Time Series Basics

A time series is a sequence of data points ordered in time: daily closing prices, minute by minute volumes, monthly inflation figures. Almost all market data is time series data. Unlike a random sample, the order matters: today's value may depend on yesterday's, volatility can cluster and patterns can change over time. Time series analysis provides tools to describe, model and forecast this kind of data, and to avoid mistakes that come from treating it as a simple list of numbers.

Components of a time series#

Decomposing a series into these parts helps separate signal from noise.

Prices vs returns#

Prices usually trend and wander without returning to a fixed level, making them non stationary. Returns (percentage or log changes) are much closer to stationary: their average and variance are more stable over time. Most statistical analysis is therefore done on returns, not prices. See Stationarity, Differencing and Unit Roots and Lognormal Distribution.

Key properties to check#

Basic tools#

Time series pitfalls#

• Look ahead bias: using future information to calculate today's values, for example with a centred moving average or a full sample normalisation. See Look-Ahead Bias.
• Overlapping returns: using rolling multi day returns creates artificial autocorrelation.
• Non synchronous data: markets in different time zones close at different times. See Timestamps, Time Zones and Daylight Saving.
• Data snooping: testing many patterns on the same history. See P-Hacking and Multiple Testing.
• Changing regimes: relationships that held in the past may break.

Forecasting realities#

Short term returns of liquid assets are very hard to forecast; they are close to random. Volatility, however, is much more predictable because it clusters. This is why many quantitative models focus on forecasting risk rather than direction, or look for small, persistent edges across many assets. See White Noise and Random Walks.

Frequently asked questions#

What is a time series?#

A sequence of data points recorded in time order, such as daily prices or monthly economic data.

Why do traders analyse returns instead of prices?#

Because returns are closer to stationary, with more stable statistical properties, while prices trend and can create misleading correlations.

What are the components of a time series?#

Trend, seasonality, cycles and random noise.

Next, learn the benchmark for randomness in White Noise and Random Walks.