I am very puzzled on risk percentages to assign to a group of systems.
Tests show that every system has its own best risk %.
Giving differents weights, testing results looks better.
On the other hand, like everyting in testing, systems do not perform always at their best, knowing nothing about the future, seem to me more wise to give the same risk.
Any thought on this?
thank you
Systems weights
Some people want the individual systems within a suite (a group of systems traded simultaneously) to be balanced. You can think up lots of different system characteristics that you might want to balance:
*perhaps, the equity curve of the Credit Suisse/Tremont Hedge Fund Index. Or the equity curve of the Barclay CTA Index. Or the equity curve of Winton Capital.
##either the measured value from simulating the system, or the Monte Carlo 95% Confidence value from iterating the system 20,000 times.
- Adjust each individual system's risk-per-trade so that every system produces the same CAGR% ##.
- Adjust each individual system's risk-per-trade so that every system produces the same MaxDD% ##.
- Adjust each individual system's risk-per-trade so that every system produces the same Daily Standard Deviation of Returns %.
- Adjust each individual system's risk-per-trade so that every system has the same long term average value of the Total Risk Profile graph produced by Blox (shown in Figure 1 below)
- Choose a reference equity curve (perhaps, Buy-and-Hold-the-SP-500)*. Measure the correlation between each individual system's equity curve, and the reference equity curve. Weight each individual system's risk-per-trade inversely to its correlation with the reference. Lower correlation, higher weight; higher correlation, lower weight
*perhaps, the equity curve of the Credit Suisse/Tremont Hedge Fund Index. Or the equity curve of the Barclay CTA Index. Or the equity curve of Winton Capital.
##either the measured value from simulating the system, or the Monte Carlo 95% Confidence value from iterating the system 20,000 times.
- Attachments
-
- Figure 1. Blox measurement of a system's Total Risk Profile
- trp.png (9.39 KiB) Viewed 3605 times
Systems weights
Hi Sluggo,
Please excuse the late response-- I am a recent member and found your interesting post.
I came across your post regarding matching different systems to a criteria.
Some of the criteria, such as matching the drawdowns, I understand as a way to have commensurate risk (given the known data) per system. Makes sense to me.
Matching CAGR -- what's the purpose of that?
Finding non-correlation to a benchmark is desirable -- but from my perspective, seems that choosing the second system or system parameter set should be done intentionally to have a different characteristic / non-correlation, not just selecting non-correlated results which could be the result of random selection. What I mean is it shouldn't be results alone that drive the selection process, but knowing that the second system has a characteristic that makes it likely to be different / non-correlated.
What do you think of that?
Do you have any additional thoughts?
Regards,
-Michael
Please excuse the late response-- I am a recent member and found your interesting post.
I came across your post regarding matching different systems to a criteria.
Some of the criteria, such as matching the drawdowns, I understand as a way to have commensurate risk (given the known data) per system. Makes sense to me.
Matching CAGR -- what's the purpose of that?
Finding non-correlation to a benchmark is desirable -- but from my perspective, seems that choosing the second system or system parameter set should be done intentionally to have a different characteristic / non-correlation, not just selecting non-correlated results which could be the result of random selection. What I mean is it shouldn't be results alone that drive the selection process, but knowing that the second system has a characteristic that makes it likely to be different / non-correlated.
What do you think of that?
Do you have any additional thoughts?
Regards,
-Michael
Different traders have different goals and different approaches toward reaching those goals. I am sure there are many more ways to "balance" the individual systems within a suite, than the five mentioned in my post a month ago.
Each trader will need to answer the question: what does "balance" (used as a verb) mean, TO YOU? What is it that you wish to "balance"?
Trader Alfred might decide: "I want every system to contribute the same amount to the bottom line. Each must pull their own weight." And so Alfred could very well choose to adjust betsizes so that each system has the same CAGR (in historical testing).
Trader Barbara, on the other hand, might think: "It will be very easy for me to stick to my system-suite if it generates huge profits. That's a no-brainer. I can also stick to my system-suite if it generates small profits; I'll be disappointed but not destabilized. I even think I can stick to my system-suite if it generates small losses: God knows I've lived with small losses in my non-systematic investing and trading! But what concerns me is, can I stick to my system-suite if it generates large losses?? I doubt it. Therefore when I design my system-suite I will try very hard to avoid large losses! I will adjust betsizes of the individual systems in my suite so that all of them have the same Max-Drawdown-Percent (in historical testing). I refuse to let one system have a bigger and more crippling drawdown than all the others."
It's all very human. Human traders design mechanical systems, to maximize human pleasure and to minimize human pain. Their reasons Why, are human reasons. Sure, we use tools (historical data, software, computers) and we perform analysis (statistics, etc.), but at the end of the day we're simply trying to maximize gain and minimize pain. Since each of us has a slightly different definition of gain and pain, each of us will probably trade differently.
Each trader will need to answer the question: what does "balance" (used as a verb) mean, TO YOU? What is it that you wish to "balance"?
Trader Alfred might decide: "I want every system to contribute the same amount to the bottom line. Each must pull their own weight." And so Alfred could very well choose to adjust betsizes so that each system has the same CAGR (in historical testing).
Trader Barbara, on the other hand, might think: "It will be very easy for me to stick to my system-suite if it generates huge profits. That's a no-brainer. I can also stick to my system-suite if it generates small profits; I'll be disappointed but not destabilized. I even think I can stick to my system-suite if it generates small losses: God knows I've lived with small losses in my non-systematic investing and trading! But what concerns me is, can I stick to my system-suite if it generates large losses?? I doubt it. Therefore when I design my system-suite I will try very hard to avoid large losses! I will adjust betsizes of the individual systems in my suite so that all of them have the same Max-Drawdown-Percent (in historical testing). I refuse to let one system have a bigger and more crippling drawdown than all the others."
It's all very human. Human traders design mechanical systems, to maximize human pleasure and to minimize human pain. Their reasons Why, are human reasons. Sure, we use tools (historical data, software, computers) and we perform analysis (statistics, etc.), but at the end of the day we're simply trying to maximize gain and minimize pain. Since each of us has a slightly different definition of gain and pain, each of us will probably trade differently.