Phoenix 2007 (new thread) | Page 44 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Phoenix Forex Manager trading journal 25 replies

Phoenix Foundation 1 reply

Can someone fix the Phoenix EA? 0 replies

Automated Trading Championship 2007: New Challenge! 33 replies

Phoenix is here! 848 replies

Attachments: Phoenix 2007 (new thread)
Exit Attachments

Phoenix 2007 (new thread)

  • #861
  • Nov 15, 2006 7:40am Nov 15, 2006 7:40am
Quoting Palmer
Four winners in a row last night, MODE=1...why change anything....

Good luck Harry.
I want to be wrong. And todays settings are the best. But...
  • #862
  • Nov 15, 2006 8:41am Nov 15, 2006 8:41am
14 November 2006

New MetaTrader 4 build 200 has been released.

What’s new:
1. Removed Multilanguage Pack, the distributive size decreased to 3 MByte;
2. Added feature to manage bookmarks of charts bar;
3. Tester: Reduced limitations on the amount of passes in genetic optimization (minimum 200 passes);
4. Tester: Imrpoved fractal modeling of a bar. Now more smoothed patterns are used to model price movements;
5. Tester: Fixed free margin calculation, which was incorrect in some cases;
6. Tester: Fixed stopuot status calculation, in case of the absolute level, not a percent one;
7. Tester: Added the margin_stopout_mode parameter in the FXT file heading structure, the heading size not being changed. The version number became 403;
8. Tester: Corrected drawing of the Ask line at visual testing in the pause status;
9. Tester: Fixed execution of deinit function of expert advisor at stop of testing;
10. MQL4: Improved working with temporary variables of the ’string’ type;
11. MQL4: Added functions: AccountFreeMarginMode, AccountStopoutMode, AccountStopoutLevel, AccountFreeMarginCheck;
12. MQL4: Added functions: IsVisualMode, TerminalPath, WindowExpertName, WindowPriceMin, WindowPriceMax, GlobalVariablesTotal, GlobalVariableName;
13. MQL4: Changed operation logic of the GlobalVariablesDeleteAll function — it is now possible to specify the prefix for mass deletion of global variables;
14. MQL4: In the MarketInfo function, added identifiers: MODE_SWAPTYPE, MODE_PROFITCALCMODE, MODE_MARGINCALCMODE, MODE_MARGININIT, MODE_MARGINMAINTENANCE, MODE_MARGINHEDGED, MODE_MARGINREQUIRED;
15. MQL4: Due to functions restructuring, renamed the following functions: PriceOnDropped, TimeOnDropped, FirstVisibleBar, BarsPerWindow, ScreenShot, HistoryTotal, CompanyName, ClientTerminalName, ServerAddress, Lowest, Highest, CurTime, LocalTime;
16. MetaEditor: Due to functions renaming, added the «Outdated Functions» section into the Dictionary;
17. Added the ray property to the object named Fibo Expansion;
18. Fixed and improved scanning of servers when connection is broken;
19. Added downloading and import of data in the History Center;
20. Fixed saving of the objects’ default flag named background;
21. Fixed saving of charts multiline comments in configuration files;
22. Added a sound file notifying about that testing is over;
23. Improved displaying of empty databases in History Center — after the first call, data are pumped and displayed;
24. Strongly accelerated working with databases on huge arrays of historical data;
25. Strongly accelerated operation of deleting large amounts of historical data from History Center, fixed error of re-offsetting the history just deleted;
26. Fixed small errors in trade logs output.
<!-- / message --><!-- controls -->
  • #863
  • Nov 15, 2006 9:45am Nov 15, 2006 9:45am
Quoting Hendrick
Here you have the results of 4 backward test for the USDJPY Mode 1.
Period 2006.5.1 2006.11.10
MM=false (see reports for the rest of the settings)

Test A Mode 1 Without other settings
Test B Mode 1 BreakEvenAfterPips=32
Test C Mode 1 CloseAfterHours=120
Test D Mode 1 Combination of test B and test C

