stevehopwoodforex.com
https://www.stevehopwoodforex.com/phpBB3/
Print view

Macd Stochastic Fractal MA5/25 Daily Marylin
https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=1448		 Page 106 of 141
Author:  milanese [ Wed Oct 16, 2013 9:18 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
Hi all,

I opened now an extra account for Marilyn, here I test a new Empty4 version based at 0.40a but with TP calculation if SR zones are neares as 300 pips and with some different CSS test modules.

Cheers :)
Tommaso
here you can see how she looks :)
audjpydaily.png
Author:  phil_trade [ Wed Oct 16, 2013 9:26 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
How to find something usefull with CSS to detect Long Trend and Range period

EURUSD W1 example from 2009 to 2013
EURUSD W1 CSS compare to TREND 2009 2012 and RANGE 2013.PNG
to compare with FT EURUSD backtest well performing from 2009-2012 and flat since 2013
EURUSD 2009 2013 Graph result sans fractal brek.PNG
Author:  befreemow [ Wed Oct 16, 2013 9:26 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
Hi All,

Tommaso, checking the CSS on both currencies within a pair may be a good option, also given that a lot of work has been invested in tools and indicators related to CSS.
Here is another ... out of the box thought:

Please excuse these ramblings ...... Yes, the search to know what the trend is ..... flat, up, down, up over time or just now in the short-term.

For a moment let's forget about Time Frames (TF's) because this is just a way to cut up price in chunks for certain purposes, e.g. to read price behaviour or create indicators, etc. Let's just look at price over time and try to measure price using a horizontal straight line that changes its angle like a scale, if you take todays price of a currency and compare it to the currency price 3 days ago .... does the scale tip up or down and at what angle? Compare it to 3 weeks ago, does the scale tip up or down and at what angle. In between those 2 points of measurement it may have changed drastically up and down. Maybe an intelligent scale is needed that has multi-dimensional measurements and time sequences purely looking at price and the angles of movement is needed to truly know where the price has been and where it may continue in time?

I don't have the answers and I hope that posing questions may encourage further thought to help this project.

Tim
Author:  milanese [ Wed Oct 16, 2013 9:30 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
Hi Tim,

yes your input is great and very welcome :) Looking only at the price like you say is very interesting and promessing, but not easy to code ..
But what is easy :mrgreen: I will look to give a try :)

Cheers :)

Tommaso
Author:  phil_trade [ Wed Oct 16, 2013 9:38 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
befreemow wrote:Hi All,

Tommaso, checking the CSS on both currencies within a pair may be a good option, also given that a lot of work has been invested in tools and indicators related to CSS.
Here is another ... out of the box thought:

Please excuse these ramblings ...... Yes, the search to know what the trend is ..... flat, up, down, up over time or just now in the short-term.

For a moment let's forget about Time Frames (TF's) because this is just a way to cut up price in chunks for certain purposes, e.g. to read price behaviour or create indicators, etc. Let's just look at price over time and try to measure price using a horizontal straight line that changes its angle like a scale, if you take todays price of a currency and compare it to the currency price 3 days ago .... does the scale tip up or down and at what angle? Compare it to 3 weeks ago, does the scale tip up or down and at what angle. In between those 2 points of measurement it may have changed drastically up and down. Maybe an intelligent scale is needed that has multi-dimensional measurements and time sequences purely looking at price and the angles of movement is needed to truly know where the price has been and where it may continue in time?

I don't have the answers and I hope that posing questions may encourage further thought to help this project.

Tim

Hmmmm... need some cool off after reading that ;)
relacher la pression.jpg
Author:  zTB [ Wed Oct 16, 2013 10:20 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
Look at the chart below. There are quite a few market types – even more than those noted on the chart.

What is one to do?

Design a ‘super smart’ trading system good for multiple market types?

A ‘super smart’ trading system – if there is, or can be, such a thing – would require lots of inputs. The problem is the more inputs a trading system has, the less robust it is likely to be. KISS is best. Remember Long Term Capital Management had two Nobel laureates, and the joke has it:
It seems LTCM could have survived one Nobel prize-winner, but with two, they were doomed.
To make steady money one needs a good and robust trading system which allows for rigorous risk control and judicious position sizing. A super system is not required.

All evidence suggests Marylin is a fairly good and robust trading system. Based on evidence I could find, it is the best the Forum has. Adding more and more inputs, filters, etc. may be good for the ego, but it may not be good for Marylin.

Phil is to be commended for demonstrating Marylin real money live. I do not think there is another live real money test on the Forum. If for no other reason, for this reason alone, one hesitates to be critical, and feels bad about it.

Nevertheless, in my opinion, Marylin would benefit from (i) control of the really big losses (ii) better position sizing (iii) a less correlated, and more diversified, portfolio (iv) and a longer back-test period including 2008.

