Quoting PalmerDislikedFour winners in a row last night, MODE=1...why change anything....
Good luck Harry.Ignored
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
Quoting HendrickDislikedHere 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.Ignored
Quoting corner_hDislikedbut i do not have data till Nov.11 to compare exact numbers.Ignored
Quoting billworldDislikedI'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.
BillIgnored
Quoting billworldDislikedI'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;Ignored
QuoteDislikedAnd 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.
Quoting wackenaDislikedI 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.
WackenaIgnored
Quoting wackenaDislikedBill,
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%.
WackenaIgnored
Quoting wackenaDislikedI 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.
WackenaIgnored
Quoting billworldDislikedThanks 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).Ignored
Quoting DagnarDislikedGreat 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 developmentIgnored
Quoting DagnarDisliked
...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.
Ignored