NFT & DeFi Bubble Detection Research

Detecting price bubble formation using SADF, GSADF, and LPPLS frameworks

Research Paper

Detecting Price Bubble Formation in NFT and DeFi Markets

In Progress

Lennart J. Baals [ORCID]

Based on: Wang et al. (2022), Journal of Chinese Economic and Business Studies

Digital asset markets, including those for non-fungible tokens (NFTs) and decentralized finance (DeFi) instruments, have expanded rapidly and are commonly viewed as highly speculative. This research examines the extent and timing of price bubbles in these markets using daily USD prices for two capitalization-weighted benchmarks (NFT Index and DeFi Pulse Index), alongside major infrastructure coins and DeFi-related tokens. We apply SADF and GSADF tests to detect and date-stamp mildly explosive episodes, with LPPLS framework as robustness check.

WP2 [PDF]
NFT DeFi Bubble Detection SADF GSADF LPPLS Cryptocurrency
Key Findings
Recurrent Bubble Regimes

All assets exhibit recurrent bubble regimes with heterogeneous frequency and intensity. ETH shows highest bubble coverage (19.53% SADF), while DPI spends more time in bubble states than NFTI.

Three Boom-Bust Phases

Bubble episodes cluster around: (1) DeFi Summer 2020, (2) Crypto/DeFi boom early 2021, and (3) NFT-centered bull run late 2021. Co-occurrence patterns show synchronized exuberance.

Liquidity Association

Bubble regimes systematically associated with elevated trading activity. Joint bubble days display pronounced increases in log trade volume relative to non-bubble periods.

Assets Analyzed (10)
NFTI

NFT Index

DPI

DeFi Pulse Index

ETH

Ethereum

BNB

Binance Coin

SOL

Solana

ADA

Cardano

UNI

Uniswap

AVAX

Avalanche

WBTC

Wrapped Bitcoin

LINK

Chainlink

Methodology
SADF Test

Supremum Augmented Dickey-Fuller test for detecting single bubble episodes using forward-recursive regression windows.

GSADF Test

Generalized SADF for multiple bubble detection with flexible start/end windows. More powerful for recurring episodes.

LPPLS Framework

Log-Periodic Power Law Singularity model for robustness check. Identifies bubble run-ups and crash predictions.

Date-Stamping

Joint bubble indicator requiring 7+ consecutive days of overlapping SADF and GSADF intervals for conservative identification.

GSADF Analysis (12 figures)
GSADF NFTI

NFTI

Bubble periods (red) when test statistic exceeds critical value

GSADF ETH

ETH

Highest bubble coverage (19.5%), multiple episodes 2017-2021

GSADF BNB

BNB

Bubble regimes during 2021 bull market

GSADF SOL

SOL

Sharp explosive episode late 2021

GSADF ADA

ADA

Multiple bubble periods in 2021

GSADF DPI

DPI

Extended bubble state during DeFi Summer 2020

GSADF UNI

UNI

Bubble at launch and 2021 boom

GSADF AVAX

AVAX

Late 2021 explosive behavior

GSADF WBTC

WBTC

Tracks BTC bubble patterns

GSADF LINK

LINK

DeFi-correlated bubble episodes

Co-occurrence Heatmap

Co-occurrence

Pairwise bubble overlap frequencies

Dynamic Co-occurrence

Dynamic Zoom

Time-varying co-explosivity across assets

SADF Analysis (10 figures)
SADF NFTI

NFTI

Forward-recursive test for single bubble

SADF ETH

ETH

Detects single largest explosive episode

SADF BNB

BNB

Single bubble detection analysis

SADF SOL

SOL

Forward recursive test results

SADF ADA

ADA

Single bubble detection

SADF DPI

DPI

DeFi Summer 2020 detection

SADF UNI

UNI

Launch period bubble detection

SADF AVAX

AVAX

Single explosive episode detection

SADF WBTC

WBTC

Bitcoin-correlated bubble

SADF LINK

LINK

DeFi ecosystem bubble detection

LPPLS Confidence Indicators (10 figures)
LPPLS CI NFTI

NFTI

Positive/negative confidence indicators

LPPLS CI ETH

ETH

Run-up and crash probability signals

LPPLS CI BNB

BNB

Bubble probability confidence bands

LPPLS CI SOL

SOL

Crash warning indicators

LPPLS CI ADA

ADA

LPPLS confidence over time

LPPLS CI DPI

DPI

DeFi bubble probability signals

LPPLS CI UNI

UNI

Crash/run-up probability

LPPLS CI AVAX

AVAX

Bubble confidence indicators

LPPLS CI WBTC

WBTC

BTC-linked crash signals

LPPLS CI LINK

LINK

LPPLS probability bands

LPPLS Composite Fits (10 figures)
LPPLS Fit NFTI

NFTI

Model fits overlaid on price series

LPPLS Fit ETH

ETH

Fitted bubble trajectories

LPPLS Fit BNB

BNB

LPPLS model overlay

LPPLS Fit SOL

SOL

Fitted crash predictions

LPPLS Fit ADA

ADA

Model trajectory fits

LPPLS Fit DPI

DPI

DeFi bubble model fits

LPPLS Fit UNI

UNI

Fitted LPPLS trajectories

LPPLS Fit AVAX

AVAX

Model overlay analysis

LPPLS Fit WBTC

WBTC

Bitcoin-linked model fits

LPPLS Fit LINK

LINK

LPPLS trajectory overlay

Liquidity Analysis (6 figures)
Liquidity NFTI

NFTI

Price and volume with bubble periods shaded

Liquidity ETH

ETH

Volume spikes during bubble regimes

Liquidity BNB

BNB

Price-volume relationship in bubbles

Liquidity SOL

SOL

Trading activity during exuberance

Liquidity ADA

ADA

Liquidity surge in bubble periods

Liquidity DPI

DPI

DeFi volume during bubble states

Open Science Repository

Detecting Price Bubble Formation in NFT and DeFi Markets

Lennart J. Baals (University of Twente)

This deposit provides the full research code and executable analysis pipeline for Chapter 2 of a doctoral dissertation on risk management in digital finance. The work examines price bubble dynamics in NFT and DeFi markets using a reproducible R/Quarto workflow combining PSY bubble tests (SADF/GSADF), LPPLS robustness analysis, and liquidity assessment.

License

MIT License

Language

R (>=4.0)

Size

15.6 MB

Published

Jan 5, 2026

Repository Contents:

6 Quarto notebooks 6 R source scripts 60+ PNG visualizations 2 CSV summary tables

R Dependencies:

tidyverse, tseries, moments, exuber, arrow, patchwork, janitor

Related Publication:

Wang, Y., Horky, F., Baals, L. J., Lucey, B. M., & Vigne, S. A. (2022). Bubbles all the way down? Detecting and date-stamping bubble behaviours in NFT and DeFi markets. Journal of Chinese Economic and Business Studies, 20(4), 415-436. [DOI]

Funding: Swiss National Science Foundation (Grant IZCOZ0-213370): "Narrative Digital Finance: a tale of structural breaks, bubbles & market narratives"

Note: Raw market data excluded due to third-party licensing restrictions. Users must obtain data from original vendors (WorldCoinIndex, CoinGecko) independently.