System Testing

[suri112000]

I am testing two systems in order to assess whether they can be traded in isolation or combindly. The parameters are as follows :-

Timeframe : 15 minutes
Settings : Usual
Outlook : Technical
Trades : Longs and Shorts
Instrument : Banknifty
Bias : Trend (Short term trends)
Stops : No stops.
Duration : Always in trade whether long or short.
Capital : Rs.5,00,000
No of lots : 3
Result : Monthly basis.........not based on trades in isolation.
Slippage and brokerage : 10 points deducated for a round trade

Jan 2015

System 1 (Swing)
Profit = Rs.38,377 (ie 7.6% on Rs.5 lakh capital)
40% win rate with a total of 10 trades.
Drawdown (Hole in capital): 31,428.

System 2 (Agrawal)
Profit = Rs.1,73,872 ( ie 34% on Rs.5 lakh capital)
66% win rate with a total of 18 trades.
Drawdown (Hole in capital) : NIL

Combined Profit = Rs.2,12,249 (21%)

Feb 2015

System 1 (Swing)
Profit = Rs.64,922
45% win rate with a total of 11 trades.
Drawdown (Hole in capital): 29,905

System 2 (Agrawal)
Profit = Rs.82,910
52% win rate with a total of 25 trades.
Drawdown (Hole in capital) : NIL

Combined Profit = Rs.1,47,832 (14%)



Mar 2015

System 1 (Swing)
Profit = Rs.76,222
77% win rate with a total of 9 trades.
Drawdown (Hole in capital): 30,695

System 2 (Agrawal)
Profit = Rs.28,576
42% win rate with a total of 19 trades.
Drawdown (Hole in capital) : Rs.56,255

Combined Profit = Rs.1,04,798 (10%)



Apr 2015

System 1 (Swing)
Profit = Rs.4,207
33% win rate with a total of 9 trades.
Drawdown (Hole in capital): 19,140.

System 2 (Agrawal)
Loss= Rs.1,02,532
29% win rate with a total of 27 trades.
Drawdown (Hole in capital) : Rs.1,25,476

Combined Loss = Rs.98,325 (-9.8%)

 

[ROCKSTAR55]

Hi friend, are you using EMA crossovers for this strategy?

 

[suri112000]

Hi friend, are you using EMA crossovers for this strategy?

Its a bit complex than simple EMA crossovers but the essence is the crossover.

 

[suri112000]

1 Use a mechanical system. The trading system must be 100% mechanical without any human input or overrides. It must also not be tweaked or adjusted as time goes on to fit current data. In addition, the system's rules must not be curve-fitting to short-term, non-repetitive patterns of past data that eliminate otherwise losing trades. A good way to screen for curve-fitting is to look for consistently good results over a minimum of five years of past data that meet all of the other criteria outlined below.

2. Maximum drawdown. There are two aspects to consider: the amount of the drawdown (it should not exceed a small percentage of the account value) and the duration of the drawdown until a new peak is realized (should not exceed six months). Some trading systems hype great profits over the past several years but don't disclose drawdowns that exceed the initial capital invested and last for a year or more. Before selecting a trading system, you must be able to quantify the drawdown risk and find it acceptable, both financially and emotionally.

3. The account size. The maximum past drawdown (over a minimum five-year period) plus the margin required for one contract is the absolute minimum account size required to trade a system. To be conservative, it is prudent to add a buffer, since the maximum drawdown for any trading system is always in the future.

4. Calculate annual returns. Two things are important here. First, the average annual net profit should be a minimum of twice the maximum drawdown of the past five years. Second, ideally there should be no losing years.

5. Trade profile. There are two important aspects here. First, the percentage of profitable trades should be in the 40% to 60% range. The ratio of average win to average loss should be in the 1.3 - 2.0 range. Second, the average trade net profit (total net profits divided by the total number of all trades) should be at least three times greater than real-world per-trade slippage and commission assumptions. Beware of systems claiming to deliver greater than 60% winners. Such systems usually exhibit a very poor average-win-to-average-loss ratio, where a few losing trades can wipe out profits from several winning trades.

 

[suri112000]

Hi...friends,

I have been searching for a mechanical system since start of Jan 2009 with a minimum of 60% win ratio, a profit factor 2 or above, a maximum system drawdown of less than 20%, a single trade draw down of 6% (a worst case scenario of gap against the trade), sharpe ratio of 2 or above. Either the system could be an AFL or chart pattern/price action with a well defined rules.

I have checked up hundreds of AFLs available over the internet/forums. None of them met my criteria.

My system of checking the system is very simple. I check the system on two years of data. Say from 2010 to 2011. If the result satisfies my criteria, then I check the system on random data choosen from 2012 to 2015. (May be three months or five months etc). If the result satisfies my criteria, I consider the system as robust and tradable.

Recently, I met a Trader-cum-system developer who has given me an AFL converted into a plugin (so that we cannot see the actual code of AFL). This AFL is astonishingly satisfies my criteria. I thought that it is the AFL I have been looking for all these years. I asked him to give it to me. He said it costs Rs.2,00,000/-. Whether I am going to buy it or not that is another issue. But, see the backtesting report, what the AFL is giving on 15 min TF on banknifty by suitably deducting slippage, brokerage, taxes. My idea is here that.....some good AFLs are existing but we need to spend lot of energy and time to develop it.

