CFA Level II ExamQuantitative MethodsMedium
A quantitative researcher is performing a backtest on a new trading strategy. The strategy involves daily rebalancing based on signals generated from fundamental data. The researcher notices that the backtest results show exceptionally high, unrealistic returns, far exceeding what would be expected in real-world trading. Upon closer inspection, it is discovered that the strategy's rebalancing decisions are made using financial statements that were published several days *after* the trading day they are used to make decisions. This is an example of which type of backtesting bias?
- AOut-of-sample bias
- BLook-ahead bias
- CSurvivorship bias
- DData snooping bias
Show answer & explanationAnswer & explanation
Correct answer: B. Look-ahead bias
Look-ahead bias occurs when a backtest uses information that would not have been available to a trader at the time the decision was made. Using financial statements published after the trading day they influence is a classic example, as the strategy is 'looking ahead' into future information.
Why the other options are wrong
- A. Out-of-sample bias is not a standard term in backtesting; the more common concept is the need for out-of-sample testing to avoid overfitting, but it's not a bias in the same way as the others.
- C. Survivorship bias occurs when a data set only includes companies that have survived up to the present, leading to an overestimation of historical returns.
- D. Data snooping bias arises when a strategy is developed by repeatedly testing hypotheses on the same dataset until one appears to work, leading to an overestimation of its effectiveness.
Look-Ahead Bias
Look-ahead bias in backtesting occurs when a trading strategy uses information that would not have been available to an investor at the time the investment decision was made, leading to inflated and unrealistic backtest performance.
- Uses future data to make past decisions.
- Common examples: using restated financials, using future prices, using data released with a lag.
- Leads to overoptimistic backtest results.
Memory trick: Backtests can be biased by looking ahead or picking winners.