So it makes sense to use the setting BreakEvenAfterPips to improve your results. Net Profit will increase but what's more important MD will decrease.
Just one more addition to this,.....
BreakEven AfterPips=0,and Trailing stop=32 is even better but i do not have data till Nov.11 to compare exact numbers.
  • #864
  • Nov 15, 2006 9:51am Nov 15, 2006 9:51am
Quoting corner_h
but i do not have data till Nov.11 to compare exact numbers.
http://www.alpari-idc.ru/en/dc/databank.php
  • #865
  • Nov 15, 2006 10:01am Nov 15, 2006 10:01am
I've asked these a couple of times before, but, does anyone know why the defaults for USDJPY in the contest version are different than the released versions (both for 4.x and 5.x)?

For example, here's a partial list of differences (there are additional differences):

defaults for contest version

SMAPeriod = 7;
SMA2Bars = 14;

defaults for released version

SMAPeriod = 2;
SMA2Bars = 18;
  • #866
  • Nov 15, 2006 10:06am Nov 15, 2006 10:06am
I've asked this before, but, just wanted to ensure the question was clear.

Has anyone been able to get backtests of this EA to confirm forward tests? Since extensive backtesting is used to derive optimal settings, it would sure be nice to see clear correlations between backtests and actual forward tests.

For example, it would be nice to see backtests of the period from 10/1/09 to present using the contest version of the EA which match the results posted on the championship site. I've tried, but, haven't been able to get them to match up. I know that the precise starting point can have a lot to do with it, but, if you set Strategy Tester to begin the same day as the contest began, presumbably even if the actual start time is off by some hours, one would think that after a certain period the backtests would sync up with the live results. That hasn't been the case with the basic tests I've done. Just wanted to see if anyone else here has done more in depth tests and has been successful with this, or at least has information as to why its not working.

Thanks in advance for any input on this.

Bill
  • #867
  • Nov 15, 2006 2:23pm Nov 15, 2006 2:23pm
Please find attached the first version of my attempt to explain how to optimize <?xml:namespace prefix = st1 ns = "urn:schemas-microsoft-com:office:smarttags" /><st1:City w:st="on"><st1lace w:st="on">Phoenix</st1lace></st1:City>’s settings. Please realize that, as always in Forex, back testing doesn’t guaranty anything for the future!

First of all thanks to Hendrick of course! Without him there wouldn’t have been anything to optimize at all…

Hopefully you will be able to run the tests yourself now and there won’t be any need anymore for the usual: “Can anybody post the most resent/ best settings for currency pair …”
My thought behind this is that I’ve always been a strong believer in trying to educate oneself. I’m not convinced that just spoon feeding everyone by posting 100 and one found “best settings” will help anyone. Why would you want to sift through an x amount of settings produced by others, when you can find them relatively easy yourself?
I can imagine it can be daunting at first when you are new to it all. But I’m still relatively new to EA’s too. I have learned so much in the past few months just by reading, experimenting, playing with settings and asking questions when I couldn’t find my question already being answered somewhere.

Another reason for wanting to do optimizing yourself is that one person prefers to optimize for maximum profit while someone else might prefers lowest % of drawdown.

Finally, I found that different MT4’s downloaded from different brokers gave different results… (even when using the exact same data…). So what works for someone else might not work for you..

You have all the available tools to make it work best for YOUR preference. All you need to do is put some effort in!

But hey, it’s just my humble opinion. Everyone is different so it is up to you!

Good luck testing and please let me know what you think of the manual. Your feedback is much appreciated!

Vince
Attached File(s)
File Type: pdf How to optimize Phoenix.pdf   364 KB | 949 downloads
  • #868
  • Nov 15, 2006 2:52pm Nov 15, 2006 2:52pm
Quoting billworld
I've asked this before, but, just wanted to ensure the question was clear.

Has anyone been able to get backtests of this EA to confirm forward tests? Since extensive backtesting is used to derive optimal settings, it would sure be nice to see clear correlations between backtests and actual forward tests.