SEE THE BACKTESTED REPORT FROM 2010 TO 2011.

 

[jagankris]

Hi...friends,

I have been searching for a mechanical system since start of Jan 2009 with a minimum of 60% win ratio, a profit factor 2 or above, a maximum system drawdown of less than 20%, a single trade draw down of 6% (a worst case scenario of gap against the trade), sharpe ratio of 2 or above. Either the system could be an AFL or chart pattern/price action with a well defined rules.
----------------------


Hmm this seems to be a holy grail trading system :thumb:
Can you please share the backtesting results YOY till date ?

 

[Smart_trade]

One should fully convince himself that the system works in different sample periods.If it gives similar performance in different samples then it may be worthwhile testing it on different scrips such as Nifty,stocks futures etc....a good system normally works well in other counters too.

If the system is giving this consistant performance, then ask him to let you test the system for 2 months in real markets. Many times systems test great on system testers but fail in actual trading , so whether we can do all trades as per system is also important.

If the system satisfies all the above ( I am still skeptical though) then Rs 2,00,000 is a small price for that system.

Posting just to caution if someone is thinking of buying it....be 100 % sure that it will make money after you buy and trade it.

Why I am saying this is I know 2-3 cases of systems designers who designed good systems...but they failed to make any money from the markets trading on their own systems......they took others money to manage thinking that the systems will make tons of money for them and for the trader and ended up with huge losses in those accounts...so this caution...not to dampen the enthusiasm.

Smart_trade

 

[suri112000]

Hi...friends,

I have been searching for a mechanical system since start of Jan 2009 with a minimum of 60% win ratio, a profit factor 2 or above, a maximum system drawdown of less than 20%, a single trade draw down of 6% (a worst case scenario of gap against the trade), sharpe ratio of 2 or above. Either the system could be an AFL or chart pattern/price action with a well defined rules.
----------------------


Hmm this seems to be a holy grail trading system :thumb:
Can you please share the backtesting results YOY till date ?

I donot have BNF and NF futures data from Apr 2012 to Dec 2014 to backtest. If you have please upload, I will test it.

Whatever futures stocks or indices data I am touching (Of course which is available with me) it is giving super results. I have test it with bar replay also to ensure that the AFL is not looking into future data to adjust the signal.

As suggested by Smart-trade, it should tested live for atleast 2 months before deciding.

 

[Blackhole]

I donot have BNF and NF futures data from Apr 2012 to Dec 2014 to backtest. If you have please upload, I will test it.

Whatever futures stocks or indices data I am touching (Of course which is available with me) it is giving super results. I have test it with bar replay also to ensure that the AFL is not looking into future data to adjust the signal.

As suggested by Smart-trade, it should tested live for atleast 2 months before deciding.

try this for data.

http://www.traderji.com/data-feeds/89889-nifty-historical-ieod-2008-till-date-free.html

there is for one more but not sure of data format...just check it

http://www.traderji.com/data-feeds/11267-intraday-1-minute-future-data-8.html


system testing


afl into dll or any plugin..... can be easily manipulated.

metrics changes even if u do position sizing or change fix position say from 50 to 250....u might see drastic swing in all the metrics or can ruin the entire capital within a year

i personally believe in

backtesting on atleast 7-10 years of data (for nifty i would prefer 2007-8-9) mandate.

use a fix lot size(for nifty and banknifty same lot size might mess up results)

optimizing the afl for best parameter then randomly backtest on 3 m 6m different data sample.

walk forward test in very essential before going live. Please make sure walk forward test is done based on different parameters kratio / ulcer index /car-mdd.

not to forget Monte carlo test....new version of amibroker has it....can be done on excel sheet as well.

---------

for instance supertrend nowadays is popular and freely available on some popular sites.

i check past few months it was giving u pity good loss. select 3 months sample optimize(or) randomly change atr/factor in afl and do backtest . U would be surprised by results.


Real trades and results are very different from afl results. Its trader that trades the system good trader controls drawdowns survives and impoves on other factor as he enjoys the process. Consistent performance is needed from trader not the afl. As afl can give u double the continuos lossing trades without giving u a profitable trade( exhausting ur capital adding to ur mental stress) after that u might not like the idea of taking the winning trade and repent afterwards.




sorry for this post, i know u might already know all this.

happy trading

 

[jagankris]

I donot have BNF and NF futures data from Apr 2012 to Dec 2014 to backtest. If you have please upload, I will test it.

Whatever futures stocks or indices data I am touching (Of course which is available with me) it is giving super results. I have test it with bar replay also to ensure that the AFL is not looking into future data to adjust the signal.

As suggested by Smart-trade, it should tested live for atleast 2 months before deciding.

Suri,

Pls find below the data uploaded in 4shared (dot) com

BNF - 2012 - 2013 2 years data.
Don't have the 2014 data.

http://www.************/office/eWHECh1Ice/BANKNIFTY_F1.html

Nifty 2014 data
http://www.4shared (dot) com /office/MCNlTua3ce/NIFTY_F1.html

For now let us know the results. Eagerly waiting :thumb:

Will get the missing data from fellow members.