Experiments in Technical Analysis

Hello Karthik

I am happy to have you here you don't know how much help you where to me and I thank God for showing me people like you inthis market from whom people like me can learn a lot. Well In India we don't have advantage like in US where you get quality technical institutes where you can go enroll and learn all about technicals here we have to do our own homeworks and our own studies and even if it is wrong no one is here to tell us Oh!! Boy u doin the wrong stuff its not done like this...so people like me go on working in a wrong direction so thats where I find people like you and beleive me people like you are like a lifeline for us beacause you all save us from blunders and mistakes and give us a chance to do something inour life. God bless you for this and if you like or not have taken you as my Mentor and am happy to follow your footsteps. I will be looking forward for your valuable advices from yor end.

your student,
Rahul

Thank you Karthik for the encouragement. I will definitely post whatever I try and seek the comments from experts like you. At present I am on the job. To start with I am trying out crossover only from your technique. You have used an extra condition for Buying. I am avoiding that and using only cross over for Buy. This gives me alternate BUY and SEll signals (Any crossover technique gives it). I am sure you might have already tried it out. But I am trying out with a set of thirty stocks randomly selected. All my experiments will be with these thirty. Also I am trying out these under different conditions. I will post the details if I succeed in getting some good results.

Thanks once again for your views and clarifying the points I raised. I am also waiting for some more expert opinions on the same.

-Anant

Hello Seniors

I am a junior in the world of stocks. Once I read a post about simple moving average of 5, 10, 20 in this site. I applied this and found the following scrips good to buy. Here I would request Seniors to atlest pick 2 or 3 stock from my list and guide me whether I am doing right or wrong.

ACC
AMTEKINDIA
ANDHRABANK
ASIANPAINT
CIPL
DALMIACEM
EASTSILK
GESHIPPING
IDFC
PNB
PNC
SCI
TVSELECT

Thanks in advance
Geo

Hello Anant,

Its good that you will be testing a system and intend to share the results in this thread. I would suggest you to share the results of your test in either case, good and bad results are equally educative.

Further if you will test any cross-over system on indian stocks for recent history they will give you excellent results, specially on the long side . The crossover system works very well with a trending market, which has been the case with our markets for last few years.

If you can get/have data, do try out your system with diff markets, for example if you use AB the default data that comes along can be used to test your system, you may be surprised to get totally different results.

But then again, you can share them both!

Regards And Best wishes
Sanjay

Hello Sanjay,

Thanks for your encouragement and guidance especially for the suggestion to test other markets. In fact I have not given details of the testing I am going to do. Most systems give excellent results when the trend is upwards. But I want to test the system under different conditions. Broadly I would describe it here:

1) Test the system for a period of 15 to 20 months in the past, continuously paper-trading the signals.
2) Test for a period when the individual stock is not moving up or down but within a narrow range especially over extended periods (a couple of months). The entry would be just before the sideway movement starts and continue till a little after the side way movement gives way to a new trend.
3) Test when the individual stock is going down. Again the entry would be just before the down trend starts and continue till the down trend ends.
4)Test under the crashing market conditions. That is when almost everything is going down, similar to the situation in May 2006.
5) Test the set of 30 randomly selected stocks under all the above conditions. That is the performance of the portfolio as a whole although the individual stocks are behaving differently.
6) The above portfolio would be tested in two different ways: (a) having a fixed number of each of the stocks, say 1 each of them. (b) Allot a fixed amount of money to each stock (say Rs 10000 each) and purchase whatever number of the shares that can be bought with that amount. These two methods are bound to give different results.
7) All Buy and Sell signals will be religiously followed without looking at the profit or loss generated by that trade. If more than one buy/sell signals are generated consecutively, only the first signal will be followed rejecting all others till an opposit signal is generated.

There is much more, which I would elaborate when the need arises. I want the system to undergo as rigourous testing as possible. At present I am working on four different strategies, inclding Karthik's MABITUS strategy. Let us see which gives best/worst results.

More when I get some results.

Thanks once again for your guidance.

Regards

-Anant

Rahul

Happy to note that this thread has been useful to you and Thanks for the kind words.

About ..mentor... we are all learners and we will help each other and learn together ...

Warm regards

Karthik

Hello Geo,

I had written about the 5, 10, 20 method. Good that you are able to get encouraging results. But be aware that this method works well in uptrend. During side-way movement or down trend it would not perform well, even generating big losses.

Best wishes for success to you

-Anant

Dear Anant,

Excellant, very elaborate and through testing. And of course very interesting too. Please do keep us posted on your findings.

Looking forward to more from you.

Warm regards

Karthik

Hello Anant,
You have indeed made a very through plan for testing the different systems.

I infer from your post that you are planning to walkthrough a time period of 15 to 20 months rather than using Automated Testing. That would definitely give you a better feel of the system.

Besides others, do consider the period between Jan 2002 and May/June 2003 for your studies.

Money Management is the most important aspect of any mechanical system.

Normally performance would be calculated in % terms and allocation of funds and position sizing should be formulated separately.

I would suggest you to store/save the results of your back tests in an excel sheet. Then you can use the features of excel to do what if analysis for many different scenarios and money management/position sizing options.

I agree with you that the overall performance of the entire portfolio is more important than individual scripts.

We all are aware that there is always some Slippage between the signal being generated and trade being executed. To get consistence results you can use some predefined rules for trade execution. (for e.g. closing of same day or opening of next day or high of first 30 mins/1 hour of next day etc.)

I am sure this Much More does include Stop Loss.

While testing a crossover system using data from recent Indian markets, I made an interesting observation. As I increased the level of Stop Loss the profits increased like if Stop Loss level was increased from 2% to 3%, there was a substantial increase in overall average profits. Most surprisingly the best results were with SL disabled.

Looking forward to your experimentations.

Regards and best wishes
Sanjay

Thanks Sanjay and Karthik,

These are quite good suggestions. To start with I will keep the system as simple as possible. Then introduce more and more conditions. Probably, when I reveal the results of the simplest strategy, many of the boarders here would apply their own modifications and try out. This in itself gives a good amount of feed back. It would be like making a wheel which people can modify to fit it to a cycle, scooter, car, lorry, aeroplane etc.

Yes, I am saving the results in Excel with one worksheet reserved for each stock.

Although different people will have different reasons to choose the buy/sell prices when the signal is generated, I am formulating a definite rule for buying and selling which would be strictly followed during testing.

To keep the system simple, in the beginning I am not following any stoploss . The SL, MM etc can always be added later to make the system more robust. I too have got results similar to what you mentioned when there is no SL. One of the four strategies I am testing has almost always produced profits only (for all the 30 stocks tested) without needing any SL when tested over a continuous period of about 20 months. But need to test at different points in time (like 1 year before, 3 years before or 5 years before etc.).

Well it is no use describing without giving any results. Let me go back and continue the "Experiments in Technical Analysis".

Thanks to all those who are giving me suggestions and encouragement.

-Anant