Algorithm-Based Low-Frequency Trading Using a Stochastic Oscillator and William%R: A Case Study on the U.S. and Korean Indices
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe authors have presented an excellent topic and merged it from a mathematical and financial perspective. The methodology is sound and the results depict algorithm trading approach benefits for both individual and institutional investors.
The conclusion can be further enhanced to add policy recommendations as two very diverse indices are being used. Also what are other alternatives other than deep signal. This should be explained.
Comments on the Quality of English LanguageN/A
Author Response
Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.
1. Reviewer Feedback
The conclusion can be further enhanced to add policy recommendations as two very diverse indices are being used. Also what are other alternatives other than deep signal. This should be explained.
1. Author Response
Thank you for your valuable advice. In this study, the max drawdown was successfully limited. However, we would like to apply a loss-cut rule to our investments to limit unexpected losses in the future. In fact, institutional investors and individual investors have loss-cut rules based on their investment goals and fund types. During the test period, the max drawdown of the S&P 500 was -15.1% weekly base, and the max drawdown of the MSCI Korea was -12.3% weekly base. In order to manage the investment within this smaller loss, we add a -10% weekly base loss-cut rule.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsHello All,
Apparently it is an interesting read. However, plenty of research gaps and missing theories needed to be bridged.
1. Why specifically US and Korea? Are they similar? Are they different? Justify in the manuscript.
2. Why only Stochastic oscillator and William%R? Justify in the manuscript.
3. Please explain the Heuristics behind the trading algorithm 1 (6th page)!
4. Typically markets deviate from EMH and MPT, therefore, why 'Drift' is ignored here? In fact drift is integral for any stochastic time series!! Please justify in the manuscript.
5. Scientific premise of 'DeepSignal' model is rather unclear to me!! Explain in detail (in the manuscript).
6. What is the novelty of this work, other than showcasing some empirical data analysis? What is the theory/hypothesis underneath?
7. Some strategy (random) which works in the past data set does not necessarily mean that it will work well in future and safeguard the investor wealth. How to defend this sort of a rebuttal? Properly analyse and possibly add a section on the 'shortcomings'.
8. Additionally, the knowledge of 'Drift Burst Hypothesis' would help you to make the paper scientifically sound.
9. How did you suggest 80% weight of the U.S. market and 20% weight of the Korean market? There are plenty of ad-hoc assumptions are there. Far too many, I suppose for a research paper. Each of them needs justification, if possibly backed by an existing work.
Sincerely
Reviewer
Author Response
Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.
1. Reviewer Feedback
Why specifically US and Korea? Are they similar? Are they different? Justify in the manuscript.
1. Author Response
Thank you very much at all.
The S&P 500 US market index is the biggest index in the MSCI ACWI and MSCI WORLD, which are representative benchmark indices. The institutional and individual investors compare the return rate to the benchmark. In this situation, the biggest index is an important index to invest and simulate. Also, the MSCI KOREA index is the second biggest index in the MSCI EM index, which is the benchmark in the emerging stock market. After 2018, the U.S. had heavy conflicts with China that brought a long downtrend in the Chinese stock market. So, we select the two key indices in both markets, which are the developed market and the emerging market. Of course, the two indices are different: developed economy vs. emerging economy, key currency vs. non-key currency, size of the market, the power of politics, etc.
2. Reviewer Feedback
Why only Stochastic oscillator and William%R? Justify in the manuscript.
2. Author Response
Thank you very much.
First of all, predicting the stock market using stochastic has been widely discussed in academia (Bartolozzi, 2009; Neely, 2009; Mariani, 2021; Davies, 2019). We combine the oscillators to get better results or avoid risk. We tested various representatives of the stock market technical oscillators such as William%R, Bollinger band, MACD, MA, Envelope, RSI, Pivot, etc. The William%R technical oscillators could eliminate the noise and increase the hit ratio with the stochastic oscillator.
3. Reviewer Feedback
Please explain the Heuristics behind the trading algorithm 1 (6th page)!
3. Author Response
Thank you for your comments.
The stochastic oscillator ranges from 0 to 100, and the William%R ranges from -100 to 0. We tested every five changed parameters to determine decisions. Moreover, we figured out which parameters are optimized. Finally, we established the standard of algorithm trading rule. 1. Buy = Stochastic oscillator < 30 and William%R < -75. 2. Sell = Stochastic oscillator > 80 and William%R > -20. 3. Else be cash
4. Reviewer Feedback
Typically markets deviate from EMH and MPT, therefore, why 'Drift' is ignored here? In fact drift is integral for any stochastic time series!! Please justify in the manuscript.
4. Author Response
Thank you very much.
