Advice requested of this system

D

Dhalsim

Hi guys, i have been a discretionary trader for 5 years now and this is first time i have created an automated system. Don't know if i should run it live as it performs far batter during high market volatility and has some glaring issues. I do know that i will never find the perfect system but i am currently more interested in advice regarding my testing process and results.

Anyway guys have a look, here are some of the stats:
Each trade has commission already added at IB $2.40 for futures.

Also, even though entries are all limit orders and exits are limit orders i am going to factor in $5 slippage per round-trip. This is system was tested on the Emini NQ.


Monthly PnL Entries per month
Jun-07 -$1,611.80 16
Jul-07 $2,157.60 13
Aug-07 $2,283.80 19
Sep-07 -$557.00 15
Oct-07 -$266.40 18
Nov-07 -$702.80 11
Dec-07 $1,182.40 12

Jan-08 $1,171.00 15
Feb-08 -$767.20 14
Mar-08 $258.40 17
Apr-08 $1,793.60 18
May-08 -$847.20 14
Jun-08 $2,163.60 18
Jul-08 -$2,816.40 18
Aug-08 $1,728.40 17
Sep-08 $3,687.60 13
Oct-08 $5,222.60 13
Nov-08 $3,362.20 11
Dec-08 $882.80 14

Jan-10 $122.80 14
Feb-10 $628.20 16
Mar-10 -$572.20 14
Apr-10 $463.60 18
May-10 -$1,171.40 18
Jun-10 $1,918.00 15
Jul-10 $218.40 17
Aug-10 -$1,282.00 15
Sep-10 $1,408.20 16
Oct-10 -$147.20 14
Nov-10 -$217.00 15
Dec-10 $18.00 15

Jan-11 $692.80 14
Feb-11 $788.00 15
Mar-11 $8.00 15
Apr-11 -$257.20 14
May-11 -$671.80 16
Jun-11 $633.20 16
Jul-11 $1,682.60 13
Aug-11 $1,563.60 18
Sep-11 $5,408.20 16
Oct-11 $2,843.20 16
Nov-11 $2,268.20 16
Dec-11 $1,267.80 14

Jan-12 -$837.00 15
Feb-12 -$382.20 14
Mar-12 $1,363.60 18
Apr-12 1268.40 17

Net $37,352.00
Avg monthly $1,037.56


Years Annual PNL Entries Avg Per trade
2007 $2,485.80 104 $23.90 (only 7 months tested 2007)
2008 $15,839.40 182 $87.03
2010 $1,387.40 187 $7.42
2011 $16,226.60 183 $88.67
2012 $1,412.80 64 $22.08 (only 4 months tested 2012)


Losing months 16
Winning months 31
Monthy W/L 65.96%

Based on expectation of $20,000 account start Start accnt $20,000.00

Trade Data Number Net Percentage
Largest Winner $2,480.20 12.40%
Largest Loser $1,204.80 6.02%
Wins in a row 11 $4,457.20 22.29%
Losses in a row 5 $1,134.00 5.67%
Biggest Draw-down 31 trades -$3,286.00 -16.43%

Total Gain/Loss after commission and slippage $37,352.00
avg per trade $51.88
Total trades 720
Winners 431
Losers 289
W/L 59.86%

Net Gains $118,940.00
Net Loss $73,527.00
Profit Factor $1.62
standard dev $367.08
confidence level 3.79
confidence level 99.40%

I have some better performing systems in the pipeline but this one has been fully tested and all the data has been validated. I am not looking for the perfect system as i know it does not exist so i have not tried to optimize the data much and system plays of very few variables.

I did randomize the data and equity curves all were fairly decent with max draw-down of 23% after 20 randomized data sets.

I just need some advice from experienced guys of whether i am going in the right direction with my system design or am i missing something. What should i be aiming at for avg$ per trade?? Also, am i missing something in the testing process?

I have ran future tests on the data based on what i read of few threads on ET. I am no expert on stats or programming hence my trepidation when it comes to having confidence in my results.

I plan to run the system live in the future with small account and at the same time run the system under simulated conditions so i can see if i will get similar results.

I have attached some charts for you guys to have a look at.
 

Attachments

It looks like that your system makes no trades for all of 2009. Something is not right. What timeframe are you working with? By the way, the low profit factor is not acceptable in professional environments.
 
Quote from intradaybill:

It looks like that your system makes too few trades from Dec 2008 to middle of 2010. Something is not right. What timeframe are you working with? By the way, the low profit factor is not acceptable in professional environments.
More...

Sorry I forgot to write that no data was collected for 2009 hence why no data.
 
Quote from Dhalsim:

Sorry I forgot to write that no data was collected for 2009 hence why no data.
More...

Also, can you explain what you think is an acceptable profit factor and why. I just read this post on another board and think it makes some sense, what do you think??

"Some of the most robust simple trading methods have profit factors of between 1.2 and 1.6 which are highly successful. Why you would put 3 to 10 on the scale is beyond me because the overall longevity of such PFs is really scarce, if available at all. For a PF of 1.5 you are averaging a return of $1.50 for every $1 you risk which over the greater timespan is phenomenal. To point out another PF fact, the game roulette at the casino has a 1.05 PF in favour of the house. That measly 5c turns into 10s of millions every year so you can see that a profit factor of 3 or more isnt really necessary unless your an adrenline junkie and your system calls for overleveraged over risked trades on a short term, the long term will usual unvail these systems to be less than 1.

