notadamking / RLTrader

A cryptocurrency trading environment using deep reinforcement learning and OpenAI's gym
https://discord.gg/ZZ7BGWh
GNU General Public License v3.0
1.71k stars 537 forks source link

Question: why action changed to all in/out instead of partial(percentage of fund) #94

Closed ieow closed 5 years ago

ieow commented 5 years ago

Why action changed to all in/out instead of partial(percentage of fund) which used by previous version.

notadamking commented 5 years ago

The model has been extremely simplified to remove any sources of bias/overfitting/noise. One of these changes was to decrease the number of discrete actions, though it is unclear how this has affected the performance. It would be great for someone to test the performance of the agent with more discrete actions to see how it affects performance.