What if my project is completely unique?

The skill includes a 'Funnel Base Rate' workflow to decompose novel situations into smaller, sequential stages that have existing historical data points.

What is Reference Class Forecasting?

It is a methodology for making predictions by identifying a class of similar past events and using their outcomes as a statistical baseline, also known as taking the 'Outside View.'

Does this skill require specific data files?

The skill provides the framework and search strategies; it helps you find the data through web search, academic studies, and industry reports to set your anchor.

How does this skill prevent bias?

It challenges the 'inside view'—the tendency to think a specific case is unique—by forcing the user to justify any deviation from historical averages with hard evidence.

Can I use this for software development estimates?

Absolutely. It is specifically designed to combat the planning fallacy, helping teams set more realistic deadlines based on how long similar projects actually took in the past.

Reference Class Forecasting

Name: Reference Class Forecasting
Author: lyndonkl

bylyndonkl

•

Data Science & ML

Establishes objective statistical baselines for predictions and projects by anchoring them in historical reality and the 'Outside View.'

The Reference Class Forecasting skill is a specialized tool for decision-making and project management that helps users avoid the 'planning fallacy' and 'inside view' bias. By identifying a class of similar past events and using their statistical frequency as a starting point, it forces an objective assessment before getting distracted by the specific details of a current case. This skill is essential for anyone needing to forecast project timelines, startup success, or complex multi-stage outcomes using a disciplined, data-driven framework that prioritizes historical base rates over optimistic narratives.

Key Features

018 GitHub stars

02Uniqueness audit to challenge 'this time is different' bias and test similarity.

03Guided identification of specific, searchable reference classes for any event.

04Integrated search strategies to find academic and industry success statistics.

05Multi-stage funnel modeling for calculating compound probabilities in sequential processes.

06Reference class validation tests for homogeneity, sample size, and data relevance.

Use Cases

01Assessing the probability of success for new business ventures or product launches.

02Calculating the likelihood of legislative, regulatory, or complex organizational changes.

03Estimating software project completion timelines by using historical industry base rates.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lyndonkl/claude reference-class-forecasting

For use in Claude.ai and ChatGPT

Download Skill