Professional Analysis of an MCX Gold Trading System with Positive Expectancy.

GuluGulu

Well-Known Member
#1
Hi Traders,

May God Bless all of you. Returning to this forum after June, 2013. I was silent everywhere during this period and busy designing a system.

My intention is to analysis of my or any system from a mathematical point of view. This will follow through series of some posts. Today is first installment.
 

GuluGulu

Well-Known Member
#2
Glória in excélsis Deo


I am an Engineer and now a Professional Trader and by Grace of God, trading for a Living with my family.

After trading the markets with many methods, in July 2013, I started designing a Commodity Trading System that will trade some commodities in India's Multi Commodity Exchange (MCX). So I started gathering real time tick data for several commodity futures like Gold, Silver, Crudeoil, Natural Gas etc. My intention was to build a system that will give beautiful profit when operated in the long run.

Some preliminary requisite for the trading system was the following:

1. It should trade the commodity on pure technical basis. No discretionary and impulsive trades are allowed.

2. It must have a positive expectancy over many trades. Without this, the system will not be able to fetch money in the long run.

3. Total drawdown at a time should not be above the overnight margin required to hold the commodity. And it should recover from the drawdown maximum within a month of timeframe.

4. Average win to average loss ratio should be 1.5 or above. And as the system is a trend following system, the winning percentage could be around 45 to 50%.

5. It must give at least around 18% to 20% return per annum over the cash value of the futures contract; when traded with a single contract . That means, I am targeting more than double the bank fixed-deposit return per single contract per annum. Compounding this return with systematic position sizing will fetch brilliant profit in the long run - say 5 years period.

6. Consistence equity-curve.

Designing such a system is not that easy especially when you intend to make a consistent profit over any market condition. However I took a chance and started spending countless hours designing various models of such a system.
I tried various common system, namely, Fibonacci, Gann, all general indicators, Renko, everything one can imagine with three commodities: GOLD, SILVER and CRUDEOIL. I wanted to choose only these three because they have very good liquidity. And I choose MCX Exchange because it operates 13 to 14 hours per day - the longest of any Indian exchange - so that I can study with short time-frame like 1-hour candle.

So, in short, after a countless hours of study-design-testing and all combination of my work, I came to a working "Trend-Following" system that perhaps meeting all my above criteria.

And also I found MCX GOLD FUTURES contract gives consistent and best performance with this system.
 
Last edited:

GuluGulu

Well-Known Member
#3
Why Trend-Following:

As Tom Basso pointed out in an interview with Dr. Van Tharp, that every trader needs a trend, whether up or down, to make money. If you don't have a trend, you don't make money. You follow a trend - that means you 'let your profits run an cut your losses short'. And many successful and professional traders belongs to this 'Trend-followers' club. So I choose a trend-following system that will always be there in the market; either long or short. And the system is traded without any Stop-Loss defying all general assumption of Stop-Loss concept. Also there is no manual closing of position during a big news day such as Fed announcement etc.

The rest of this article will describe what I have found in this testing days with GOLD Futures. One very important point, I shall not mention the Trading System Logic, mathematics and indicators behind this particular MCX GOLD Futures Trading System. This is very much confidential to the success of this system and some parameters are proprietary to me only.

Here I shall only analysis the data that I have gathered from July 2013 to December 2013 and decide whether I should trade this Live or not.

And the method of analysis I shall discuss here will enable readers to evaluate their own systems in a complete mathematical point of view so that beginners as well as professional traders can analyze their own system and find out if it is suitable for trading in the long run. And that is the aim of this document.

And to the benefit of maximum traders, I am posting this to two very well-known trading forum.

The MCX GOLD system is designed in Amibroker. Everybody wants to know at first hand, what is the generated profit? So in next page, there is the result for the first 65 trades from 1st July 2013 to 24th December 2013.

Actual entry price is calculated with 6 points spread difference, equally divided in 3 + 3 points in entry and exit for Gold Future.

For example, if Buy signal is there at 28008, purchase was made at 28011 and if Sell signal is there at 28008, sell was made at 28005. In practice, as the GOLD future is very liquid, I have seen that this spread varies between 2 to 8 points, so on average 6 points spread is a very good estimation.

And the net profit is calculated after deducting all government fees and low-cost brokerage like Zerodha.

Rollover at the contract end is done mostly 1 day prior to the calendar month end to avoid last day unnecessary volatility.

This 65 trades occurred during roughly 6 months period from July to December 2013. During this time, we have seen some very good trending market as well as directionless sideways market. Although you can say just 6 months of data is not enough and it could be more like some sort of curve-fitting - that's just one part of the system design if I agree with you.

It is sure market dynamics changes over long time. For example, when you enter a sideways market? What kind of sample trades did you expect to get in that market conditions? You never know unless you have tested your system in that market condition and gathered a new set of trade samples. For example, a trend-following system like this may perform very good during a trending market but performs badly in sideways market. So we take here a different approach.

