Bitcoin Burn Addresses: Unveiling the Permanent Losses and Their Underlying Causes

2025-03-18

AI Summary

This work addresses the challenges of identifying Bitcoin burn addresses and accurately estimating permanently lost coins. We propose the first automated detection method based on a multilayer perceptron (MLP), trained on a manually labeled dataset comprising 198,000 addresses and engineered from on-chain behavioral features. Applied to a universe of 1.28 billion addresses, our model identifies 7,905 genuine burn addresses with only 1,767 false positives, confirming 3,197.61 BTC as irrecoverably lost (valued at $295 million as of November 2024). We find that 99% of lost coins concentrate in just three addresses—revealing an extreme power-law distribution and diverse use cases, including Proof-of-Burn token issuance and OLGA Stamps steganography. This study establishes the first large-scale, high-precision burn address identification framework, providing critical empirical grounding for modeling Bitcoin scarcity and assessing its macroeconomic implications.

Technology Category

• Data Mining & Knowledge Management: Data Compression
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Discourse, Pragmatics & Argument Mining

Application Category

• Economics, Online Markets and Human Computation: Economic aspects of blockchain and cryptocurrencies
• Web Mining and Content Analysis: Large pretrained models with web data
• Security and Privacy: Large-scale security measurements

Abstract

Bitcoin burn addresses are addresses where bitcoins can be sent but never retrieved, resulting in the permanent loss of those coins. Given Bitcoin's fixed supply of 21 million coins, understanding the usage and the amount of bitcoins lost in burn addresses is crucial for evaluating their economic impact. However, identifying burn addresses is challenging due to the lack of standardized format or convention. In this paper, we propose a novel methodology for the automatic detection of burn addresses using a multi-layer perceptron model trained on a manually classified dataset of 196,088 regular addresses and 2,082 burn addresses. Our model identified 7,905 true burn addresses from a pool of 1,283,997,050 addresses with only 1,767 false positive. We determined that 3,197.61 bitcoins have been permanently lost, representing only 0.016% of the total supply, yet 295 million USD on November 2024. More than 99% of the lost bitcoins are concentrated in just three addresses. This skewness highlights diverse uses of burn addresses, including token creation via proof-of-burn, storage of plain text messages, or storage of images using the OLGA Stamps protocol.

Problem

• Identify and quantify Bitcoin losses in burn addresses.
• Develop a method to detect burn addresses automatically.
• Analyze economic impact and usage patterns of burn addresses.

Innovation

• Multi-layer perceptron model for burn address detection
• Trained on manually classified dataset of addresses
• Identified 7,905 true burn addresses accurately