DislikedGlenhaven Capital....ouch {image} Probably no need to go on...but it at least demonstrates how the long term fortunes are dictated by Outliers. The rest is just noise...in the scheme of things :-)Ignored
Diversified Trend Trading Approach
- #4,503
- Jun 23, 2021 8:27pm Jun 23, 2021 8:27pm
- | Commercial User | Joined Apr 2013 | 4,366 Posts
Here is how we can definitively test for an edge.
Take a very long term 'real' market data sample. Randomise it with an array to eliminate auto-correlation in the series. Then assess the performance of your system against that array of random result. A system with an edge applied to real market data will clearly outperform an array produced by applying that system to random market data.
Here is a Breakout Trend Following Model (uncompounded) applied to real market data (top performer) against that same model applied to that same market data which has been randomly shuffled to eliminate auto-correlation in the series.
Now that we know there is an edge to that system, we can apply compounding to the series to lift it to the heavens.
You only need a weak edge under compounding to generate significant wealth building returns. Only 5.3% of these trades from outlier events produced this result. The rest was inconsequential to total performance.
This is the result of a system with a weak edge under a very large trade sample size which through diversification captures outliers in market data. It is only about 5% of total trades that turn the following profitable performance to a random performance result which is displayed by the lower group of curves which are hard to visually see over the large sample size.
Take a very long term 'real' market data sample. Randomise it with an array to eliminate auto-correlation in the series. Then assess the performance of your system against that array of random result. A system with an edge applied to real market data will clearly outperform an array produced by applying that system to random market data.
Here is a Breakout Trend Following Model (uncompounded) applied to real market data (top performer) against that same model applied to that same market data which has been randomly shuffled to eliminate auto-correlation in the series.
Now that we know there is an edge to that system, we can apply compounding to the series to lift it to the heavens.
You only need a weak edge under compounding to generate significant wealth building returns. Only 5.3% of these trades from outlier events produced this result. The rest was inconsequential to total performance.
This is the result of a system with a weak edge under a very large trade sample size which through diversification captures outliers in market data. It is only about 5% of total trades that turn the following profitable performance to a random performance result which is displayed by the lower group of curves which are hard to visually see over the large sample size.
- #4,507
- Jun 23, 2021 8:59pm Jun 23, 2021 8:59pm
Your arguments appear convincing and may very well be correct. But I remain uncertain.
I require a large amount of data from at least a similar population before drawing conclusions and comparisons.
Consequently, I wont be convinced until the long-term data from forex prop traders is available. Given that’s unlikely I will have to rely on my own personal research. Will reply to this post in 5 years time.
I require a large amount of data from at least a similar population before drawing conclusions and comparisons.
Consequently, I wont be convinced until the long-term data from forex prop traders is available. Given that’s unlikely I will have to rely on my own personal research. Will reply to this post in 5 years time.
- #4,508
- Jun 23, 2021 9:09pm Jun 23, 2021 9:09pm
- | Commercial User | Joined Apr 2013 | 4,366 Posts
Here is the market data on which this simple breakout system was tested. The red market data is the 'real market data' in the series. All the other curves are random market data sets where the auto-correlation in the market data series has been removed.
Now if you had a breakout system, you would assume that some of the more volatile curves would produce a better performance result for that system than the real market data.....but here is the rub....they don't.
The real market data (red) produced the best performing equity curve above in green.
It is not the form of the market data that is important to an edge. It is the 'auto-correlation' in the series that makes a profitable system. So technical traders looking for 'form or pattern' are not necessarily going to able to detect the 'actual bias' that exists in a market data series.
Here is a random trend......plotted against a non random trend (a price series with auto-correlation in it).
There is no edge to be extracted from the random trend in the future....but there is an edge to be extracted from the non random trend in the future. Visually you will not be able to decide between the two....but the auto-correlation in the non random series produces the drift in the series. It is the drift...or the bias in the series that leads to a system with an edge.
During outlier events....often called capitulation tails......there is an auto-correlated bias in the underlying data series. No such bias exists in random market data....hence your results are just that....random.
Now if you had a breakout system, you would assume that some of the more volatile curves would produce a better performance result for that system than the real market data.....but here is the rub....they don't.
The real market data (red) produced the best performing equity curve above in green.
It is not the form of the market data that is important to an edge. It is the 'auto-correlation' in the series that makes a profitable system. So technical traders looking for 'form or pattern' are not necessarily going to able to detect the 'actual bias' that exists in a market data series.
Here is a random trend......plotted against a non random trend (a price series with auto-correlation in it).
There is no edge to be extracted from the random trend in the future....but there is an edge to be extracted from the non random trend in the future. Visually you will not be able to decide between the two....but the auto-correlation in the non random series produces the drift in the series. It is the drift...or the bias in the series that leads to a system with an edge.
During outlier events....often called capitulation tails......there is an auto-correlated bias in the underlying data series. No such bias exists in random market data....hence your results are just that....random.
- #4,509
- Jun 23, 2021 9:11pm Jun 23, 2021 9:11pm
- | Commercial User | Joined Apr 2013 | 4,366 Posts
Your arguments appear convincing and may very well be correct. But I remain uncertain. I require a large amount of data from at least a similar population before drawing conclusions and comparisons. Consequently, I wont be convinced until the long-term data from forex prop traders is available. Given that’s unlikely I will have to rely on my own personal research. Will reply to this post in 5 years time.
- #4,515
- Jun 30, 2021 10:52pm Jun 30, 2021 10:52pm
- | Commercial User | Joined Apr 2013 | 4,366 Posts
Why a Trend Follower Uses Portfolio Realised Balance rather than Portfolio Equity for Position Sizing
https://atstradingsolutions.com/why-...sition-sizing/
https://atstradingsolutions.com/why-...sition-sizing/
- #4,520
- Jul 25, 2021 10:09pm Jul 25, 2021 10:09pm
- | Commercial User | Joined Apr 2013 | 4,366 Posts
It's great to be a part of the Top Traders Unplugged crew. Such a privilege.
Looking forward to more sessions and bringing different perspectives into the discussion on Trend Following.
https://www.toptradersunplugged.com/...uly-25th-2021/
Looking forward to more sessions and bringing different perspectives into the discussion on Trend Following.
https://www.toptradersunplugged.com/...uly-25th-2021/