Deep Reinforcement Learning for Trading—A Critical Survey

Round 1

Reviewer 1 Report

The paper is well written. I was happy to read this paper. I have a minor comment for this paper. I suggest the authors to include short selling in stock market examples by DRL method in the revised paper.

 

Author Response

Thank you for your feedback.

We searched in our corpus what papers allow short selling and found out that only about 5 do, so we added a couple of phrases saying that as part of an extra point in the methodological issues section referring to the papers (page 20, middle of the page).

Reviewer 2 Report

Dear authors, The theme of your paper is very interesting but I have some comments:

1. The abstract is not obvious. The abstract has to contain: theme, goal, methods, procedure, main results.
2. You use in-text formulas, but I did not see them in the analysis. 
3. On the figure 4 to add the value of data

Author Response

Thank you for your feedback.

1. We modified the abstract as to state clearly the theme, goal, methods and main results.
2. We added a paragraph in section 10. Reward shaping where we talk about of the limitations and possibilities of these risk measures (in-text formulas).
3. We added the value of data in Figure 4 (now it's Figure 5).

Reviewer 3 Report

Dear Authors,

Here are my comments:

Overall authors have made a good attempt. However it needs further improvements as suggested below:

1. You have mentioned that you used meta analysis. Where is the meta regression equation and analysis? It is very important to map conditional estimates of effect.
2. Authors are suggested to include a detailed flow chart for the each stage of literature search in meta-analysis.
3. Why only google scholar is searched ? ... as the topic looks like more technical ... Why IEEE, ACM, SCOPUS, Web of Science databases etc., are not included or searched ?
4. Add a clear cut section on theoretical framework of the present study.
5. Authors are suggested to look into the already published few articles on meta analysis. Author may consider the below or similar articles.
   https://onlinelibrary.wiley.com/doi/10.1111/joes.12472
   https://onlinelibrary.wiley.com/doi/epdf/10.1111/joes.12399
6. Add limitations in the conclusion section.

Good Luck !!!

Author Response

Thank you for your feedback.

1. We do not use meta-analysis. We wanted to actually say corpus summarization or systematic review or something in between, more qualitative than quantitative. Which is, in a way, the beginning of a meta analysis, or a part of it. We did use the term meta-analysis lightly, we removed that and replaced it with corpus summarization. 
2. We added a flow chart in Figure 1, detailed in the text, in section 1. This was a reason to improve our search, investigate more, and download all papers by hand which allowed then to use recoll on the documents, which is a more reliable way of making this kind of investigation. 
3. We added an item in page 2 (lower half of the page), where we say the reason for searching only on Google Scholar. The reason is that Google indexes all articles, present on other portals as well.
4. We added Section 7 where we look at the financial market from a DRL perspective, as an environment and associate items from the financial domain to the DRL domain.
5. Thank you for the references. They were useful to understand what really is a meta-analysis and not take the term lightly.
6. We added a limitations section, where we state clearly the limitations of this study.

Round 2

Reviewer 3 Report

Good Luck !!!