I'm not using the new indi (see previous posts) but if you want an 80% T1 win ratio:-
H1
12 bars
40 pip range
4 pip buffer
2 min bars outside
Extracted Thread (per thread starter's request)
- #5,464
- Edited Feb 19, 2013 4:42pm Feb 18, 2013 6:13pm | Edited Feb 19, 2013 4:42pm
Good day everyone,
I have been a regular reader of this thread for a while now. I do not trade optimized systems as a rule. However, I have been interested in the fact that this modified volatility expansion strategy (which I can testify is a sound trading concept) may also reach some kind of long-term profitability using optimization.
This method started off on promising grounds but as it unfolds, I find that posts after posts, many traps are still left open and, what is worse, are being held as core principles which are trusted by every one and used every day for trading. My contribution here is not meant to criticize the absence of mathematics in the system design. To each his own as long as it works. I rather feel compelled to highlight some crucial aspects so that people here may avoid this method's weaknesses - and ultimately lose money trading what may otherwise be a profitable one if soundly researched first.
What everybody needs to understand is that this volatility expansion strategy (VES) is first and foremost an exercise in optimization. Anyone may argue that this is a "naked system", that drawing boxes around CZs is what teaches you the edge etc...But doing so will only familiarize the user with the spirit of the method. The true system's mechanics rely on inducing profit stability through the identification of recurring and stable series of similar volatility patterns (not price patterns) born from the CZs. Given that this method essentially relies on 3 core parameters (CZ height, CZ length and TP), optimization becomes the one and only focus in the process of acquiring a stable, repeatable and trustworthy strategy. Claims may be made that one should not follow every single signal of the indicator, that the eye should decide what the best CZs are, that judgment calls are needed. But such claim inherently contradicts the initial premise of this strategy which tells us that the trader cannot and should not try to predict how the market will behave next - which is by the way the right bias to have.
Here are a few key remarks. I will try to be concise and specific.
PROPOSITION #1 - This Strategy has an Edge
Wrong. This strategy could have an edge indeed if it was not heavily parametric and did not require such serious optimization to approximate stability. Each new parameter set will invariably end up in another degenerating process to finish its course into a situation where the trader will not know how to make the difference between a mere drawdown and a broken trading logic. An edge, on the contrary, captures the dynamics of a given market class without any parameters or recurring back & forward tests. There is a way to decipher a drawdown from a broken trading logic when it comes to optimized system. I may address it in a later post though as there are more urgent aspects to focus on.
PROPOSITION #2 - A Forward Test is Key to Trade this Strategy Successfully
Inaccurate. The only upside of a forward test is to check whether the strategy can actually be executed with a decent level of replication versus theoretical results. It should never be used as a way to validate a given strategy or parameter set. Validation should have already been covered through the back test by the time a forward test is started. Otherwise, a forward test will bring nothing more than what a slow out-of-sample back test will, because this is exactly what this forward test will look like upon completion. A more efficient way to do it is to spare a large sample of data and test your optimized strategy over it. This will be immediate and will add the same statistical value to assessing your system's stability.
PROPOSITION #3 - One Should Pick the Settings that Suit Him
Wrong. EVERYBODY, regardless of its risk-aversion, should necessarily pick THE best combination of return versus risk. The rest is merely a question of money management. For every given instrument, time frame and data sample, there is one and only one set of parameters that is uniformly superior to all others. Not choosing it over the others has nothing to see with risk aversion or personal liking, but rather with flawed logic.
PROPOSITION #4 - The Tighter and the Longer the CZ, the Better
Wrong. Claims have been made of the above and also that a "sweetspot" should be found. These assertions are mutually exclusive. Finding the "sweet spot", without any "tighter CZ" bias, is the one way to go. The only filter that should be applied is that for the same risk/return combination and a similar pip profit, we should prefer the lesser number of trades so as to minimize the impact of spread cost. Apart from this notion, overlapping CZs are of no value as to tell whether a parameter combination is better or worse. The proof is below.
