To improve the explainability of our model outputs, Gauntlet is working on a series of improved simulation features. We are particularly excited about scenario tools, which allow us to test specific hypothetical price levels while accurately replicating a volatile and correlated market.
While the overall trend of crypto markets is notoriously difficult to predict, we can replicate the typical behavior of asset price paths with multi-variable geometric Brownian motion (MultiGBM), given historical correlation and volatility data. However, in some cases, we may want to test paths that reach specific price levels. With the inherent randomness of Brownian motion models, it is difficult to define conditions like this in the simulated environment. We could run many simulations and use only those that hit the target, but this would be inefficient in practice.
Brownian bridges are a type of stochastic process that preserves the conditional probability distribution of a typical Brownian motion process, while “pinning” the start and end points. This allows us to simulate realistic paths between current prices and a specific future market state. For example, when modeling the depegging of stable assets or derivatives, it can be useful to specify a deviation from the underlying value that needs to be reached.