Fortunately, in this 6 months of observation contains both trending as well as sideways phase.

Statistician says that you will need at least, yes, at least 30 trade sample as bare minimum to even consider calculating effectiveness of a trading system. More is better. At least 100 trade samples. Much better and robust is 100 trade sample from all kind of market conditions. But we have 65 trade samples. Consider this is good enough than bare minimum required. So below is the trades.
 
#4
Hello Gulu

Nice thread . . . :thumb:

I have started trading a system based on daily TF which was back tested on 41/2 years of data
and coincidentally it gave me just 67 trades from back testing . . .

Regarding the sample size for the number of trades I feel the figure 100/or whatever is not so much relevant,
perhaps the more important factor would be to ask, have we covered maximum number of different scenarios that can occur . . .

Ideally we should be saying all possible scenarios,
but then there is no way to define this illusive all possible :)

With my back test, my view was 67 is a v small sample size.

When I did a bar by bar walk-though I was trying to look at different possibilities for what ifs.
Putting the system through the rigor of testing for all known/imaginable what if scenarios, and
walking through the entire know history of the ticker made me feel better but still there was
this lingering thingy, before taking it live, about should have had more back-data for testing.


Looking forward to your trade list and other posts detailing various aspects of evaluating the performance of our trading systems.


One aspect that you have touched upon is about curve fitting the system to the past data.
They say for avoiding or minimizing this one needs to keep the degree of freedom to minimum.
i.e keep to minimum the number of variables that are used in decision making
Not sure how we can truly achieve this? Maybe you can also say more about it.


One again thanks for the thread.


Happy :)
 

GuluGulu

Well-Known Member
#6



Above are the Trade results.

Here the starting capital was taken as the cash value of one GOLD Futures contract. 1 GOLD FUT is 1000 Grams and that is roughly rounded-off to INR 30,00,000. So we assumed the capital is that much.

Although you will need around INR 150,000 to hold such one GOLD Future overnight., that's just around 5% of the total cash value.

After trading 6 months, the system made a profit of around INR 10,50,000, deducting all spread, brokerage and expenses.

This is around 35% profit in 6 months in cash value. If you have a capital of just double the margin required, i.e. INR 300,000 the profit is 350%. These figures are very very lucrative.
So I need much more mathematical interpretation regarding validity of the system before I intend to put real money in it.

In the above table, there is total 65 trades out of which winning trade is 39 and loosing trade is 26.

So Winning percentage is (100 x 39) / 65 = 60%
And Loosing percentage is (100 x 26) / 65 = 40%

This means till now, suppose you are in a trade. The chance of success of this trade is always 60%. In fact, any trade you take among the distribution, chance of success is always 60% whether it eventually becomes a profitable or loss trade - mathematically that doesn't matter.
 

augubhai

Well-Known Member
#8
Hi GuluGulu,

Thanks for starting this thread. I am looking forward to how the system performs....

However, I would like to mention a couple of points.

1. The 30 data points is good enough confidence for a normal distribution. However, price does not follow a normal distribution. Many trading systems like arbitrage, some option strategies give close to normal results, but trend following systems are generally fat tailed distributions.... http://en.wikipedia.org/wiki/Fat-tailed_distribution

2. When considering the win rate, also consider the probability of a string of losses. The probability of a large drawdown is inversely related to the win %. One tool that i use is http://www.sbrforum.com/betting-tools/streak-calculator/, but even this does not calculate the largest peak to valley drawdown.

I am trading a low win% system, and am going currently through a huge drawdown... :)
 

anuragmunjal

Well-Known Member
#9
The rest of this article will describe what I have found in this testing days with GOLD Futures. One very important point, I shall not mention the Trading System Logic, mathematics and indicators behind this particular MCX GOLD Futures Trading System. This is very much confidential to the success of this system and some parameters are proprietary to me only.


Sir

If I may ask, what is the purpose of this thread?

regards
 

GuluGulu

Well-Known Member
#10
Hi Angu,

Saw your performance in one of your thread here:

http://www.traderji.com/trading-diary/90116-dasara-system-63.html

This approach of mixing a lot of systems may eat your trading capital slowly even if you ended up positive. Please don't do that. Instead, please analyze your methods first.

Hello Anurag,

The purpose of this thread is to educate common traders like Angu, to scientifically analyze his or her system whether it could be profitable or not in the long run. My MCX Gold System is just being used here as an example. Please go through it as I write and you will understand slowly.

I have seen all traders jump from system to system in search of Holy Grail. I have done this too once upon a time.

Then I realized there is no Holy Grail in the trading system. But of course it is there - it is inside 'You'. You are the Holy Grail. Your system is just 10 to 15% of that - whether you use ORB, SAR, MA Cross or whatever you use as long as you have positive expectancy.

From that point of my realization, I started marching towards the avenue of a Professional Trader and never looked back.

Give a same system to 10 people - and I bet results will be different for all 10 people.

So, please wait for my next post and judge yourself.
 

Similar threads