Symphonie EA

[SteveHopwood]

19th JUNE 2012. THIS EA IS A FAILURE AND HAS BEEN ABANDONED. PLAY WITH IT ON DEMO IF YOU WISH, BUT DO NOT EXPECT SUCCESS.

SERIOUS WARNING

• Most Forex traders lose all their money.
  Using the robot posted here in trading Forex does not guarantee success.
  Trading this robot could lead to serious financial loss.
  Trading this robot without understanding its underlying trading strategies guarantees traders will lose their money.
  This is not a set-and-forget ea. There is no such thing and anyone who tells you there is, is either a liar or an idiot. This EA will need manual intervention when NoOfTrades buys and sells are open, and the only recourse from there is to manage the basket manually.
  At best, a trading robot is only 90% as good as the manual strategy it trades. At best. At worst, it can be much less effective. If the strategy is rubbish, so is the robot.
  To trade this robot, you have to understand:
  • - How to use EA's.
    - Evaluator's Symphonie trading system described at http://www.stevehopwoodforex.com/phpBB3 ... 2402#p2402

DO NOT TRADE THIS EA ON A LIVE ACCOUNT. IT IS EXPERIMENTAL.

Ehup all

Intro
This EA has its genesis at http://www.stevehopwoodforex.com/phpBB3 ... 2402#p2402 Read about it first before continuing.

We have cuzgeorge to thank for the EA. A contributor asked me to take a look at Symphonie shortly after I started this forum, but I declined because of my dislike of trading systems that require several pretty-coloured thingies to line up. Given the time and research that Evaluator put into developing his system, this was a mistake. George has been exceptionally generous to me this week, so when he asked me to look at Symphonie, I could not refuse - a good thing, as it turns out.

We have Evaluator to thank for the trading system. Take a bow, Evaluator.

Required indicators
You need the four Symphonie indi's attached here in your Indicators folderfor the EA to work. You do not need them on your chart. Also note that you need the inputs in the EA's Symphonie indi inputs to match those of the indi's on your charts if the ea output is to match what you see on the chart.

Extreme (actually TVI V2) colours the histo's red when they are falling, and blue when rising. If two histo's are the same, the second retains the colour of the first. The position of the histo's re the zero line is irrelevant; Symphonie is only concerned with their direction.

Time frames
Read this post by Evaluator: http://www.forexfactory.com/showthread. ... ost5470997 for an explanation of his 60-5 method. Symphonie defaults to this ootb. The time frame inputs are:

• - HigherTimeFrame: enter this input as minutes, so M1 is 1, D1 is 1440 etc. A value of 0 means the time frame of the chart onto which S is loaded.
  - LowerTimeFrame: enter this input as minutes A value of 0 disables mtf trading, should you not wish to use this.

Trade Exit
valuator's three exit strategies are represented by the exit strategy inputs. If you choose UseExitStrategy2, make sure your lot size is divisible by 2; there are no idiot checks. If you are using the multi time frame option, these strategies are tested on the lower time frame also, if ApplyStrategiesToLtf is set to true (default false).

ARS
ARS is fully described in post 1 of Evaluator's thread. UseARS is the input that turns this off. It is on by default. The total of lot sizes is calculated from all the open trades on the account, not just those owned by Symphonie. Evaluator rates this money-management as central to his trading success, so turn it off at your peril and do not blame me if you subsequently blow your account.

Other stuff
The usual stuff is included - Hanover, CBI, trend determination, ATR for stops/tp's etc. No Recovery or Caterpillar as I see no use for them here. I have attached my Shell user guide which has explanations for a lot of the inputs. Ignore inputs described in the guide that you cannot find in the EA inputs.

Symphonie variants by other members
Please feel free to start threads within this forum to offer your own variants on all this.

In conclusion

Have fun.



Matters of general interest

Read this post:
http://www.forexfactory.com/showthread. ... ost4923611 Newbie traders, read this for inspiration and warning. Those of you who have suffered losses, read it for inspiration. There is a thread devoted to Sam's trading method at http://www.forexfactory.com/showthre...25#post5118825

Masterly summary of hedging: http://www.forexfactory.com/showthread. ... ost4977179

EA coding
I receive may requests to code EA's. Traders, if I code your EA for free, then I will share it here. If you want me to code your EA without sharing, then my fee is $100 payable into my paypal account. I explain why at http://www.stevehopwoodforex.com/phpBB3 ... p=803#p803, at the bottom of the post.