As the reviewer said, the market's random walk neutralizes EMH and MPT. In response to these trends that appear in the actual stock market, we stuided to help with investment in other ways. In fact, many past studies have attempted to do this through descriptive, quantitative, and qualitative analysis. We tested how to respond to the market's random walk-through technical indicators.
5. Reviewer Feedback
Scientific premise of 'DeepSignal' model is rather unclear to me!! Explain in detail (in the manuscript).
5. Author Response
Thank you for your comments.
We presented the equitation of the stochastic oscillator. First of all, set the parameters for %K. We tested various n data from one to fifty-two based because we used the weekly base data. Moreover, we figured out that ten was fitted for n. Next, we should select the n of the %D. We tested the same as before and selected number six for n of the %D. The stochastic oscillator ranges from 0 to 100. A lower number of stochastic oscillators means a low stock price and a higher number means a high stock price. We tested every five changed parameters to determine decisions.
For the William%R oscillator, we tested various n data from one to fifty-two because we used the weekly base data. Moreover, we figured out that ten was fitted for n of the Low n and the High n. The William%R ranges from -100 to 0. A lower number of the William%R oscillators means a low stock price and a higher number means a high stock price. We tested every five changed parameters to determine decisions.
Moreover, we figured out which parameters are optimized. Finally, we established the standard of algorithm trading rule. 1. Buy = Stochastic oscillator < 30 and William%R < -75. 2. Sell = Stochastic oscillator > 80 and William%R > -20. 3. Else be cash
We presented the simulation trading table in the manuscript.
We presented the simulation of the SPY ETF, which is a representative S&P 500 index. In the table below, the first buy trading signal was in 20.JUN.2010. On that day, the stochastic oscillator was 26.0 points, and William%R was -94.4 points. After that, we follow the algorithm rule. There were eight trading times. The max gain was +28.7% and max drawdown was -0.1%.
We presented the simulation of the EWY ETF, which is a representative MSCI KOREA index. In the table below, the first buy trading signal was in 05.SEP.2011. On that day, the stochastic oscillator was 14.5 points, and William%R was -88.0 points. After that, we follow the algorithm rule. There were eleven trading times. The max gain was +47.8% and max drawdown was -2.5%.
6. Reviewer Feedback
What is the novelty of this work, other than showcasing some empirical data analysis? What is the theory/hypothesis underneath?
6. Author Response
Thank you very much for your comment.
Infrequent trading, a high hit ratio, and better returns versus the index were obtained. This allows us to address transaction costs for real investors, allowing them to make profits and feel secure.
Market-timing investing is possible through indicators in the random walk of the stock market.
7. Reviewer Feedback
Some strategy (random) which works in the past data set does not necessarily mean that it will work well in future and safeguard the investor wealth. How to defend this sort of a rebuttal? Properly analyse and possibly add a section on the 'shortcomings'.
7. Author Response
Thank you for your comment.
We achieved optimal results for this study through historical optimization. We advised that when investing in the future using this algorithm, the results may vary. It is clear that the statement that past analysis may not guarantee future investment performance cannot be denied. However, since there have been confirmed cases where no trading occurred or losses were minimized in a situation where extreme volatility occurred or annual declines were severe, the portfolio can be defended against a market long-only strategy.
8. Reviewer Feedback
Additionally, the knowledge of 'Drift Burst Hypothesis' would help you to make the paper scientifically sound.
8. Author Response
We appreciate your constructive comments. We strongly believe that the ‘Drift Burst Hypothesis' makes our article better; thus, future our works too.
9. Reviewer Feedback
How did you suggest 80% weight of the U.S. market and 20% weight of the Korean market? There are plenty of ad-hoc assumptions are there. Far too many, I suppose for a research paper. Each of them needs justification, if possibly backed by an existing work.
9. Author Response
Thank you very much.
The assumption of the weight of each country is based on the MSCI ACWI’s DM(developed market) and EM(emerging market) weights. The MSCI ACWI is the most used benchmark for global investors. The composite of the MSCI ACWI is 80% of the DM and 20% of the EM for the long-term period of two decades. So, we set the fixed the DM(S&P 500) and EM(MSCI Korea), 8:2 ratio.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors
The purpose of this research is to investigate a simple and general trading model for low-frequency, low-risk, high-return, and high-hit ratio model for individual investors. It offers a DeepSignal model. The trading algorithm is quite simple. Buy if Stochastic oscillator < 30 and William%R < -75; Sell if Stochastic oscillator > 80 and William%R > -20. Howerver, this research donot provide nither theoretical basis nor strong evidence for determining the parameters. Why are they 30, -75, 80, -20? The paper claims "we verified that our method is generic in an uptrend market such as the S&P 500 and a side walk trend such as the MSCI Korea. This implies that our algorithm trading model was working at different type of the trends." It would be improper to generalize by just one uptrend market and one MSCI Korea. This research still has a lot of work to do and should not leave for future work.
