Open MysticDakra opened 2 years ago
One thought on this -- in time series modeling there's a concept of a "naive model": the value you would select if you did not have a model built. If you're predicting temperature tomorrow, the naive model would be average weather in the past n days.
Is there an obvious "naive delta" that we could use here that would be better than random selection? Maybe something like "best delta in past n days?"
@egweber2 we want to use this more as a benchmark, this proposal could be another ticket for a different strat to compare against other strats, lets connect because I think you may have work that can be converted to be used on this issue.
Describe the Solution Randomly select deltas between 0 - 0.5 ( maybe include a param that will prevent us from selling if it isnt worth) Purpose of this would be to test against other “strategies” and see if we can do better than it over a large number of paths. This will allow us to backtest our strategies against random ones.
While we are doing this please create a function that will allow us to plug other strats into our simulation so we can easily call them. Pretty much try to make it modular so it can be useful for other pieces of simulation
Expected behavior We should
Tasks
When is it considered Done? This is considered done when
Assignees Needs: Python coder - implement into our code base