For example, it would be nice to see backtests of the period from 10/1/09 to present using the contest version of the EA which match the results posted on the championship site. I've tried, but, haven't been able to get them to match up. I know that the precise starting point can have a lot to do with it, but, if you set Strategy Tester to begin the same day as the contest began, presumbably even if the actual start time is off by some hours, one would think that after a certain period the backtests would sync up with the live results. That hasn't been the case with the basic tests I've done. Just wanted to see if anyone else here has done more in depth tests and has been successful with this, or at least has information as to why its not working.

Thanks in advance for any input on this.

Bill
Bill,
I just did a Backtest of usdjpy with Phoenix-v4.2.03 and compared to same period of demo (forward test) of same EA. Attached are results in an Excel file. One sheet is Demo and next sheet is Backtest. Big difference. Lots sizes are slightly different, but not enough to invalidate results. Demo produce much better. Also, Model Quality on Backtest was 90%.

Wackena
Attached File(s)
File Type: zip Demo-Backtest-usdjpy..zip   6 KB | 462 downloads
  • #869
  • Nov 15, 2006 3:15pm Nov 15, 2006 3:15pm
I had an telephone conversation with IBFX to investigate why Phoenix Contest results in the Championship were not the same as my IBFX account over the same time period. They were not sure why, but thought server response time or data feed difference might be factors.

They did advise that they plan to have a live tick data feed for Demo testing availabe in a couple of months.

Unless your Broker is using Live tick data for Demos, your results will be somewhat different than your Live trading account.

Wackena
  • #870
  • Nov 15, 2006 3:18pm Nov 15, 2006 3:18pm
Quoting billworld
I've asked these a couple of times before, but, does anyone know why the defaults for USDJPY in the contest version are different than the released versions (both for 4.x and 5.x)?

For example, here's a partial list of differences (there are additional differences):

defaults for contest version

SMAPeriod = 7;
SMA2Bars = 14;

defaults for released version

SMAPeriod = 2;
SMA2Bars = 18;
Each update had further optimization of the preferred settings. Hendrick explained this in the last thread as well as the beginning of this thread. This Hendrick quote sums up the theory while v4.2 was in use.
Quote
And that’s the idea behind Phoenix2007: keep on testing and changing until you have found this special combination of settings, that gives you a smooth up going balance-curve. There’s no doubt about it that this balance-curve will continue in the same direction for at least the coming four to eight weeks because the law that’s causing this won’t stop or change. Only thing to be done is backward test every 2 weeks to see if the settings are still up to date.
Because of the success of the contest version I think the optimization time frame has been open for re-evaluation.
  • #871
  • Nov 15, 2006 3:19pm Nov 15, 2006 3:19pm
Quoting wackena
I had an telephone conversation with IBFX to investigate why Phoenix Contest results in the Championship were not the same as my IBFX account over the same time period. They were not sure why, but thought server response time or data feed difference might be factors.

They did advise that they plan to have a live tick data feed for Demo testing availabe in a couple of months.

Unless your Broker is using Live tick data for Demos, your results will be somewhat different than your Live trading account.

Wackena
Good work, that explains a few things.

Thanks
  • #872
  • Nov 15, 2006 6:06pm Nov 15, 2006 6:06pm
Excellent. Thanks Wackena. This corroborates what I've been seeing. Not to cause undue concern here, but, if backtesting is used as the primary means to derive settings for a system which openly admits to being curve-fitted based on recent backtesting, and said backtesting does not corroborate with live or forward tested performance, then, settings derived from this method would not appear to accurately represent the reality of the situation. Mind you, there's nothing wrong with the premise here as long as forward/live tests match backtests over the same period. If one can't get to that point, it would appear one is shooting in the dark...

This is just an observation which in no way denigrates or in any way whatsoever shows lack of appreciation or respect for all of the wonderful contributions made toward this project.

Quoting wackena
Bill,
I just did a Backtest of usdjpy with Phoenix-v4.2.03 and compared to same period of demo (forward test) of same EA. Attached are results in an Excel file. One sheet is Demo and next sheet is Backtest. Big difference. Lots sizes are slightly different, but not enough to invalidate results. Demo produce much better. Also, Model Quality on Backtest was 90%.

