# Reinforcement learning environment design for crypto trading [closed]

• how should one design the observation and reward for a crypto trading environment?
• suppose the action design is simple: do nothing, buy, sell
• What have you already tried and what are the bad results you're getting? Aug 5 at 10:01
• many things... hard to say... just share you thoughts please
– Rick
Aug 5 at 10:16
• Hello. You need to provide more details. You need to explain in more detail what a "crypto trading environment" is. You're explaining what the actions are, what about the rest? Some people may not be familiar with it. You should also explain what you've tried so far (as someone else above suggested) and why it doesn't "work". Please, take the time to read How do I ask a good question?.
– nbro
Aug 5 at 13:13

See my answer at https://ai.stackexchange.com/a/29944/5763 where I refer to Bruce Yang's excellent article "Deep Reinforcement Learning for Automated Stock Trading".

Crypto trading isn't much different than stock trading if you base it on price and volume data and derivations of it such as technical indicators. Using the same model:

Observations: crypto price, portfolio balance, amount, and technical indicators

Rewards: profit and loss