Log return approximation
For small moves, the log return and the simple return are almost the same. For large moves they part ways, and the log return is always the smaller of the two.
01The formula
- the simple return, so +10% is 0.10
- the log return: the continuously compounded return
- the second-order correction, which matters once moves get large
02See it move
03Where it is used
- Quant researcherUses log returns so that returns over many days simply add up, then converts back to simple returns for reporting.
- Risk managerKnows that for daily moves of 1–2% the two are interchangeable, but for crash days the difference is material.
- Options deskModels log returns as normal, which is the assumption behind Black–Scholes.
04Why traders use it
Log returns add over time, are symmetric (a log return of −0.1 undoes +0.1) and suit statistical models. The approximation tells you when you can swap one for the other in your head.
05Worked example
A 10% simple return is a log return of 9.53%; the shortcut ln(1 + x) ≈ x is off by 0.47%. Over 5 periods in a row the log returns add to 47.66%, while the total simple return is 61.05%.
06How it is built
- Expand as a Taylor series around 0:
- For small the higher terms are tiny, so .
- The first correction, , is always negative, which is why log returns sit below simple returns.
- Over periods, : log returns add, simple returns compound.
07The common mistake
Adding simple returns across days, or averaging log returns and reporting them as simple returns. For volatile assets the gap, roughly half the variance, can be several percent a year.
08In Python and LaTeX
import numpy as np x, n = 0.10, 5 print(round(np.log1p(x), 4), round(x - np.log1p(x), 4), round(n * np.log1p(x), 4), round((1 + x) ** n - 1, 4)) # 0.0953 0.0047 0.4766 0.6105
\ln(1 + x) \approx x - \frac{x^2}{2} \approx x \quad \text{for small } x09Where it fits
Rather not code the maths yourself? Describe your trading idea and let AI build the bot.
Build a bot with AgenKitThe worked example and the model run the same code, so every number matches. Models simplify markets on purpose. This is education, not financial advice.