Cross-validation — the honest way to fail before you ship
Cross-validation estimates out-of-sample error before deployment, but time-series and grouped data break the i.i.d. assumption behind vanilla k-fold, leaking the future into the past.
Not yet scoped.
Why
The PoC contrasts random k-fold with time-aware splitting on the same forecasting task and shows the optimism that leakage manufactures.
How it works
Not yet built.
Where it lands in Jayverse
Number: use walk-forward splitting, never random k-fold, on time-series readings. Any backtested reading or indicator built on time-series data leaks the future into the past under random k-fold; the published performance number must come from a time-aware split.
DeFi: apply the same split to liquid-staking backtests. Yield or risk-parameter backtests built on historical data need time-aware validation before any parameter is called validated, not just a k-fold score.
Key expressions
Words and phrases from this page worth keeping, with the Korean meaning and the sentence they come from.
Expression
뜻 · 쓰이는 자리
out-of-sample
학습에 쓰이지 않은 데이터(표본 밖)의 · 모델이 처음 보는 데이터에서의 성능을 말할 때. "estimates out-of-sample error before deployment"
break the assumption
전제를 깨뜨리다/위반하다 · 이론의 기본 가정이 실제 데이터에서 성립하지 않을 때. "break the i.i.d. assumption"
leak into
(몰래) 스며들다/새어 들어가다 · 미래 정보가 부당하게 과거 데이터 처리에 섞여 들어갈 때. "leaking the future into the past"
manufacture (optimism)
(인위적으로) 만들어내다/조작하다 · 실제보다 좋아 보이는 결과를 데이터 유출이 만들어낼 때. "the optimism that leakage manufactures"
the honest way to fail
정직하게 실패해 보는 방법 · 실전 배포 전에 미리 제대로 실패를 확인하는 검증법을 말할 때. "The honest way to fail before you ship"
i.i.d.
독립동일분포(Independent and Identically Distributed) · 데이터 포인트가 서로 독립이고 같은 분포를 따른다는 통계적 가정. "break the i.i.d. assumption behind vanilla k-fold"