The Simple System | Page 29 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Please suggest a simple & effective Method/System of "Trading System" Forum to newbie 182 replies

Day Trading system - Simple system to make money 51 replies

Another great Breakout GBP/JPY System (Dachel's Simple system) 62 replies

My Simple System (Can we automate this system?) 96 replies

An Easy System (With the help of ES Trader - Another Simple System) 13 replies

Attachments: The Simple System
Exit Attachments

The Simple System

  • #561
  • Jun 11, 2007 9:27pm Jun 11, 2007 9:27pm
Quoting farimir
Snobol,

is the EA posted in post 558 the same one you are talking about in post 556?
Yes. I am talking about FishTSS but with my new algorithm.

Chandra, I posted a statement for the GBPUSD pair hope that is helpful

Last night (Sunday night EST), I attached that EA to several charts, this Monday afternoon, 3 pm the USDJPY was +13p the AUDUSD up +12p the EURUSD down -3p. other charts had no trades.

I have attached the zip folder again so that there will be no confusion as to what I am writing about. I use one HOUR charts, but it should work on a 15 min chart too. The AUDUSD is 15min, all others are one hour.

Chandra, thanks for the info about the tester, I always though that 90% modelling quality meant just that, modelling of the EA against applied data.

Cheers,
Pip2
  • #562
  • Jun 11, 2007 9:41pm Jun 11, 2007 9:41pm
Quoting achandrasekaran99
Hi Pip2,

I checked the files and the results are the best I have ever seen. But I have difficulty getting same backtest results. For example, the best Profit Factor I got was 7.4, win ratio was 85% and my total trades were only 20 (you had them as 18.5, 93% and 46 trades). This is after I ran the optimization of the parameters which took about 100 minutes (which were very close to those in the .set file you have given). Still impressive results for a low drawdown. But my number of trades were so less. I checked the code and still I don't find anything that could have affected my results. Any pointer will be greatly helpful.
Thanks.
chandra
Chandra, I don't know what to say. Yes the optimisation takes a long time. Basically I optimize all parameters except "Lots" so it chugs away and I do other things while the optimizer works. I have no explanation for the difference and am certainly not padding the results in any way. I leave Lots at .5. If you use the .set parameters I have included, and use the time period I have used the results should be identical, if your data are identical to mine.

As an experiment I will try to send you my "History" and you could test it against those data to see if that's where the difference lies. Beyond that am stumped. I too, thought the results were too good to be true but also note, I have tested other pairs which were NFG . That's no F.. Good, lol. GBPUSD gave the best results with the EURO not far behind. Drawdowns in all cases were minimal or zero.

Cheers,
Pip2

  • #563
  • Jun 11, 2007 10:20pm Jun 11, 2007 10:20pm
I have put this in my expert adviser file. But the little hat is gray and it should be blue. Any ideas?


Quoting Snobol1
Great Matt, I have been using H1 for all my tests. See the files posted below. The zip file has 4 files a settings file an ex4 a FishTSS and a readme file. The FishTSS was sent in by Jam, a contributor and I added the algorithm to buy/sell without indicators.

The EA should wait until a top or bottom is established and a retracement starts then it will take a position. So be patient, it will take a few hours on a one hour chart. I have not tried lower TF's, so that means you can do so and report back. Hope you have good luck with it.

Cheers,
Pip2
  • #564
  • Jun 11, 2007 10:39pm Jun 11, 2007 10:39pm
Quoting Snobol1
Chandra, I don't know what to say. Yes the optimisation takes a long time. Basically I optimize all parameters except "Lots" so it chugs away and I do other things while the optimizer works. I have no explanation for the difference and am certainly not padding the results in any way. I leave Lots at .5. If you use the .set parameters I have included, and use the time period I have used the results should be identical, if your data are identical to mine.

As an experiment I will try to send you my "History" and you could test it against those data to see if that's where the difference lies. Beyond that am stumped. I too, thought the results were too good to be true but also note, I have tested other pairs which were NFG . That's no F.. Good, lol. GBPUSD gave the best results with the EURO not far behind. Drawdowns in all cases were minimal or zero.

Cheers,
Pip2
Hi Pip2,

Thank you for the system. I can check with your History files. I think for this type of trades, the data will not be generally that different. But worth to check it. I think, if I can get same number of trades, the results should be similar. So, I am mostly puzzled by less than half of the trades I had.