Useful utilities/EA's/scripts
Matt's chart-population script at http://www.stevehopwoodforex.com/phpBB3 ... 5733#p5733. Use this every time you start a fresh platform.
MaryJane has posted a fabulous utility to allow us to resize EA input windows: http://www.forexfactory.com/showthread.php?t=299017
Multi purpose trade manager: http://www.forexfactory.com/showthread.php?t=89371
Account equity analyser: http://www.forexfactory.com/showthread. ... ost4693716 Also at http://www.forexfactory.com/showthread. ... ost4709098, which includes a template for both USD and the Euro.
Past-trades display indi by squalou; shows a graphical display of the success/failure of trades in your History tab. Breathtaking piece of work. http://www.forexfactory.com/showthread. ... ost3830610
My shell EA – the code I use to produce EA's so quickly. http://www.stevehopwoodforex.com/phpBB3 ... p?f=5&t=15

[mbkennel]

One concern. Symphonie-Sentiment-Indicator uses information from the future, meaning that on the chart is it different from real-time.

I bet also that when doing Empty4's backtesting, it will erroneously use the value of the indicator as computed on full historical data (i.e. using data from the future).

Offending code (ugly decompiled stuff):

Code: Select all

   for (int li_120 = 0; li_120 < li_116; li_120++) {
      l_high_104 = High[iHighest(NULL, 0, MODE_HIGH, period, li_120)];
      l_low_96 = Low[iLowest(NULL, 0, MODE_LOW, period, li_120)];
      ld_16 = (High[li_120] + Low[li_120]) / 2.0;
      ld_48 = 0.66 * ((ld_16 - l_low_96) / (l_high_104 - l_low_96) - 0.5) + 0.67 * ld_56;
      ld_48 = MathMin(MathMax(ld_48, -0.999), 0.999);
      g_ibuf_80[li_120] = MathLog((ld_48 + 1.0) / (1 - ld_48)) / 2.0 + ld_80 / 2.0;
      ld_56 = ld_48;
      ld_80 = g_ibuf_80[li_120];

You see that the loop is going up from 0 to li_116, meaning from later time to earlier time.
But "ld_56" and "ld_80" values are from the previous iteration---which is in the future relative to the one in li_120.

A correction would be to change the for loop to go from li_116-1 down to zero. I also bet the results will look not quite as good.

Note, this technique is probably a great way for EA scammers to sell fake results.

[SteveHopwood]

One concern. Symphonie-Sentiment-Indicator uses information from the future, meaning that on the chart is it different from real-time.

I bet also that when doing Empty4's backtesting, it will erroneously use the value of the indicator as computed on full historical data (i.e. using data from the future).

Offending code (ugly decompiled stuff):

Code: Select all

   for (int li_120 = 0; li_120 < li_116; li_120++) {
      l_high_104 = High[iHighest(NULL, 0, MODE_HIGH, period, li_120)];
      l_low_96 = Low[iLowest(NULL, 0, MODE_LOW, period, li_120)];
      ld_16 = (High[li_120] + Low[li_120]) / 2.0;
      ld_48 = 0.66 * ((ld_16 - l_low_96) / (l_high_104 - l_low_96) - 0.5) + 0.67 * ld_56;
      ld_48 = MathMin(MathMax(ld_48, -0.999), 0.999);
      g_ibuf_80[li_120] = MathLog((ld_48 + 1.0) / (1 - ld_48)) / 2.0 + ld_80 / 2.0;
      ld_56 = ld_48;
      ld_80 = g_ibuf_80[li_120];

You see that the loop is going up from 0 to li_116, meaning from later time to earlier time.
But "ld_56" and "ld_80" values are from the previous iteration---which is in the future relative to the one in li_120.

A correction would be to change the for loop to go from li_116-1 down to zero. I also bet the results will look not quite as good.

Note, this technique is probably a great way for EA scammers to sell fake results.

Hehe. Matt, indi code robs me of the will to live. The moment I see decompiled code, my interest vanishes faster than the factual content of a politician's interview reply.

I am not going to argue and so have made the change you recommend - uploaded to post 1.

[cuzgeorge]

If i were the King, and the Queen was my wench, I would bestow the legendary knighthood upon thee, as well as the key to my cities and all the virgins with-in them.

i cannot thank you enough Steve, i really think you are a genius, and in such record time too! Will explore and reply here.

very good point MBKENNEL, well spotted and thank you to Evaluator for spending months to come up with his system.

kind regards to all and happy pipping to.
Hopwood, you're the best.
yours unconditionally,
cuzgeorge

[mbkennel]

Attached is a modified version (for visual use) where you can change a boolean parameter (UseInformationFromTheFuture from true to false) and see how it changes.

