Development of MyStockScreener for Intraday & Short-Term Trading

[murthymsr]

MyStockScreener for Intraday & Short-Term Trading

Hi traderji and all members of traderji family,

Happy Diwali ! The new year to follow may bring prosperity to all members of the traderji family.

I would like to present the StockScreener developed by me to the learned members of this august forum for review and suggestions for improvements. (I neither consider myself as a good programmer nor a good investor). The output of this software (identified stocks) will be posted and made available to members of this forumon a day to day basis.

I intend to share the ideas indicated below so that the seniors can give their opinion on the domain value such an application would deliver to the community so that i can freeze the requirements and get started on the release of the output of the StockScreener.

Now, about the stock screener and about the process of identifying the stocks.

The process of identifying a stock with a potential to give reasonable returns from out of more than 800 stocks traded on NSE is a herculian task, particularly for day/short-term traders as the picture changes by the end of every trading day.

I wished that a stock screener may be there to help in shortlisting the traded stocks to a few which can be studied in detail, as no software can predict the market nor can replace an expert. I searched the internet without much success. I attempted to develop one and the StockScreener is born.

In the current version, the stocks are ranked by their potential for giving returns and top 30 stocks are selected each day. The input data used is the previous few days' (3 to 5) stock price movements based on the the eod data downloaded from the NSE site. A few parameters based on volumes, prices, volatility and pivot points etc. are identified and the performance of the stock evaluated on these parameters. Low turnover and special category stocks are excluded before processing.

Each parameter identified is tested for correlation to the 'Potential' for giving returns. The ranked stocks are tested for gradient in 'potential for earnings'. My preliminary tests revealed that the results are about 65 % reliable when the market is steady.

The process of StockScreener and parameters considerd for identifying the stocks are provided in the attachment. I appeal to all learned members (with either trading edge or programming edge) to go through and offer their comments/suggestions for improvements.

The StockScreener output based on the eod data (current version) can be posted by about 7 PM on each trading day (i.e., 1 hour after Bhav copy is available at the NSE site). This leaves enough time for the investor to study the set of about 30 stocks and to decide on the game plan for the following day.

Pl. refer attached text document for detailed of process and parameters.

with kind regards

murthymsr

 

[kgsirasangi]

very good idea.
congratulations on developing this stock screener.
shall look forward to your daily reports
regards

 

[karthikmarar]

Dear Murthy

Very commendable effort.

My immediate thoughts after going through your write up are as follows.

1. In order to evaluate the sufficiency and efficiency of the chosen parameters it would be better to put down the objective of the parameters. I can understand the need or objective of some of the parameters for example
-

• Ratio of last days turnover to the 3 day EMA would indicate the immediate volume trend is up or down.
  
• Ratio of close to the 3 day EMA would reflect bullish or bearish

In the similar fashion if you chart out the objective of each parameter our analysis of the parameters would be objective

2. Consider the near duplicate parameters for example 1 &2 and 3&4. In parameter 2 and 4 you are comparing the turnover and close to the third days values. If the third values happens to be bad it will affect the weightage drastically even if the next two days values are good. Since you are evaluating the performance for a very period of 3 days is such redundancy of parameters required.

3. Comparing the close to the close a month back (parameter no. 9) does not provide some objective information. Wont it be better to compare to the moving average instead of comparing to one single value? At least this would tell us if the overall trend is up or down.

4. How is the volatility measured? ATR?

Some suggestion

Since the sentiment for the day is dependant on the overall market sentiment, is there some way to reflect it on the stocks. Like the performance of the particular stock compared to the nifty for the last 3 days. Stocks outperforming the indices should get more weightage etc.

The overall trend of the stock could also be reflected on the evaluation. Naturally if the medium trend is bullish the chance of the stock going is more and vice versa. If the stock is moving sideways the movement becomes less predictable.

I will revert back after further detailed analysis. You are doing excellent work.

Why dont you post the list from your screener. It will help us to evaluate the parameters better.

Wishing you the best

karthik

 

[murthymsr]

Dear karthik,
Thanks for your valuable suggestions, inspite of your being busy.
Please read my 'Reply' to each of your 'Comment'.

