search

shenyichen105 / Deep-Reinforcement-Learning-in-Stock-Trading

Using deep actor-critic model to learn best strategies in pair trading
309 stars 127 forks source link

Negative port value #2

Open joaosalvado10 opened 7 years ago

joaosalvado10 commented 7 years ago

Hello, first of all, thank you for your great contribution. After successfully run the code I found out by locking to csv output files that the final port value is negative. The results are not the same as the ones in the data_for_vis_final.csv. Am I missing something? Also I would like to ask you how is calculated the price of buying/selling - Long/Short action? How would you change the project to take into account only one equity.

lorrp1 commented 6 years ago

same, not only the result are not promising, the data in the cvs is wrong (close value are wrong, adj close are more similar to real the real stocks price but still are not the same as stocks price i saw on google).

"Also I would like to ask you how is calculated the price of buying/selling - Long/Short action" there is not calculation the bot should "learn" to buy and sell on its own, but i this .py seems to trade randomly, as 20 interactions got worse result then 5 (maybe it would require 10000 interactions but its not written)