If it gave similar results for another pair, that is a good thing (shows this is not just coincidence). Backtesting on longer period of time and forward testing should confirm the results (even part of the performance will be extremely great).

-chandra
  • #565
  • Jun 11, 2007 11:49pm Jun 11, 2007 11:49pm
Hi Snobol1, Chandra, all,

Newbie that I am, I've never backtested before (but I will!). However, in researching it, I've come across some documents that could be helpful.

Check out the Yahoo Metatrader group email file collection at:
http://www.forexmt4.com -> _MT4_Tutorials -> ...

There are several documents on backtesting, as well as plenty of other useful MT4 training documents. One has the enticing title, "Backtesting with 99 percent modelling V1.doc". Besides the tutorials folder, there are also indicators, EA's, scripts, etc. So much I've barely scratched the surface...

I also found another document elsewhere I'll upload here. Perhaps there may be some tips that will help improve backtesting accuracy.

pips4life
Attached File(s)
File Type: pdf 16524-90-modelling-quality-modellingquality.pdf   331 KB | 602 downloads
  • #566
  • Jun 12, 2007 12:05am Jun 12, 2007 12:05am
One more thing about backtesting...

The documents I've glanced at talk about importing M1 data (or raw tick data is better). You then use Period_Converter script to create longer time period charts, like M15 or H1(M60) or H4(M240) using multipliers.

However, if two people use different MT4 brokers, the "0:00" time may differ. Some use true GMT, others use London, others use EST, and so on. If, for example, Snobol1 and Chandra use brokers with a 1-hour offset difference, then the candles for M15 and H1(M60) *should* still be nearly the same, however any time period above H1 (e.g. H4) may have a different starting time for the candle period. The H4 candles may look different because of the 1,2, or 3 hour offset.

I came across a periodcon_v1p5.mq4 indicator which includes a time-offset feature so in theory, one should be able to create H4 bars that match what another broker uses. For usage, please read the notes in the indicator. It's kind of tricky...

pips4life
Attached File(s)
File Type: mq4 periodcon_v1p5.mq4   16 KB | 631 downloads
  • #567
  • Jun 12, 2007 8:11am Jun 12, 2007 8:11am
Hi Snobol1,

I tried to test the EA and attached it on Sunday, but it opened multiple positions at once on the same currency and blew my account (demo fortunately). Any idea what happened?

By the way, thatnks for your work on this.

Paul.
  • #568
  • Jun 12, 2007 11:12am Jun 12, 2007 11:12am
Have the indicators changed since thread one?
  • #569
  • Edited 10:12pm Jun 12, 2007 3:45pm | Edited 10:12pm
Quoting stav01
Hi Snobol1,

I tried to test the EA and attached it on Sunday, but it opened multiple positions at once on the same currency and blew my account (demo fortunately). Any idea what happened?

By the way, thatnks for your work on this.

Paul.
Hi Paul, I haven't a clue as to what went wrong. If you tried it on the GBPUSD and used the .set files I included you should have some nice demo profits.

If you tried on another pair, you will have to first optimize for a period say Mar 07 to June 07 or something like that. I have the "Lots" set at .5 and optimized for .5. If you use larger lots you're on your own. I typically use 5K for my demo and 0.5 for Lots. I know some people use a million dollar demo account which is absolutely ridiculous.


Also all my testing has been on the one hour TF H1. You must optimize before testing. Thats all I can say.

Cheers,
Pip2
  • #570
  • Jun 12, 2007 10:46pm Jun 12, 2007 10:46pm
Yeah, it opened and closed multiple orders on me also. Took 8 trades in less than 2 minutes and luckily I was watching and deleted the EA before it cleaned out my account. This EA is definately corrupted. It took 200$ from my live acct before I got it shut off. Gos to show you u should test it on demo. Bad mistake on my part.



Quoting stav01
Hi Snobol1,

I tried to test the EA and attached it on Sunday, but it opened multiple positions at once on the same currency and blew my account (demo fortunately). Any idea what happened?

By the way, thatnks for your work on this.

Paul.
  • #571
  • Jun 12, 2007 10:49pm Jun 12, 2007 10:49pm
Someone needs to fix this EA before it gets spread all thru FF and people start losing their money...Corrupted big time




]Hi Paul, I haven't a clue as to what went wrong. If you tried it on the GBPUSD and used the .set files I included you should have some nice demo profits.

