Modeling Time

In the Balancer Simulations setup V1.0, we model Balancer Pools as discrete-time and discrete-state dynamical systems.

A transaction in a Balancer Pool is modeled as a discrete-time event updating the pool's state. They are referred to internally as Actions.

• similarly, price updates are processed as state updates

• the timestamp of every transaction, included in transaction data such as BigQuery Ethereum-Balancer, is preserved in the simulation output

• the simulation returns a datetime64 in 'change_datetime'

• datetime allows to use any Python library for analysing and manipulating time series data