Here is an analysis of what I call “really big losses”. Empty4-907238-JPG has had 33 of its trades each individually lose more than the average profit of € 50.18. These 33 trades all together lost more money than the rest of the 139 losing trades. 15 of these 33 are XAUUSD trades, and 2 of those (the second entered within 63 minutes of the first which was already more than € 90 down) 15 collectively lost more than € 500, or equivalent to 10% of total money lost.
The elements of good trading are: (1) cutting losses, (2) cutting losses, and (3) cutting losses. If you can follow these three rules, you may have a chance.
Pyramiding instructions appear on dollar bills. Add smaller and smaller amounts on the way up. Keep your eyes open at the top.
Apologies for the missing chart. :oops: :oops: :oops:

Z
Author:  phil_trade [ Wed Oct 16, 2013 11:33 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
Z

Good post
Nevertheless, in my opinion, Marylin would benefit from (i) control of the really big losses (ii) better position sizing (iii) a less correlated, and more diversified, portfolio (iv) and a longer back-test period including 2008.
"in my opinion" is one thing, but could you prove this assertion with real fact, result or deep manual analyse ?
Here is an analysis of what I call “really big losses”. Empty4-907238-JPG has had 33 of its trades each individually lose more than the average profit of € 50.18. These 33 trades all together lost more money than the rest of the 139 losing trades. 15 of these 33 are XAUUSD trades, and 2 of those (the second entered within 63 minutes of the first which was already more than € 90 down) 15 collectively lost more than € 500, or equivalent to 10% of total money lost.
There was some bug with XAUUSD at the beginning of Marylin. So to analyse trades, you should only look last 6 months or avoid 3 first month only for Gold.

Philippe
Author:  zTB [ Wed Oct 16, 2013 1:54 pm ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
phil_trade wrote:Z

Good post
Nevertheless, in my opinion, Marylin would benefit from (i) control of the really big losses (ii) better position sizing (iii) a less correlated, and more diversified, portfolio (iv) and a longer back-test period including 2008.
"in my opinion" is one thing, but could you prove this assertion with real fact, result or deep manual analyse ?

Philippe
Sure, Phil.

I have already referenced certain points from your myfxbook performance data, but shall go into greater detail tomorrow.

Z
Author:  aSushiRoll [ Wed Oct 16, 2013 11:47 pm ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
yes and let me please share my link again :) 116% in 4 months :D version 0.36 I think..
https://www.myfxbook.com/portfolio/1-mt ... mon/606234

P.S. I took all the profits 2 days ago, it looked tot good :)
Author:  zTB [ Thu Oct 17, 2013 6:04 am ]
Post subject:  Re: Macd Stochastic Fractal MA5/25 Daily Marylin
zTB wrote:
phil_trade wrote:Z

Good post
Nevertheless, in my opinion, Marylin would benefit from (i) control of the really big losses (ii) better position sizing (iii) a less correlated, and more diversified, portfolio (iv) and a longer back-test period including 2008.
"in my opinion" is one thing, but could you prove this assertion with real fact, result or deep manual analyse ?

Philippe
Sure, Phil.

I have already referenced certain points from your myfxbook performance data, but shall go into greater detail tomorrow.

Z
Herewith a bit more detailed analysis of Marylin’s real money performance.

In the attached Excel file Marylin are a number of sheets. The Overview is simply a reproduction of information from myfxbook and need not be commented upon.

The next sheet Expectancy is interesting. The expectancy per trade is 0.33, which means that every time Marylin trades, the expected return per € 1 risked is € 0.33; but the main area of concern is the standard deviation of expectancy – too high at 2.49. This high standard deviation is adversely impacted by the large losses. One may do a what-if analysis in Excel. If you lower the numbers in the red cells to about € 50, you will see significant improvement. The following sheet, Loss Trades, breaks down the losses. 19% of (big) losses account for more than 50% of the money lost. Equally significantly, the 5 biggest of those 33 losses account for more than 33% of the money lost by the 33 trades. If you pay attention to the control of the big losses, Marylin will see dramatically better returns.

The sheet Exit Efficiency Ratio reinforces the need to control the losses incurred in the early stages of a trade. The Exit Efficiency Ration suggests, Marylin quite good at retaining the profits once it is in the profit territory.

The sheet Pairs Performance breaks down returns on the basis of the pairs in the Marylin portfolio. The returns range from € 23 per trade profit for GBPUSD to € 20 loss per trade for EURJPY. Given such a variation in the performance of the portfolio constituents, I find it difficult to understand why, with the exception of XAUUSD, all pairs are being allocated the same lot size. I would allocate more money to the winners, and reduce the lot size of the losers. XAUUSD get the highest lot size allocation even though is only a middling performer. Higher allocation of investment money to the winners, and lesser to the losers can only increase the overall profitability of the system. Hence, I said better position sizing is needed.

The Correlation sheet shows how closely are the portfolio constituents correlated. The way I see it Marylin is practically trading only 4 instruments. The average correlation of all 8 instruments is 0.45. This is not enough of a diversification to help in getting better returns against the risks taken. A number closer to, or lower than, 0.25 would be a big step forward and you will see improvement in the return/risk ratio by at least a factor of 2.

Marylin trades the daily chart. For this granularity the back test period should be at least 5 years, never mind the number of in-sample trades. Moreover, the present test period does not have enough of the ‘trending and volatile’ market type. Inclusion of the year 2008 will adequately redress the balance.

Finally, please see an updated 10,000 iteration Marylin-Random Equity. The orange line represents the actual performance. Indeed, there is room above up to the 5% probability green line. But there is also the danger of the 95% probability red line.

It always helps to keep the risk measured.

Z
All times are UTC Page 106 of 141

Powered by phpBB® Forum Software © phpBB Limited