Reply

Portfolio Effect Simulation / Money Management

6 replies

MFXS

Subscriber, bbp_participant, community, 9 replies.

Visit profile

7 years ago #116291

When a colleague put me on to QA a few days ago, I thought I’d found the holy grail. I thought to myself ‘no more combining backtest results in spreadsheets and manually calculating growth, drawdown and stagnation – this is too good to be true!’ … Sadly it was!

 

Now this thread has been done multiple times by different users, but the issue has still not been addressed as far as I can tell. The portfolio analysis component of QA is essentially useless until proper portfolio effect simulation is implemented. Previous responses to older threads seem to either say A. it’s impossible / too hard or B. no need due to the money management component.

 

Regarding A: I agree that implementing this perfectly, ie calculating new lot sizes using open and close time of trades would be difficult (though far from impossible!), but a basic simulation would be incredibly simple and much more accurate than what we have currently. QA currently calculates portfolio profit and loss by adding the $ value results of trades correct? All we need to do is get the program to use the % gain / loss values instead … I don’t know Java, but I imagine you guys could code this in to an extension extremely quickly! There is no need to recalculate lot sizes with this method, we are simply using the original % figures to calculate profit / loss instead of $ values.

 

Now granted, this solution is not perfect in that often a trade is opened before the preceding trade is closed etc … but I maintain this work around would be infinitely more accurate and useful than what we are currently working with.

 

Now regarding B, Money Management: As far as I can tell, this component simply does not work on strategies with dynamic stop losses. Irrespective of whether I ‘Recognize SL from orders’ or not, when I run a ‘Fixed % of Balance’ or ‘Fixed % of Account’ simulation and then send the results to data bank and analyze, the % risked on each trade always fluctuates and the fluctuations are quite substantial. I’m really not sure as to why this doesn’t work, as SL info is readily available in MT4 backtests and QA even ‘recognizes’ this info as shown by the calculation of average SL and max SL.

 

I would really like to purchase this program, but until one of the above issues is addressed there is not much point. On the other hand, if one of these issues were addressed, I would definitely subscribe for life and tell anyone and everyone in the world of auto-trading about it. This program has so much potential and has obviously had a ton of work put into it – let’s go the final yard!

 

Cheers

0

tomas262

Administrator, sq-ultimate, 2 replies.

Visit profile

7 years ago #141463

Hello,

 

thanks for your response and suggestions. Will notify developers about this

0

mabi

Customer, bbp_participant, community, 261 replies.

Visit profile

7 years ago #141478

SInce computing open equity using historical minute data was added  to QA some time ago calculating new lot sizes using open and close time of trades ought to be possible.

0

mikeyc

Customer, bbp_participant, community, 877 replies.

Visit profile

7 years ago #141482

Can someone please explain what is wrong with the current implementation?  The lot size is currently equal for all portfolio component strategies, what is it calculating incorrectly?

0

MFXS

Subscriber, bbp_participant, community, 9 replies.

Visit profile

7 years ago #141637

Hello,

 

thanks for your response and suggestions. Will notify developers about this

 

Thanks Tomas, that would be great. Really keen to buy this program, but can’t justify til this is addressed.

 

SInce computing open equity using historical minute data was added  to QA some time ago calculating new lot sizes using open and close time of trades ought to be possible.

 

^What he said. Though simple stop gap solution should be implemented in meantime.

 

Can someone please explain what is wrong with the current implementation?  The lot size is currently equal for all portfolio component strategies, what is it calculating incorrectly?

 

Okay, so if your strategies use fixed lot sizes there is no problem. Also, if you’re running, for example, 4 strategies which use balance proportionate sizing (MM), but are run on separate balances (10K per strategy / 40k total) there is also no problem.

 

Problem is when you are running 4 strategies which use MM on a single balance / deposit. At this point the results are completely corrupted. Profit and more importantly, Drawdown are grossly understated.

 

Here is what QA says would happen if I combine my two strategies which are supposed to risk 1.5% per trade (but don’t):

… and here is a more accurate simulation created by bouncing out list of trades to excel and correcting risk per trade manually before bouncing back into QA:

… Note QA’s default simulation understates Drawdown by 30%, CAGR by close to 50% and total profit by nearly a factor of 3! The results from the default output are completely corrupted and inherently useless. Do you see the problem?

0

mnlpad

Subscriber, bbp_participant, community, 7 replies.

Visit profile

6 years ago #142491

 Hi,

 

I noticed that risk MM in % do not work well. The size of the position is not well calculated to get the input of risk placed. There is other setting that i need to adjust?

 

Thanks.

 

  

0

Alejandro

Subscriber, bbp_participant, community, 0 replies.

Visit profile

6 years ago #197284

I’m surprised this issue hasn’t been fixed. It’s critical and it seems that people don’t get it. This should be fixed when QA detects the stop loss for each trade and calculates the position size taking into consideration the initial stop loss for each trade. I’m not an expert in java but this seems pretty easy to implement. Just detect the stop loss when each order is executed. The information is available in MT4 reports. If this is done I would happily buy this product

0

Viewing 6 replies - 1 through 6 (of 6 total)