Wackena
  • #873
  • Nov 15, 2006 6:19pm Nov 15, 2006 6:19pm
Thanks for doing this wackena (cool name btw!). I've been rather closely monitoring the championship servers' quotes vs what I have on my IBFX, FBFX and FXDD accounts (amount other brokers) and haven't noticed that big of a difference. Certainly not a big enough difference to through off the envelopes and SMAs used for triggers in Phoenix.

Also, what this doesn't explain is why some entries are near perfectly aligned and others are way off.

Since Phoenix is hard to visualize on charts (without any real indicator which matches the way it operates) and its "secret" is reliant upon settings derived from extensively optimized (e.g. curve-fitted) backtesting, if we can't correlate backtests to something which occurs in the real world like forward/live results, then, I say we're surely on shakey ground.

I venture Hendrick is concerned with this as well (although I haven't had the pleasure of any direct contact with him to date) and that's one of the reasons why he's decided to go back into research and development mode without distractions from the group here (nothing wrong with that mind you).

Quoting wackena
I had an telephone conversation with IBFX to investigate why Phoenix Contest results in the Championship were not the same as my IBFX account over the same time period. They were not sure why, but thought server response time or data feed difference might be factors.

They did advise that they plan to have a live tick data feed for Demo testing availabe in a couple of months.

Unless your Broker is using Live tick data for Demos, your results will be somewhat different than your Live trading account.

Wackena
  • #874
  • Nov 15, 2006 6:35pm Nov 15, 2006 6:35pm
Quoting billworld
Thanks for doing this wackena (cool name btw!). I've been rather closely monitoring the championship servers' quotes vs what I have on my IBFX, FBFX and FXDD accounts (amount other brokers) and haven't noticed that big of a difference. Certainly not a big enough difference to through off the envelopes and SMAs used for triggers in Phoenix.

Also, what this doesn't explain is why some entries are near perfectly aligned and others are way off.

Since Phoenix is hard to visualize on charts (without any real indicator which matches the way it operates) and its "secret" is reliant upon settings derived from extensively optimized (e.g. curve-fitted) backtesting, if we can't correlate backtests to something which occurs in the real world like forward/live results, then, I say we're surely on shakey ground.

I venture Hendrick is concerned with this as well (although I haven't had the pleasure of any direct contact with him to date) and that's one of the reasons why he's decided to go back into research and development mode without distractions from the group here (nothing wrong with that mind you).
Wackena is a North Carolina Iroquoian Indian word for pond or slow moving water. My given name is also Bill.

Wackena
  • #875
  • Nov 15, 2006 11:31pm Nov 15, 2006 11:31pm
Great job, cool EA, my respect, Hendrick.

Trying some backtests of different phoenix versions, results differs from doubling (and more) deposit to margin call. Looking at indicator for this EA (dot't remember by who, posted at championship forum), I've got an idea - open trade only by some consecutive signals.

added some code to EA( like this for sell signal) :

if (CurTime()-s_time>900) {ss=ss+1; bs=0;}
if (ss>=signal_count) {
res=OrderSend(...SELL....);
ss=0; s_time=CurTime();}

I think, it will help to filter out single signals at gaps, etc..

Made this for ph_4 (contest ver) and for ph_5 at mode1 (last ver) - the result is a new significant parameter for optimizing using pref_settings=true; For EURJPY, for example, I've got 50% increase in profit turning signal_count to 5 (not less than five consecutive signals opens trade). Surprisingly, this didn't reduces total amount of trades, only placing orders more precise.