Comment: "1. In order to evaluate the sufficiency and efficiency of the chosen parameters it would be better to put down the objective of the parameters. .... "

Reply: The overall objective is to identify stocks that have potential to raise/fall more than the rest with a horizon of a) one day, b) a few days and 3) a few weeks etc.,. The various parameters that could effect in any of the horizons are considered but different set of weighatages will be given to the parameters depending on the chosen time-frame. (this includes zero weightage)

Commennt: "2. Consider the near duplicate parameters for example 1 &2 and 3&4. In parameter 2 and 4 you are comparing the turnover and close to the third days values. If the third values happens to be bad it will affect the weightage drastically even if the next two days values are good. Since you are evaluating the performance for a very period of 3 days is such redundancy of parameters required. "

Reply: I agree that comparision to 3 day's back value can sometimes give false signals. Actually, i have first used the comparision with 3 day's back values only, but noticing less correlation, i have added the 3 day ema's but still retained the old one also, in the hope that it does not harm, but 'may' improve correlation. Now i will look into, if i should remove the 3 days back comparision.

Comment: "3. Comparing the close to the close a month back (parameter no. 9) does not provide some objective information. Wont it be better to compare to the moving average instead of comparing to one single value? At least this would tell us if the overall trend is up or down."

Reply: Yes, instead of 1 month back values, closing price will be compared with 20 day ema.

Comment: "4. How is the volatility measured? ATR?"

Reply: This software can do only some simple arithmatic calculations. Volatility is measured by the ratio of day's high price to the day's low price and it's 3day ema also considered.

Comment: "Since the sentiment for the day is dependant on the overall market sentiment, is there some way to reflect it on the stocks. Like the performance of the particular stock compared to the nifty for the last 3 days. Stocks outperforming the indices should get more weightage etc"

Reply: Yes, i know, this is very important. But the limitation is data source. NSE in it's PR.ZIP file provides index data, but the files are not archieved. If i miss the data for a day, because of my non-availability or any other resource non-availability, the chain is lost because of non-continuity and the calculated ema's are not reliable. On the other hand, Bhav copy of NSE site does not give index data, but are archieved. The data formats of these two sources are different. So in the future versions, i may think of using PR.zip and include the assesment of nifty relative performance, and use Bhav copy data as a back-up when i could not retrieve PR.Zip.

Comment: "Why dont you post the list from your screener. It will help us to evaluate the parameters better."

Reply: I can post anytime. But my plan was to receive comments, implement as possible, do back testing with the October's data and post the Test results alongwith the first output of screener. I think i will be able to make my first output in the second week of November.

The output is just a series of NSE codes, comma separated, in the decreasing order of their potential to grow.

Comment: "I will revert back after further detailed analysis."

Karthik, I look forward to your valuale suggestions.

 

[karthikmarar]

Dear Murthy

Just a little clarification on my comment no.1

In order to judge the effectiveness a particular parameter we should first clearly understand why a particular parameter is chosen. How effective the particular parameter is in influencing the price? Is the wightage given for the parameter is correct?..like no point in giving a less influencing parameter more weightage....For that only I requested you to chart out the reasons / influence of each parameter (according to you) on the stock value. then the members can comment / judge the effectiveness of each parameter.

With the correct parameters and the right weightage your screener would do a wonderful job.

best regards

karthik

 

[murthymsr]

Dear karthik,
Thanks for your response.

The correlation between the potential and each of the identified parameter was studied for 10 trading days and accordingly the weightages are identified. For this the CORREL function of excel spreadsheet was used. Correlation coefficient was positive and the average values are taken to ascertain the weightage.

But there was wide variation in the correlation coefficient, and this is where i felt i may take the help of experience also. This correlation was also one of the main tests i plan to do after finalisation of the parameters, to ascertain the best possible weightages. This testing will be done for the data of at least a month (20 trading days) and averages will be taken, unless i find some advice contrary in this forum.

As i said earlier, i intend to post the final criteria and the weightages for each parameter as also the summary test results also along with the first output.

This may help others to come out with a better weightages or even a better StockScreener.

Thanks for your time
murthymsr

 

[karthikmarar]

Dear Murthy

Nice to now that the the parameter and their weightage were arrived with some statistic analysis.

The parameters can be improved overtime after analyzing the results after testing in the actual market situation and adjusted later.

Eagerly waiting for you to back test the screener and post the results. It would be interesting to study that.

I found your work quite interesting. I will get back to you on further observation and suggestions in any.

All the best.

regards

Karthik

 

[lvgandhi]

Where and when do you post the results?

 

[murthymsr]

hi ivgandhi,

The selections of the stock screener are being posted since 10 days in the thread "Aces in the deck"
in >>> The Markets >>> Equties.

You may find the EOD Selections for the day in the above thread between 6 PM & 9:55 AM on the trading day.

murthymsr

 

[younggenius]

Do U Plan to share the software with us ;-) Or would it be for sale.
The reason I am asking this is if such a software is made available then people can perform analysis on their own.