Spoiler: everything lags by a few bars and real life you'd get whipsawed in contrast to the pretty picture in the stock version.

The underlying methods use a bit of math which I recognize, it took somebody semi smart to program it in the first place. So I'm surprised there was such an elementary blunder.

How it works: It transforms the position of the current price to within hi-low bounds to -0.5 and 0.5 and then does an EMA in it. Then it takes a Fisher transformation, converting uniform into Gaussian distributions (which have central limit behavior) and then takes another EMA of that.

The provenance as a decompiled version makes me suspicious/

Here's what my spidey sense thinks happened. A real indicator with some modest value written by a mathey guy was intentionally modified by a very clever and unscrupulous commercial scammer at some point (don't think anybody here is involved).

Then it was sold as a system which lo-and-behold could even be backtested by the customer-victims with seemingly great results (in Empty4), and yet it was much more likely to fail in reality.

Later it got decompiled and pieces of code passed around not knowing the downside, and it washes up here.

I think the other indicators have some forward-snooping too. In my day job (machine learning software) the jargon for this !@#*! is "target leak".

The difference between a newbie and an oldbie in my business. Newbie invents new variable, trains model finds it's really good and predictive, goes to the top of the list of most contributing variables. He is proud and shows plots and numbers to his boss. Oldbie, e.g. his boss, looks at a great new variable which is really good and predictive and thinks, "Not again! What kind of new bug is crawling out of the woodwork?"

[NicoScholtz]

I am not going to argue and so have made the change you recommend - uploaded to post 1.

Steve, thank you and all the other contributors for this wonderful site!

An observation: I think you slipped when you made the change referred to above.
Your present Symphonie_Sentiment indi in post 1 as at 2012/03/18 09:00 GMT has the updated iteration loop in line 64 as:
for (int li_120 = li_116; li_120 > 0; li_120--) {

I believe this should be:

for (int li_120 = li_116; li_120 >= 0; li_120--) {

Regards
Nico

[mbkennel]

for (int li_120 = li_116-1; li_120 >= 0; li_120--) {

[SteveHopwood]

I am not going to argue and so have made the change you recommend - uploaded to post 1.

Steve, thank you and all the other contributors for this wonderful site!

An observation: I think you slipped when you made the change referred to above.
Your present Symphonie_Sentiment indi in post 1 as at 2012/03/18 09:00 GMT has the updated iteration loop in line 64 as:
for (int li_120 = li_116; li_120 > 0; li_120--) {

I believe this should be:

for (int li_120 = li_116; li_120 >= 0; li_120--) {

Regards
Nico

Cheers Nico. Change made and uploaded to post 1. Coders, you know what to do.

[SteveHopwood]

Matt, I added a note to post 1 to make sure that people read your posts.

[squalou]

The provenance as a decompiled version makes me suspicious/

Here's what my spidey sense thinks happened. A real indicator with some modest value written by a mathey guy was intentionally modified by a very clever and unscrupulous commercial scammer at some point (don't think anybody here is involved).

Then it was sold as a system which lo-and-behold could even be backtested by the customer-victims with seemingly great results (in Empty4), and yet it was much more likely to fail in reality.

Later it got decompiled and pieces of code passed around not knowing the downside, and it washes up here.

I think the other indicators have some forward-snooping too. In my day job (machine learning software) the jargon for this !@#*! is "target leak".

The difference between a newbie and an oldbie in my business. Newbie invents new variable, trains model finds it's really good and predictive, goes to the top of the list of most contributing variables. He is proud and shows plots and numbers to his boss. Oldbie, e.g. his boss, looks at a great new variable which is really good and predictive and thinks, "Not again! What kind of new bug is crawling out of the woodwork?"

Matt, this is an excellent summary of the situation !

This is what made me run away from Symphonie system a minute after having "dissected" Symphonie's indicators code when the system came out...
Same fallacy (fraud?...) as TMA, SolarWind, Snake, and so many other similar so-called "indicators" out there...

I think your post should be put on its own as a sticker at the same level as "Important things about Empty4" at least... if not even at a higher level !

Empty4 users MUST be warned of such practice.


Sq