FinRL icon indicating copy to clipboard operation
FinRL copied to clipboard

Published 20 hours ago • verified publisher icon AI4Finance-Foundation

Model converging to buy and hold strategy every time it is trained/optimized

Open flyingsandwich1 opened this issue 2 years ago • 4 comments

Does anyone have recommendations on what to do to fix this? My model essentially learns to just buy and hold stocks instead of exploring trading strategies. My learning rate spread is quite large (between 1 and 1e-8) for Optuna optimization so its weird it gets stuck on this everytime. Is there some way to punish the model for holding a stock too long or make it trade more?

flyingsandwich1 avatar May 23 '22 05:05 flyingsandwich1

Well, interesting. Previously, several people reported too frequent trading and want to punish it. From your description, your codes may include some bugs. Fix them before trying to do logic change.

YangletLiu avatar May 23 '22 14:05 YangletLiu

I am having the same problem with TD3 and doesn't know why yet.

trungtv avatar Jun 02 '22 04:06 trungtv

@flyingsandwich1 I might be able to help. Send me an email: [email protected]

george-adams1 avatar Jul 13 '22 21:07 george-adams1

Same issue .Is there a way to punish the model for holding a stock too long?

ae17b008 avatar Mar 13 '23 06:03 ae17b008