WIP burn_in feature for all simulations

sbenthall commented 2 years ago

Addresses #122 by functionalizing out the burn-in period from Chum into a method for all simulations, and working this through Calibration and Attention simulations.

A couple open questions about this:

What to do about the dividend process during burn-in? Chum let it walk normally.
There are inconsistencies in how Chum and AttentionSimulation have been doing tracking slightly differently. Currently, there are errors when using the burn_in when doing the AttentionSimulation because of these discrepancies. Fixing these is issue #81.

The upshot is I need to talk to @mesalas and @gms158 because this change will effect downstream data output (and aggregation scripts) as well as upstream market behavior.

sbenthall commented 2 years ago

Decision: We're going to (optional) output the burn-in data but not track it with the main simulation results.

sbenthall commented 2 years ago

That means we should put the relative price difference on a day as a column in the _data output

sbenthall / SHARKFin

WIP burn_in feature for all simulations #137