The Luna flash crash came just eight days after Terra co-founder Do Kwon told American-Canadian chess star Alexandra Botez that 95% of cryptocurrencies would fail, adding that “watching companies die is entertainment.”
Do Kwon: “95% will die (coins), but it’s also fun to watch companies die”
8 days ago. ironnically. pic.twitter.com/fEQMZIyd9a
— Pedro (@EncryptedPedro) May 11, 2022
Investors lost over $40 billion in assets in the May 5-13, 2022 crash. Less than a year later, Do Kwon was arrested on suspicion of attempting to evade prosecution for loss-related criminal activity.
The debacle has since been discussed in numerous articles, including the Luna (LUNC) token plunge and Terra’s UST stablecoin depegging from the U.S. dollar.
Now, for the first time, scientists appear to have discovered application Statistical mechanics essentially reverse-engineers collisions using the same techniques used to study particle physics.
The research, conducted at King’s College London, focused on trading events and orders that occurred during the crash. According to the team’s preprint research paper:
“We think of order as physical particles moving on a one-dimensional axis. The order size corresponds to the particle mass, and the distance the order moves corresponds to the distance the particle moves.”
These same techniques are used Plot thermodynamic interactions, molecular dynamics, and atomic-level interactions. By applying them to individual events that occurred during a specific time period in a closed ecosystem (such as the Luna market), researchers were able to gain a deeper understanding of the microstructure of tokens and the root causes of crashes.
The process involves moving away from the snapshot approach, Z-score-based anomaly detection involved in current state-of-the-art methods, and into a granular view of events as they occur.
https://www.youtube.com/watch?v=16vAjsnazEM
By viewing events as particles, the team was able to incorporate Tier 3 data into its analysis (which, on top of Tier 1 and Tier 2 data, includes data related to order submissions, cancellations, and matches).
This, the researchers say, led them to uncover “widespread instances of deception and layering in the marketplace,” which largely contributed to the Luna flash crash.
Subsequently, the team developed an algorithm to detect delamination and spoofing. This presents a major challenge, the paper says, because no known dataset related to the Luna crash contains accurately labeled instances of spoofing or layering.
To train their model to recognize these activities in the absence of such data, the researchers created synthetic data. After training, the model will be applied to the Luna dataset and benchmarked against existing analysis via the Z-score system.
related: Top UK University Partners with AI Startup to Analyze Cryptocurrency Market
“Our method successfully detected fraudulent events in the LUNA trading market raw data set,” the researchers wrote, before noting that the Z-score method “not only failed to identify fraudulent behavior, but also incorrectly flagged large limit orders as fraudulent Behavior.”
Going forward, the researchers believe their work could serve as a basis for studying the microstructure of entire financial markets.
Svlook