PROPOSITION #5 - SSBO is like Most Optimized Systems
Wrong. As I said, I do not trade optimized systems. But this is an answer to other threads making it a daily business in criticizing this one. The foundations of this VES are extremely sound. For this reason, and unlike 95% of the systems out there, positive results are not just available locally. In the case of SSBO, they are found across a wide array of parameter combinations and over solid historical data samples. This is where spectrum optimization comes into play. Because the profitability of optimized methods cannot be proven mathematically, the one and only way to show some evidence of consistency is to implement a long-term, comprehensive back test. Long-term means collecting at least 250 trades. Comprehensive means optimizing only the key parameters of the strategy, that is Height, Width and TP around the core trade of the strategy, that is the initial breakout trade, and using a neutral 1:1 risk-to-reward ratio to normalize outcomes. All other parameters are left fixed. Everything else linked to money-management such as martingale or other deceptive artifacts are foreign to a strategy's profitability and so they should be excluded from the analysis.
Below I used Bloomberg to quickly perform spectrum optimization over 400 days for a given time frame and a given pair, collecting at least 250 trades for each one of the 750 iterations that I investigated. The parameter combinations explored fit the average range over which contributors to this thread have focused on over time.The table shows each Gross Pip count.
Two conclusions jump out of this table.
(1) The BLUE CLOUD shows that there is clear stability of positive results across a wide but consistent neighborhood of parameter combinations (the "sweet spot") which allows for volatility conditions to vary greatly before a winning parameter set degenerates into a losing one. This is reassuring. The overall sanity check of the strategy is even more self-evident by looking at this 3D density display.
(2) The GOLD AREA in the table (bottom corner on the 3Ddensity) shows that in contrast with the sweet spot cloud, the assertion saying that "the Tighter, The Longer, The Better the CZ" is clearly pointing to a losing or sub-efficient set of parameter combinations. There is no correlation between profit and tighter CZs. In fact, the most stable and profitable combinations yield 5 to 8 Czs per week with lots of overlapping boxes involved.
As a side remark, many times I see people in here "back testing" over 30 trades or so, then start forward testing. This is as statistically hopeless as it is dangerous trading-wise. You will never ever get any sense of strategy stability by investigating locally.
PROPOSITION #6 - Martingale should be used only as a Tool
Inaccurate. What the assertion does not state though, is to what extent Martingale should be used.
This crucial aspect was never addressed as to question the viability of even a successful implementation of martingaling: that is the adverse effect of Spread. Below is a table assuming a 30 pip tall box, equal TP, No AutoReduce, a 2.0 pip spread, and martingale = x2 at each increment.
Of course, the relative proportion of CZ height versus TP is of some importance to the above chart. However, trade frequency is not. Everyone will get the point. The same logic remains for every trade. The main takeaway from the above table is that the spread sterilizes the use of martingale past the 3rd trade per box. Beyond that, martingale is merely a trick used to mitigate losses, but in no way is it a tool to maintain the profit magnitude of the initial breakout trade. The broad logic behind it is that martingaling does not only increase risk exponentially, but also trading costs. For once, choosing how many steps to commit to is this time a real choice in "what suits you best".
PROPOSITION #7 - The Newer Indicator's Issues are MT4-related
Inaccurate. As shown by the statistical properties above, a thorough back test is essential to identify what parameter combinations to choose from to start live trading on stable grounds. This VES cannot be traded seriously without it. To get there, reliability in trade data collection is of prime importance. The new indicator has indeed brought new bells and whistles, with the main improvement being that it now manages gapping signals correctly. Some MT4 issues do affect it alright, which anyone can live with. However, unlike the initial indicator, according to many remarks from the participants, and contrary to what has been systematically dismissed so far, it is seriously flawed with major repainting issues affecting CZ boxes and buffer lines. Below is a straightforward example. In the first picture, the early box drawing shows that a long breakout happened. We go long. A little later, picture two shows that this long trade has "vanished", not because the arrow has disappeared, but because the box shifted upwards, lifting the upper bound from 1.58085 to 1.58127.
Again, the old indicator did NOT do that. As a result, this instability issue with the new indicator renders it totally useless for any serious back testing. I am using my own algorithm to test this VES, and found that the first indicator was close to the truth, but it is very important that everyone willing to back test and investigate parameters accurately know what they can rely on as their trading results and consistency will essentially depend on it.
Thank you for your attention.
Princeton
I have been a regular reader of this thread for a while now. I do not trade optimized systems as a rule. However, I have been interested in the fact that this modified volatility expansion strategy (which I can testify is a sound trading concept) may also reach some kind of long-term profitability using optimization.
