Phoenix optimization - page 8

 
m6m6:
I am happy to do so. Just call FXDD customer support at 1-212-791-3933 and ask them to transfer your account for it to have Phoenix Fund as an IB (Introducing Broker). They will need you to resend by fax a few of the document you initially signed and then they will assign you a new account number. Then make sure you ask them to close the other account. Then follow up closely to accelerate the procedures. It took me 4 open days to transfer mine.

For those who do not have an existing account with FXDD (www.fxdd.com), just make sure, when registering, to fill in the "IB" section on the top of one of the first registering document.

All users should also read the "Phoenix Fund" thread, it will explain that Daraknor negotiated a commission for the Fund so out of every trade we take (or Phoenix !), a part of the spread fees goes to the Fund. Phoenix Fund Website : http://pf.bestforextools.com/,

Thanks you all for doing the extra effort to help the cause !

M6M6

Peux tu m'expliquer en bon francais car mon anglais est parfois un peu... court ? Merci

(can you explain me that in French coz my english is sometimes a bit...short! Thks)

 
bertbin:
M6M6

Peux tu m'expliquer en bon francais car mon anglais est parfois un peu... court ? Merci

(can you explain me that in French coz my english is sometimes a bit...short! Thks)

I will reply in your Private Messages...

 
m6m6:
I will reply in your Private Messages...

Got it... Thks... And answered you ! BTW I have to donate to Phoenix Fund : I totally forgot about that !!!!

 

A few Questions.............

How long back should my history data be when back testing? It seems that if I use just 2006 I get good results, if I use 2005-2007 I get terrible results for 2005 and then the good results for 2006. I guess it would be good to how a few years data?

2006 also keeps coming up as a good year for Phoenix, Was it predictable?

When pref setting is set to true, I don't have to manually put in the signal settings do I?

Also, Any Idea how v6.0 is coming along?

 

The official answer for Phoenix 5 is: nobody knows but a few months are better than a few weeks. In a guide I wrote, I talked about the different optimization difficulties and they were more of an issue with 4 weeks of data than 4 months of data. Hendrick recommended 6 months.

The intended answer for Phoenix 6 is: as much as you want. Phoenix 6 will have trend avoidance (don't ice skate uphill) if we can find a good solid "Trend is underway" indicator.

 

Rather than simply increase maxrisk to increase profitability, I recommend increasing maxtrades and signal_count. These two parameters combined will allow you to have multiple trades (I tested up to 20 simultaneous trades on one currency) and rather than put a lot of money on a single trade you have a little bit of hedging and a little bit of risk reduction. Using less money each subsequent trade also ensures the system keeps going. Maxrisk is very dependent on the SL value. I recommend lower Maxrisk with high SL values to limit overall risk. Also, some brokers don't like lotsizes over a certain amount. Very few will accept large lotsizes. (FXDD accepts over 100 lots without complaint from what I heard by a trading friend.)

That said, congratulations on the performance!

 

Max lot size

daraknor:
(FXDD accepts over 100 lots without complaint from what I heard by a trading friend.)

Good to read that you feel better. Welcome back !

FXDD accept up to 50 lots per trade on regular account. As I previously wrote, it is better to go in a live FXDD mini account because it will process the fractioned lots. For example, 3.2 lot signal from Phoenix will end up at 3 lots in a regular account and 3.2 in a mini. Phoenix settings, in a live mini account, need to be exactly the same as in regular Demo.

 

Automated optimization

Last weekend I was optimizing one currency pair over many variables (6 or 7), Metatrader suggested that the process would take about 10 hours. So I left the computer on and went to sleep. When I checked the next day, the process had completed in under 3 hours.

I don't know much about the actual optimization process in Metatrader but I read somewhere that it uses genetic algorithm to improve the performance of optimization - maybe this helped Metatrader to run the optimization faster.

If I had known that the optimization will take around around 3 hours, I would have scheduled other tasks at the same time.

Anyway, I started researching into windows-macro tools (eg AutoIt) to automate the optimization partially or completely and have been surprised with the capability of such tools.

Before I spend a lot of time researching into these tools, I thought I'd better ask if anyone has tried automating (partially or completely) the optimization process in Metatrader?

 

I made a template settings file you can load to do testing on. If you load it, many many settings will be uncovered. I recommend optimizing signal and trade settings separately.

Metatrader4 uses Genetic Algorithms to focus on profitable settings to find which settings generate the most profit. I think the settings file I generated has something like 17,000,000,000 possibilities. the GA system runs about 3000 of those, spread around quite a bit. Next, it tests about 7000 settings "near" the existing settings by adjusting one parameter up or down based on the "step" value.

I think the reason it said 10 hours and then cut down to 3 hours completion time was because either A) an exhaustive search was done using settings I filter out (5.7.2 feature) or B) some of the test results were cached from previous testing and were simply reused instead of retested.

One thing I do is stop tests once they hit a certain low balance or low result. If you test with 10k and the test hits 8k in the first month of data, no point in waiting on 5 more months of data to tell you exactly how bad the results are. A savings of 5/6 time for that test.

Remember to keep at least the 2 weeks of most recent data out of the optimized test, and then try the settings on those tests. If the settings are positive on those tests, you more likely have a winner. GA system will make clumped results.

I prefer the macro program AutoHotKey, I have scripted games and repetitive tasks with it. We can likely program AutoHotKey to run the 2week filter, copy and paste the results.

 

Thanks for pointing out the template settings file and AutoHotKey.

I was thinking of completely automating the process by using Excel and MetaTrader together. Excel for keeping track of (and computing) variable start and end values and then copying them into MetaTrader and running the optimization.

I will give it a go over next few days and tell you guys how it goes.

daraknor:
I made a template settings file you can load to do testing on. If you load it, many many settings will be uncovered. I recommend optimizing signal and trade settings separately.

Metatrader4 uses Genetic Algorithms to focus on profitable settings to find which settings generate the most profit. I think the settings file I generated has something like 17,000,000,000 possibilities. the GA system runs about 3000 of those, spread around quite a bit. Next, it tests about 7000 settings "near" the existing settings by adjusting one parameter up or down based on the "step" value.

I think the reason it said 10 hours and then cut down to 3 hours completion time was because either A) an exhaustive search was done using settings I filter out (5.7.2 feature) or B) some of the test results were cached from previous testing and were simply reused instead of retested.

One thing I do is stop tests once they hit a certain low balance or low result. If you test with 10k and the test hits 8k in the first month of data, no point in waiting on 5 more months of data to tell you exactly how bad the results are. A savings of 5/6 time for that test.

Remember to keep at least the 2 weeks of most recent data out of the optimized test, and then try the settings on those tests. If the settings are positive on those tests, you more likely have a winner. GA system will make clumped results.

I prefer the macro program AutoHotKey, I have scripted games and repetitive tasks with it. We can likely program AutoHotKey to run the 2week filter, copy and paste the results.