Backtest some of your systems (with current PFs of 3 or more) for the past 30 years and you will see just how well your system compares through the boom and bust cycles, then compare them to some of the more tried and robust systems that have been in action for that long."
 
Quote from intradaybill:

So why you do not have any data for 2009?
More...

Well when is started collecting the data i wanted to take sample of couple of months randomly for last 5 years but then as i went on i decided to see how data would play out if collected till the end of year. Just by chance no data was collected for 2009 initially and therefore that data year was not expanded outwards later when i continued collecting data.

Its just random that i decided not to collect 2009 data, i could do it now but only issue is that my data is complicated in a simple way and needs to be validated hence would require even more effort to collect. I will get back to validate after i finish working on some other systems.
 
Its a good start. PF 1.6 is decent, I personally prefer > 1.8.

Reason why a PF > 1.6 is absolutely required is that unlike roulette, trading systems can lose their edge in future. IMO, a PF of 1.1 is also ok in the context of a high frequency system, since system would be taking 1000s of trades every day. So statistically, a trader should have no losing days. As soon as trader starts to see a 2% or 5% losing days, it clearly indicates that system's PF is going down, maybe strategy is dying. But for systems that trade 1-3 times a day, a PF > 1.5/1.6 is useful. I hope you get what I am saying here.

It will be helpful if you can attach an image file and remove the numbers after decimals.

One thing that strikes is that you will need to have another system with this one, for otherwise in periods of low vol, you will find it difficult to stick to the system. Since it won't make any money for extended periods of time say for 8-9 months or so. It will be very frustrating indeed and chances are you will modify the system or do some other mistake. So you need couple more systems so that equity curve of portfolio of systems is smoother.
 
Thanks for the advice, i have 5 systems which i have some preliminary testing on but they all take quite some time to develop fully.

However, it was exactly as you said that i will need more systems, my plan is too start 100,000 and have 5 systems with positive expectancy running at the same to smooth out the equity curve and allow systems performing weakly to ride out the bad periods.

All my systems are intra-day and each system never trades more than once per day.

Is their anything else i am missing or am i on the right lines here, with all things considered would this therefore be a worthwhile strategy to run alongside various other strategies? Or is it simply just a little weak?

Here is data with out the confusion:

Monthly PnL
Jun-07............... -$1,611.80
Jul-07............... $2,157.60
Aug-07............... $2,283.80
Sep-07............... -$557.00
Oct-07............... -$266.40
Nov-07............... -$702.80
Dec-07............... $1,182.40

Jan-08............... $1,171.00
Feb-08............... -$767.20
Mar-08............... $258.40
Apr-08............... $1,793.60
May-08............... -$847.20
Jun-08............... $2,163.60
Jul-08............... -$2,816.40
Aug-08............... $1,728.40
Sep-08............... $3,687.60
Oct-08............... $5,222.60
Nov-08............... $3,362.20
Dec-08............... $882.80

Jan-10............... $122.80
Feb-10............... $628.20
Mar-10............... -$572.20
Apr-10............... $463.60
May-10............... -$1,171.40
Jun-10............... $1,918.00
Jul-10............... $218.40
Aug-10............... -$1,282.00
Sep-10............... $1,408.20
Oct-10............... -$147.20
Nov-10............... -$217.00
Dec-10............... $18.00

Jan-11............... $692.80
Feb-11............... $788.00
Mar-11............... $8.00
Apr-11............... -$257.20
May-11............... -$671.80
Jun-11............... $633.20
Jul-11............... $1,682.60
Aug-11............... $1,563.60
Sep-11............... $5,408.20
Oct-11............... $2,843.20
Nov-11............... $2,268.20
Dec-11............... $1,267.80

Jan-12............... -$837.00
Feb-12............... -$382.20
Mar-12............... $1,363.60
Apr-12............... $1268.40

Net $37,352.00
Avg monthly $1,037.56


Quote from gmst:

Its a good start. PF 1.6 is decent, I personally prefer > 1.8.

Reason why a PF > 1.6 is absolutely required is that unlike roulette, trading systems can lose their edge in future. IMO, a PF of 1.1 is also ok in the context of a high frequency system, since system would be taking 1000s of trades every day. So statistically, a trader should have no losing days. As soon as trader starts to see a 2% or 5% losing days, it clearly indicates that system's PF is going down, maybe strategy is dying. But for systems that trade 1-3 times a day, a PF > 1.5/1.6 is useful. I hope you get what I am saying here.

It will be helpful if you can attach an image file and remove the numbers after decimals.

One thing that strikes is that you will need to have another system with this one, for otherwise in periods of low vol, you will find it difficult to stick to the system. Since it won't make any money for extended periods of time say for 8-9 months or so. It will be very frustrating indeed and chances are you will modify the system or do some other mistake. So you need couple more systems so that equity curve of portfolio of systems is smoother.
More...
 
You must log in or register to reply here.
Back
Top