US Large Cap Equities

Market Forecasting & Forecast Analysis Guide

Understand market forecasting methods and forecast analysis techniques, and how to apply forecast market trends to evaluate investment opportunities using financial time series data.

🏢 500+ companies📖 6 forecast methods📅 Trailing 252 days

Daily Forecasts Forecasts Guide Time Series Guide Other Data & Research

A market forecast is a data-driven estimate of a company's future share price — forecast analysis in practice. We believe it's preposterous to think the future can be estimated within a reasonable confidence level, but we'll happily talk about the math behind some of the common forecasting methods. These forecast methods use the trailing year of data, and estimate one day ahead.

This article is paired with the Daily Forecasts Dataset. For the underlying math on stationarity and returns, see our financial time series analysis guide.

Market Forecasting Methods Explained

Linear Forecast

This forecast method is perhaps the simplest. The best way to think about it is to draw a line between today and the first day you're measuring. The next value after today on the line is your forecast. This is done by finding all the information for the linear formula Y = mX + b so that we can solve for Y when X equals X + one day. To find the Slope, we need the closing prices, and the days. In this case, we use the price at Day 0 (in this case, the price one year ago), and the price today. Then to find b, we input the information into the formula with the price today as Y, and solve for b.

Since trading days don't happen every day, and we only have closing data for trading days, we make each day represent a fraction of the whole time period and that fraction becomes a sort of "Day Value". Then, the date (as a day-value) is found, and we estimate for the next trading day's day-value. Then, it's as simple as calculating Y = m(X + 1 Day-Value) + b to find the new Y for the new date.

Future Value Forecast

A Future Value forecast is also a relatively simple one, and it utilizes the Future Value formula FV = PV(1+r)ⁿ with some assumptions. This formula means that the growth rate (100% of the present value + the gain/loss as a percentage) compounded over time, times the present value (e.g. the stock price) equals the future value.

Since we have the stock price, and the time period (n) is known - 1 day, then all that we need is the rate at which the stock is going to grow. For this r value, we use the Median Daily Return, and we calculate for FV in FV = PV(1+r)ⁿ.

For annual estimates, the annualized return is often used, but in our modeling experience we find that Median Daily Return produces more favorable forecast results over Average Daily Return for daily return estimates. The reasoning for this is that median produces a more accurate window into what a "typical day" looks like. Additional details and information can be found on our Market Statistics Guide.

Exponential Triple Smoothing

This method is a bit more advanced. Instead of drawing a single straight line like the linear forecast, Exponential Triple Smoothing incorporates seasonality, recent trends, and smoothed averages, and the primary idea is that recent data matters more than older data.

This is done through the application of moving averages and exponentially weighted averages to smooth time series. In short, it gives an exponentially larger weight to data points that happened recently over data points that happened far in the past. This type of forecast is often used in cyclical industries such as the automobile industry and it is useful to estimate seasonal trends.

Volatility Adjusted GBM & Monte Carlo

Geometric Brownian Motion (GBM) is a widely used model for simulating stock prices because it incorporates randomness in the form of market shocks. It utilizes the volatility and the drift, and it incorporates market shocks through the Wiener Process.

Monte Carlo builds on models like GBM by running many simulations instead of producing just one estimate. Rather than predicting a single future price, it generates a range of possible outcomes based on repeated random sampling and it takes the average of those samples.

To do this, we simulate the price movement a hundred times using the same inputs: current price, expected return, and volatility. Each simulation produces a slightly different path due to the random component, then the average of the results is taken, and that value is the forecast.

Average of Forecasts

This Average is the average of all the forecast methods above.

Keep reading with Starter

Every article is included with the Starter plan: $8 a month, with 100,000 credits for the data tools. Cancel anytime.

Compare plans