This method started off on promising grounds but as it unfolds, I find that posts after posts, many traps are still left open and, what is worse, are being held as core principles which are trusted by every one and used every day for trading. My contribution here is not meant to criticize the absence of mathematics in the system design. To each his own as long as it works. I rather feel compelled to highlight some crucial aspects so that people here may avoid this method's weaknesses - and ultimately lose money trading what may otherwise be a profitable one if soundly researched first.
What everybody needs to understand is that this volatility expansion strategy (VES) is first and foremost an exercise in optimization. Anyone may argue that this is a "naked system", that drawing boxes around CZs is what teaches you the edge etc...But doing so will only familiarize the user with the spirit of the method. The true system's mechanics rely on inducing profit stability through the identification of recurring and stable series of similar volatility patterns (not price patterns) born from the CZs. Given that this method essentially relies on 3 core parameters (CZ height, CZ length and TP), optimization becomes the one and only focus in the process of acquiring a stable, repeatable and trustworthy strategy. Claims may be made that one should not follow every single signal of the indicator, that the eye should decide what the best CZs are, that judgment calls are needed. But such claim inherently contradicts the initial premise of this strategy which tells us that the trader cannot and should not try to predict how the market will behave next - which is by the way the right bias to have.
Here are a few key remarks. I will try to be concise and specific.
PROPOSITION #1 - This Strategy has an Edge
Wrong. This strategy could have an edge indeed if it was not heavily parametric and did not require such serious optimization to approximate stability. Each new parameter set will invariably end up in another degenerating process to finish its course into a situation where the trader will not know how to make the difference between a mere drawdown and a broken trading logic. An edge, on the contrary, captures the dynamics of a given market class without any parameters or recurring back & forward tests. There is a way to decipher a drawdown from a broken trading logic when it comes to optimized system. I may address it in a later post though as there are more urgent aspects to focus on.
PROPOSITION #2 - A Forward Test is Key to Trade this Strategy Successfully
Inaccurate. The only upside of a forward test is to check whether the strategy can actually be executed with a decent level of replication versus theoretical results. It should never be used as a way to validate a given strategy or parameter set. Validation should have already been covered through the back test by the time a forward test is started. Otherwise, a forward test will bring nothing more than what a slow out-of-sample back test will, because this is exactly what this forward test will look like upon completion. A more efficient way to do it is to spare a large sample of data and test your optimized strategy over it. This will be immediate and will add the same statistical value to assessing your system's stability.
PROPOSITION #3 - One Should Pick the Settings that Suit Him
Wrong. EVERYBODY, regardless of its risk-aversion, should necessarily pick THE best combination of return versus risk. The rest is merely a question of money management. For every given instrument, time frame and data sample, there is one and only one set of parameters that is uniformly superior to all others. Not choosing it over the others has nothing to see with risk aversion or personal liking, but rather with flawed logic.
PROPOSITION #4 - The Tighter and the Longer the CZ, the Better
Wrong. Claims have been made of the above and also that a "sweetspot" should be found. These assertions are mutually exclusive. Finding the "sweet spot", without any "tighter CZ" bias, is the one way to go. The only filter that should be applied is that for the same risk/return combination and a similar pip profit, we should prefer the lesser number of trades so as to minimize the impact of spread cost. Apart from this notion, overlapping CZs are of no value as to tell whether a parameter combination is better or worse. The proof is below.
PROPOSITION #5 - SSBO is like Most Optimized Systems
Wrong. As I said, I do not trade optimized systems. But this is an answer to other threads making it a daily business in criticizing this one. The foundations of this VES are extremely sound. For this reason, and unlike 95% of the systems out there, positive results are not just available locally. In the case of SSBO, they are found across a wide array of parameter combinations and over solid historical data samples. This is where spectrum optimization comes into play. Because the profitability of optimized methods cannot be proven mathematically, the one and only way to show some evidence of consistency is to implement a long-term, comprehensive back test. Long-term means collecting at least 250 trades. Comprehensive means optimizing only the key parameters of the strategy, that is Height, Width and TP around the core trade of the strategy, that is the initial breakout trade, and using a neutral 1:1 risk-to-reward ratio to normalize outcomes. All other parameters are left fixed. Everything else linked to money-management such as martingale or other deceptive artifacts are foreign to a strategy's profitability and so they should be excluded from the analysis.
