Price manipulation occurs when traders artificially inflate an asset's price through coordinated or fraudulent buying rather than genuine demand. Gandal, Hamrick, Moore, and Oberman's 2018 paper "Price Manipulation in the Bitcoin Ecosystem" examines suspicious trading by two automated Mt. Gox accounts, Markus and Willy. The study finds USD-BTC returns rose four to five percent on suspicious-activity days, based on approximately 18 million Mt. Gox transactions from 2011 to 2013. The suspicious activity coincided with a price run-up from around $150 to over $1,000 in two months in late 2013.
What the Study Found
On the 82 days with suspicious purchasing activity, the USD/BTC exchange rate rose 79% of the time, versus 55% on days without suspicious activity. The Markus bot acquired 335,898 bitcoins worth around $76 million over 33 of 225 active days in 2013. The Willy bot acquired 268,132 bitcoin nominally worth around $112 million over 50 of 65 active days. A regression controlling for DDoS attacks found Willy's activity raised the Mt. Gox rate by $21.65 per day, significant at p < 0.001. Willy's activity was linked to daily percentage returns 4.1 to 4.7 percent higher across all four exchanges, significant at the 99% confidence level.
Methodology
The study uses the leaked Mt. Gox transaction data dump, which links trades to individual user accounts. The dataset contains approximately 18 million matching buy and sell transactions. The data span April 2011 to November 2013, with suspicious activity identified from February to November 2013. Regressions control for DDoS attacks on Mt. Gox, the day after a DDoS attack, and other non-DDoS security incidents.
Key Statistics
| Metric | Finding | Context |
|---|---|---|
| Price rise frequency, days with suspicious activity | 79% | vs. 55% on days without suspicious activity |
| Willy coefficient, Mt. Gox rate-change regression | $21.65/day | RateChange = β0 + β1Markus + β2Willy + β3DDoS + β4DayAfterDDoS + β5Other + ε, significant at p < 0.001 |
| Willy coefficient, percentage-return regression | 4.1–4.7% | Returns = β0 + β1Markus + β2Willy + β3DDoS + β4DayAfterDDoS + β5Other + ε, across all four exchanges |
| Bitcoins acquired by Markus and Willy combined | ~600,000 BTC | by November 2013, valued at $188 million |
| USD/BTC exchange rate spike | $150 → over $1,000 | two-month period in late 2013 |
Why This Matters
Thinly traded cryptocurrency markets remain vulnerable to price manipulation by a small number of accounts. Investors and analysts evaluating cryptocurrency price movements should consider whether trading volume spikes reflect genuine demand or coordinated buying. The volume-spike pattern identified in the Mt. Gox case has recurred across other low-liquidity cryptocurrencies with small market capitalizations. Regulators and exchanges have an incentive to monitor unusual account-level trading patterns rather than relying on aggregate volume as a sign of market health.