winning backtest.
Bathymark stores curated bibliographic facts and its own paraphrase. It does not store the source abstract or full text. Open the original work to inspect the complete analysis.
David H. Bailey, Jonathan M. Borwein, Marcos Lopez de Prado, Qiji Jim Zhu · Journal of Computational Finance · 2017
Journal version of record reviewed. The SSRN identifier remains a public version trail.
Method compliance belongs in each future replication receipt.
Reviewed 2026-07-13; next review 2027-01-13. The paper record is not a recommendation, forecast, or proof of current profitability. Information, not financial advice.