Below I used Bloomberg to quickly perform spectrum optimization over 400 days for a given time frame and a given pair, collecting at least 250 trades for each one of the 750 iterations that I investigated. The parameter combinations explored fit the average range over which contributors to this thread have focused on over time.The table shows each Gross Pip count.
Two conclusions jump out of this table.
(1) The BLUE CLOUD shows that there is clear stability of positive results across a wide but consistent neighborhood of parameter combinations (the "sweet spot") which allows for volatility conditions to vary greatly before a winning parameter set degenerates into a losing one. This is reassuring. The overall sanity check of the strategy is even more self-evident by looking at this 3D density display.
(2) The GOLD AREA in the table (bottom corner on the 3Ddensity) shows that in contrast with the sweet spot cloud, the assertion saying that "the Tighter, The Longer, The Better the CZ" is clearly pointing to a losing or sub-efficient set of parameter combinations. There is no correlation between profit and tighter CZs. In fact, the most stable and profitable combinations yield 5 to 8 Czs per week with lots of overlapping boxes involved.
As a side remark, many times I see people in here "back testing" over 30 trades or so, then start forward testing. This is as statistically hopeless as it is dangerous trading-wise. You will never ever get any sense of strategy stability by investigating locally.
PROPOSITION #6 - Martingale should be used only as a Tool
Inaccurate. What the assertion does not state though, is to what extent Martingale should be used.
This crucial aspect was never addressed as to question the viability of even a successful implementation of martingaling: that is the adverse effect of Spread. Below is a table assuming a 30 pip tall box, equal TP, No AutoReduce, a 2.0 pip spread, and martingale = x2 at each increment.
Of course, the relative proportion of CZ height versus TP is of some importance to the above chart. However, trade frequency is not. Everyone will get the point. The same logic remains for every trade. The main takeaway from the above table is that the spread sterilizes the use of martingale past the 3rd trade per box. Beyond that, martingale is merely a trick used to mitigate losses, but in no way is it a tool to maintain the profit magnitude of the initial breakout trade. The broad logic behind it is that martingaling does not only increase risk exponentially, but also trading costs. For once, choosing how many steps to commit to is this time a real choice in "what suits you best".
PROPOSITION #7 - The Newer Indicator's Issues are MT4-related
Inaccurate. As shown by the statistical properties above, a thorough back test is essential to identify what parameter combinations to choose from to start live trading on stable grounds. This VES cannot be traded seriously without it. To get there, reliability in trade data collection is of prime importance. The new indicator has indeed brought new bells and whistles, with the main improvement being that it now manages gapping signals correctly. Some MT4 issues do affect it alright, which anyone can live with. However, unlike the initial indicator, according to many remarks from the participants, and contrary to what has been systematically dismissed so far, it is seriously flawed with major repainting issues affecting CZ boxes and buffer lines. Below is a straightforward example. In the first picture, the early box drawing shows that a long breakout happened. We go long. A little later, picture two shows that this long trade has "vanished", not because the arrow has disappeared, but because the box shifted upwards, lifting the upper bound from 1.58085 to 1.58127.
Again, the old indicator did NOT do that. As a result, this instability issue with the new indicator renders it totally useless for any serious back testing. I am using my own algorithm to test this VES, and found that the first indicator was close to the truth, but it is very important that everyone willing to back test and investigate parameters accurately know what they can rely on as their trading results and consistency will essentially depend on it.
Thank you for your attention.
Princeton
- #5,470
- Edited Feb 19, 2013 8:11am Feb 18, 2013 9:31pm | Edited Feb 19, 2013 8:11am
- | Commercial User | Joined May 2011 | 1,754 Posts
To improve is to change. To perfect, is to change often.
- #5,471
- Feb 18, 2013 9:39pm Feb 18, 2013 9:39pm
- | Commercial User | Joined May 2011 | 1,754 Posts
To improve is to change. To perfect, is to change often.
- #5,472
- Feb 18, 2013 10:21pm Feb 18, 2013 10:21pm
- | Commercial User | Joined May 2011 | 1,754 Posts
To improve is to change. To perfect, is to change often.
Another 35 pips.
