A command-line tool for tracking and scoring personal forecasts. Track your predictions over time and measure your forecasting accuracy using various probability distributions.

Create a virtual environment (unless you want to install globally), activate it, and install the package

python -m venv env
source env/bin/activate
pip install git+https://github.com/magoo/forecast.git

Then run the CLI

forecast

The first run will ask to create a .forecasts directory, if one is not found. Take a look at the examples to get started with .forecast files.

forecast will look for a .forecasts directory wherever it is ran, then show and score all existing .forecast files within. If .forecasts directory does not exist, it will prompt for creation. Please note that the .forecasts directory is plural, and .forecast files are not.

forecast

Filter forecasts by tag:

forecast --tag <tag>

Show all forecasts by type:

forecast --type <type>

The tool supports several types of probabilistic forecasts. PERT, LogNormal, and Pareto are distributions that are built around values you supply.

• Choice: Discrete choices with assigned probabilities
• Interval: Min/max range with confidence level
• PERT: Three-point estimation (min/mode/max)
• LogNormal: 5, 50, and 95% percentiles.
• Pareto: Power law distribution

Create your_filename.forecast files in your .forecasts directory. They look a lot like Frontmatter files, with specific keys for metadata. Example:

---
scenario: "Will it rain tomorrow?"
end_date: 2024-12-31
type: choice
options:
"Yes": 0.7
"No": 0.3
---

Some markdown text describing whether it will rain tomorrow.

Examples for each type of forecast (choice, interval, lognormal, pareto, pert) are in the .forecasts directory in the repo.