AdvancedMultiple choice
A backtest too good to be true · Part 3 of 3
A colleague’s daily stock-selection model uses company fundamentals and shows a Sharpe ratio of 4 from 2005 to 2020. The universe is today’s S&P 500 constituents, and each fundamental is stamped with its fiscal period end date.
With point-in-time data the Sharpe ratio is 1.2. What test best checks for any remaining leak?
- ADelay the features a day and rerun on shuffled labels
- BAdd more features and check whether the Sharpe ratio rises further
- CCheck that the in-sample R² is small, as a leak would inflate it
- DExtend the backtest back to 1995
The worked solution is in Premium
The answer, the full working and the one idea to take away – for this and all 1,322 questions in the bank. Answer it in practice and your working is marked, with a known mistake named when you make one.
Learn the method
More research case studies questions
- Ten observations: nine average 2 and one is 100.Foundation
- A feature is missing for 20% of rows. What do you establish first?Applied
- A series has median 10 and median absolute deviation 2.Applied
- Small companies are much likelier to leave a financial field blank when their…Applied
- Ten features are each missing in 5% of rows, independently.Applied
- You build a factor from quarterly fundamentals downloaded from a vendor today.Advanced