Scientists reverse-engineered the Luna flash crash using particle physics


Just eight days after Terra co-founder Do Kwon instructed American-Canadian chess star Alexandra Botez that 95% of cryptocurrencies would fail, including that there was “entertainment in watching companies die,” the Luna flash crash occurred.

More than $40 billion in investor property had been misplaced in the crash between May 5 and May 13, 2022. Less than a 12 months later, Do Kwon was arrested after allegedly making an attempt to flee prosecution for prison exercise related to the losses.

Volumes have since been written discussing the breakdown, which noticed the Luna (LUNC) coin plummet and Terra’s UST stablecoin depeg from the U.S. greenback.

Now, for what seems to be the first time, scientists have applied statistical mechanics to basically reverse-engineer the crash using the similar strategies used to check particle physics.

The analysis, performed at King’s College London, focused on transaction occasions and orders occurring throughout the crash. Per the group’s preprint analysis paper:

“We view the orders as physical particles with motion on a 1-dimensional axis. The order size corresponds to the particle mass, and the distance the order has moved corresponds to the distance the particle moves.”

These similar strategies are used to map thermodynamic interactions, molecular dynamics and atomic-level interactions. By making use of them to particular person occasions occurring throughout a particular time frame in a contained ecosystem, reminiscent of the Luna market, the researchers had been in a position to glean deeper perception into the coin’s microstructure and the underlying causes for the collapse.

The course of concerned shifting away from the snapshot methodology concerned in the present state-of-the-art method, Z-score-based anomaly detection, and shifting right into a granular view of occasions as they occurred.

By viewing occasions as particles, the group was in a position to incorporate layer-3 knowledge into its evaluation (which, above layer-1 and layer-2 knowledge, consists of knowledge pertaining to order submissions, cancellations and matches).

According to the researchers, this led them to uncover “widespread instances of spoofing and layering in the market,” which tremendously contributed to the Luna flash crash.

Luna spoofing revealed throughout the Terra collapse using three disparate knowledge evaluation strategies. Source: Li et al., 2023

The group then developed an algorithm to detect layering and spoofing. This introduced a big problem, in line with the paper, as there aren’t any recognized knowledge units associated to the Luna crash that include precisely labeled situations of spoofing or layering.

In order to coach their mannequin to acknowledge these actions with out such knowledge, the researchers created artificial knowledge. Once skilled, the mannequin was then utilized to the Luna knowledge set and benchmarked towards an present evaluation performed through the Z-score system.

Related: Top UK university partners with AI startup to analyze crypto market

“Our method successfully detected spoofing events in the original dataset of LUNA trading market,” wrote the researchers, earlier than noting that the Z-score methodology “not only failed to identify spoofing but also incorrectly flagged large limit orders as spoofing.”

Going ahead, the researchers imagine their work may function a basis for learning market microstructure throughout finance.