Simple and log returns
Two ways to measure the same price move. Simple returns add up across assets in a portfolio; log returns add up across time.
01The formula
- the price at the end of period t
- the simple return: the percentage change in price
- the log return: the natural log of the price ratio
- the log return over the whole stretch from period 1 to T
02See it move
03Where it is used
- Quant researcherWorks in log returns for time-series models, because a multi-day return is just the sum of daily ones.
- Portfolio managerUses simple returns to combine holdings, because a portfolio's simple return is the weighted average of its assets' simple returns.
- Retail traderLearns why a 50% loss needs a 100% gain to recover: the log returns, not the simple ones, cancel.
04Why traders use it
Mixing them up gives wrong answers in both directions: adding simple returns over time overstates growth, and averaging log returns across assets understates a portfolio. Knowing which one adds where avoids both.
05Worked example
A day of 20.0% is a log return of 18.23%. Follow it with a day of −20.0%: the simple returns add to 0.00%, but the two-day return is really −4.00%. Adding the log returns and converting back gives −4.00%, the right answer.
06How it is built
- Over two days the price ratio multiplies: .
- Take logs: . Log returns add over time.
- For small moves , so the two are close for daily returns and drift apart for big moves.
- Convert back with .
07The common mistake
Averaging simple returns over time to get a growth rate. Gains of +20% and −20% average to zero, yet you are down; the average log return tells the truth.
08In Python and LaTeX
import numpy as np r1, r2 = 0.2, -0.2 print(round(np.log(1 + r1), 4), round((1 + r1) * (1 + r2) - 1, 4)) # 0.1823 -0.04
R_t = \frac{P_t}{P_{t-1}} - 1, \qquad r_t = \ln\frac{P_t}{P_{t-1}} = \ln(1 + R_t), \qquad r_{1 \to T} = \sum_{t=1}^{T} r_t09Where 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.