Attaching both changed files, hope it's usefull contribution to phoenix development
Attached File(s)
File Type: mq4 Phoenix_4_CONTEST_changed.mq4   11 KB | 737 downloads
File Type: mq4 Phoenix_EA_v5_6_03_changed.mq4   34 KB | 760 downloads
  • #876
  • Nov 16, 2006 12:31am Nov 16, 2006 12:31am
In addition to my previous post :

with that signal counter, mean opening trade only by some consecutive phoenix signals - just for fun I've turned this counter to great number (50, 100, 200 etc.) - It was a stunning surprise - the Phoenix can catch the trend then !
For GBPUSD with open order after 200 cons.signals - only 28 trades from May.2006 till now - but only one loss, somewhere in the beginning. Look at the chart, what's happening to Equity with high MaximumRisk !
Attached Image
  • #877
  • Nov 16, 2006 12:32am Nov 16, 2006 12:32am
Cool idea Dagnar. Seeing all of the signals generated by the indicator someone whipped up, I was thinking about this as well. Thanks for doing this. Now it's time for testing the concept both forward/backward.

So, will this EA exit an open position and SAR after receiving x signals in the opposite direction? Or, are you just focusing on entries?

Cheers

Bill

Quoting Dagnar
Great job, cool EA, my respect, Hendrick.

Trying some backtests of different phoenix versions, results differs from doubling (and more) deposit to margin call. Looking at indicator for this EA (dot't remember by who, posted at championship forum), I've got an idea - open trade only by some consecutive signals.

added some code to EA( like this for sell signal) :

if (CurTime()-s_time>900) {ss=ss+1; bs=0;}
if (ss>=signal_count) {
res=OrderSend(...SELL....);
ss=0; s_time=CurTime();}

I think, it will help to filter out single signals at gaps, etc..

Made this for ph_4 (contest ver) and for ph_5 at mode1 (last ver) - the result is a new significant parameter for optimizing using pref_settings=true; For EURJPY, for example, I've got 50% increase in profit turning signal_count to 5 (not less than five consecutive signals opens trade). Surprisingly, this didn't reduces total amount of trades, only placing orders more precise.

Attaching both changed files, hope it's usefull contribution to phoenix development
  • #878
  • Nov 16, 2006 12:47am Nov 16, 2006 12:47am
Quoting Dagnar

...results differs from doubling (and more) deposit to margin call. Looking at indicator for this EA (dot't remember by who, posted at championship forum), I've got an idea - open trade only by some consecutive signals.

For EURJPY, for example, I've got 50% increase in profit turning signal_count to 5 (not less than five consecutive signals opens trade). Surprisingly, this didn't reduces total amount of trades, only placing orders more precise.
Sounds great, thank you!

Does both version shows similar increase in profit?

Did you happen to add that code, which shows buy/sell sign?

http://championship.mql4.com/2006/users/Hendrick/page2

I'm not sure if copy/paste is enough or I need to do something else.
Would you mind add those signals into EA so we can (who want) visually track them.

Btw, is it possible, to add switch, so you can turn on/off showing these buy/sell signals?

Mario
  • #879
  • Nov 16, 2006 1:56am Nov 16, 2006 1:56am
Hi Hendrik

At me in broad gullies Metatrader 4.0 it is written here
04:03:59 '54594': instant order sell 0.10 EURJPY at 150.5000 sl: 151.3400 tp: 150.0800
04:04:34 '54594': request was accepted by server
04:04:36 '54594': requote 150.5200 / 150.5600 for open sell 0.10 EURJPY at 150.5000 sl: 151.3400 tp: 150.0800

Used version EA 4,2,03. By default costs Requotes = 3. You could explain to me in what the reason?
At work of version EA 5,6,03 (Mode 2) there was such mistake in broad gullies Metatrader

2006.11.15 17:24:55 Phoenix_EA_v5_6_03 EURJPY, M15: terminated by timeout
  • #880
  • Nov 16, 2006 6:47am Nov 16, 2006 6:47am
if(Symbol() == "USDJPY")
{
SMAPeriod = 6;
SMA2Bars = 17;
Percent = 0.0032;
TakeProfit = 42;
StopLoss = 84;
TrailingStop = 0;
OSMAFast = 7;
OSMASlow = 21;
OSMASignal = 2;
EnvelopePeriod = 2;
Fast_Period = 25;
Slow_Period = 15;
DVBuySell = 0.0023;
DVStayOut = 0.023;
}
2 traders viewing now