TODO

  • plot without market bound still doesnt work

wrong figure was logged to tensorboard

  • optimal strat plot shows questionable results

wrong metric was displayed, reward instead of last wealth

  • relax the action space from to either or or

[-1_000, 1_000]

  • try different experiments where is twice num_steps

test_double_T

  • when running parallell envs try to shuffle the starting point to cover more of the episode length
  • pump up the num_steps numbers
  • train model on with with big num_steps_eval number (5_000)
  • plot the sharpe ratio on the training eval plot with a new y axis on the right to scale it properly

done

  • plot market bound for different time horizons and different hurst parameters

done

  • make a grid plot of fbm opt strats

done

  • recreate the sharpe ratio against hurst parameter plot for the optimal strategy

done

Notes

FBM on interval