Comments on the Quality of English LanguageEnglish in some sections appears to be difficult to understand and many typo errors exist. For example, “We are the set of algorithms to apply the first U.S. index S&P 500. ”, “the number of gains was eeight times”, “and stay in market strategy/”. Quality of presentation is not good. Figures 12-14 are actually tables, and their layout is messy.
Author Response
Thank you very much for taking the time to review this manuscript. Please find the detailed responses below and the corresponding revisions/corrections highlighted/in track changes in the re-submitted files.
1. Reviewer Feedback
The purpose of this research is to investigate a simple and general trading model for low-frequency, low-risk, high-return, and high-hit ratio model for individual investors. It offers a DeepSignal model. The trading algorithm is quite simple. Buy if Stochastic oscillator < 30 and William%R < -75; Sell if Stochastic oscillator > 80 and William%R > -20. Howerver, this research donot provide nither theoretical basis nor strong evidence for determining the parameters. Why are they 30, -75, 80, -20? The paper claims "we verified that our method is generic in an uptrend market such as the S&P 500 and a side walk trend such as the MSCI Korea. This implies that our algorithm trading model was working at different type of the trends." It would be improper to generalize by just one uptrend market and one MSCI Korea. This research still has a lot of work to do and should not leave for future work.
1. Author Response
Thank you for your comments.
First of all, predicting the stock market using stochastic has been widely discussed in academia (Mariani, 2021; Davies, 2019; Bartolozzi, 2009; Neely, 2009 ). We combine the oscillators to get better results or avoid risk. We tested various representatives of the stock market technical oscillators such as William%R, Bollinger band, MACD, MA, Envelope, RSI, Pivot, etc. The William%R technical oscillators could eliminate the noise and increase the hit ratio with the stochastic oscillator.
The stochastic oscillator ranges from 0 to 100, and the William%R ranges from -100 to 0. We tested every five changed parameters to determine decisions. Moreover, we figured out which parameters are optimized. Finally, we established the standard of algorithm trading rule. 1. Buy = Stochastic oscillator < 30 and William%R < -75. 2. Sell = Stochastic oscillator > 80 and William%R > -20. 3. Else be cash.
In the global stock market, the U.S. is the representative index with the highest market capitalization that maintained the MSCI DM upward trend during the test period. Among many indices that move sideways, Korea is the leading index in the EM stock market. Although we cannot generalize, we consider it positive that we have achieved good results in markets with different trends. As you commented, I would be careful about making definitive statements about generalizations.
2. Reviewer Feedback
English in some sections appears to be difficult to understand and many typo errors exist. For example, “We are the set of algorithms to apply the first U.S. index S&P 500. ”, “the number of gains was eeight times”, “and stay in market strategy/”. Quality of presentation is not good. Figures 12-14 are actually tables, and their layout is messy.
2. Author Response
We appreciate your constructive comments. We rewrote the typo errors and Figures 12-14 were changed to table style in the manuscript.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsDear Authors,
I am satisfied with the revision.
Regards
Reviewer
Author Response
Dear reviewer
Thank you very much for taking the time to review this manuscript. We appreciate your excellent comments and response.
Best regards
Authors
Reviewer 3 Report
Comments and Suggestions for AuthorsThe research tested every five changed parameters to determine decisions, then optimized the parameters and established the standard of algorithm trading rule. This is what we called grid seach, a simple technigue in data mining. The optimized parameters probally vary with samples. Therefore, the algorithm trading rule is not necessarily robust to other sistution. Investors need to be careful to implement it. The index price data comes from the Yahoo finance site. Many methods related to machine learning get stock data from Yahoo finance for convenience. However, Yahoo finance fails to process data like CRSP do. CRSP is more rigorous. Nevertheless, this research still provides some insgihts into the application of Stochastic oscillator and William%R. The authours also do their best to improve the manuscript.
Comments on the Quality of English Language
Better then the previous version.
Author Response
Dear reviewer
Thank you very much for taking the time to review this manuscript. We appreciate your excellent comments and response.
We will write ‘grid searching’ in that part of the manuscript and emphasize the caveat that the parameters may vary depending on the sample.
I emailed CRSP to subscribe but have not heard back from them in time. Instead, we compared the YAHOO FINANCE data and BLOOMBERG terminal data. Fortunately, there were no anomalies in this case. We will use more rigorous data in future research, as you mentioned.
We edited over 30 changes to the English phrase from a clarity, brevity, and flow perspective.
Author Response File: Author Response.pdf