If you tried on another pair, you will have to first optimize for a period say Mar 07 to June 07 or something like that. I have the "Lots" set at .5 and optimized for .5. If you use larger lots you're on your own. I typically use 5K for my demo and 0.5 for Lots. I know some people use a million dollar demo account which is absolutely ridiculous.


Also all my testing has been on the one hour TF H1. You must optimize before testing. Thats all I can say.

Cheers,
Pip2[/quote]
  • #572
  • Jun 13, 2007 12:27am Jun 13, 2007 12:27am
Any new EA should be backtested for many hours to test and optimize, and then forward tested on several demo accounts with many odd and unusual parameters selected. If people start loosing their money on this new EA posted in a public forum it will be their own fault from now on. They have been warned. I appreciate the release of the EA with the source code. Thanks. We can all play with the code to fix the problems that show up.

In backtesting the EA shows the repeated trade problem more often as you increase the amount of history tested. For example one of the PDFs below is an optimized EURUSD daily backtest for 4 months that looks good.

As soon as you add another 1 - 14 months of history to the beginning or end of the good test period the problem will show up. For some reason once the EA begins to rapidly open and close trades it can't correct itself. In the example below the same EURUSD daily with 2 weeks added gets stuck in the loop. See the 4 months + 2 weeks file. The problem can occur on the most recent week or several months ago.

After 4 days of tests the the EA made at least a small profit over a few months on just about every time frame if the error was avoided. If the glitch can be fixed the EA will merit further forward tests.
Attached File(s)
File Type: pdf Euro Daily 4 months.pdf   43 KB | 463 downloads
File Type: pdf Euro Daily 4 months and 2 weeks.pdf   61 KB | 408 downloads
  • #573
  • Edited 2:56pm Jun 13, 2007 2:21pm | Edited 2:56pm
I don't see why this EA was opening multiple positions. The code opens orders only after checking if the total existing order is zero for the Symbol. If the total orders for the symbol is other than zero, no order will be opened. Even if the EA was attached to several charts of same symbol that seems impossible (as per the code). So, it is difficult to understand how that happened. I have checked backtest and running forward test and never seen more than one order at a time.

The only problem I can see with this EA is that it will not open new positions if there was a position on the same Symbol (no matter who opened it, another EA or manually) and more importantly, it will apply its rules of closing and TS for any open order present for that symbol! So it is more cofusing to me how this EA was able to open more than one position at the same time.

Only one possibility I can think of is that it might have happened during highly active period, where the execution of orders were delayed by several ticks and the program did not see the earlier order filled, and continued to place new orders until some filled order showed up in the Open Positions list. All EAs do the work at every tick (unless specifically designed to work differently). This is just a possibility, but never has happened to me with this or any other EA (in demo or live).

Opening, closing and then opening is a different problem. That can happen. This EA uses two different strategies for calculating Buy and Sell (just this morning I wrote to Pip2 about it). Surprisingly (or not), at some point both conditions may be met (it looks like according to nvwine's example it happened on 01-Feb-07, on daily chart of EURUSDm; I did not even think such possibility, but now understand that it is possible!). Since the BUY order is first in the code, always it will open Buy positions and stop the calculations. Then at next tick, the checking part of the code will see the Sell signal and close the opposite position and we go in that again and again. I think this must be it.

For any EA (even if paid a lot), the proper order of use is: backtest, forward test and slowly increase the lots (micro, mini, standard). Just my opinion.
-chandra
  • #574
  • Jun 13, 2007 4:03pm Jun 13, 2007 4:03pm
Quoting achandrasekaran99
Opening, closing and then opening is a different problem. That can happen. This EA uses two different strategies for calculating Buy and Sell (just this morning I wrote to Pip2 about it). Surprisingly (or not), at some point both conditions may be met (it looks like according to nvwine's example it happened on 01-Feb-07, on daily chart of EURUSDm; I did not even think such possibility, but now understand that it is possible!). Since the BUY order is first in the code, always it will open Buy positions and stop the calculations. Then at next tick, the checking part of the code will see the Sell signal and close the opposite position and we go in that again and again. I think this must be it.
-chandra
Thanks for writing Pip2. It may just need a simple repair. I would like to stress that considering the quality of backtesting, the problem may not show up for everyone on Feb 01. For the tests in the PDFs I used history data from a live account collected from forward running. As I said before I did get the problem to repeat on almost all tests if I increased the testing period enough. On some of the other tests I had to import data so - I'm sure we all know the validity issues. I'm convinced that it could happen going forward if the conditions are right. Although it could take quite a while.

After a closer look at the statement mry2001 attached in post#548 it shows the same problem. The test is progressing normally until March 29th. The first time I looked at the report I failed to think about the increased spread on GBPJPY which is the basic size of the loses.
  • #575
  • Jun 13, 2007 7:08pm Jun 13, 2007 7:08pm
Quoting nvwine
Thanks for writing Pip2. It may just need a simple repair. I would like to stress that considering the quality of backtesting, the problem may not show up for everyone on Feb 01. For the tests in the PDFs I used history data from a live account collected from forward running. As I said before I did get the problem to repeat on almost all tests if I increased the testing period enough. On some of the other tests I had to import data so - I'm sure we all know the validity issues. I'm convinced that it could happen going forward if the conditions are right. Although it could take quite a while.

After a closer look at the statement mry2001 attached in post#548 it shows the same problem. The test is progressing normally until March 29th. The first time I looked at the report I failed to think about the increased spread on GBPJPY which is the basic size of the loses.
Hi nvwine,

If it was the above problem as I think, the following changes to the code should correct it. (Pip2 will be working on it, this is possibly a quick fix).
if( BuySig != 0 ) // old line 101
if( SellSig != 0 ) // old line 110

change them to :
if (BuySig != 0 && SellSig ==0) // new line 101
if (SellSig != 0 && BuySig == 0) // new line 110

(or one can use the file I am attaching, which has only these two corrections done)

Then run the backtest for the same period that gave problem. If you (or anyone) can report back here whether this stopped the problem or not that will be useful (I don't seem to have the data for this).

-chandra
(usually I dont look at zip files, so I did not read the previous report by mry2001)
Attached File(s)
File Type: mq4 FISHTSS_EA1acs.mq4   5 KB | 597 downloads
  • #576
  • Edited 8:03pm Jun 13, 2007 7:53pm | Edited 8:03pm
Hi all, I have not yet experienced any of those difficulties which makes the problem, if there is one, to solve. Now I have never gotten around to testing with over 0.5 lots. And I always optimize ALL read that ALL parameters at one time. Sure it takes longer but optimizing 1 or 2 or 3 parameters at a time is NOT the same as optimizing all together. Is that the problem? I don't know yet.

The "core logic" seems to operate ok for me. If not Jam could intercede and fix things for us if indeed it needs fixing.

Chandra I sent you an e-mail re buy/sell algorithm. As planned, the buy and sell algorithms are identical except for the < and> expressions which are necessary in ANY logic equation. Now the parameters are something different. Some parameters are in pips and others in bars. Take a look at my explanation of how it works in a previous post (547). Correct parameters are a must and it seems that not all pairs give as good results as GBPUSD.

As a result I do not even test weak performers. All pairs are not created equal and this algorithm surely points that out. To those who tested on a live account without proper demo testing I say. what's next? Would you drive a new car WITHOUT BRAKES!!! Its insane! Give me a break.

As Chandra said, and as all prudent investors know, backtest first, then forward test on demo, then go live with a micro account, then mini then full-blown. Its like teaching a child to walk, baby steps first with help and gradually progress to walking then running then flying then... Like I said before, use at your own risk.

I've attached a Terminal shot of my trades on a demo account and so far no probs and it looks pretty good. I've used the comment "BARS ONLY" to identify those trades.

As far as opening and closing multiple orders I cannot see how my buy/sell algorithm can trigger that. Chandra says that a delay in confirmation of a buy might cause a sell to be triggered. I cannot see that happening, because for a buy a bottom must appear and for a sell a top must exist, I don't think a bottom AND a top can exist at the same ( or nearly the same) time. Maybe am wrong, so I'll keep testing.

The 2 losing positions are from another EA am testing. It does not use stops(bad idea) and you can see the effect on the account. That -800 is now over -1000! That's why I demo trade for testing.

Cheers
Pip2
Attached Image (click to enlarge)
Click to Enlarge Name: BARS ONLY.jpg Size: 36 KB
  • #577
  • Jun 13, 2007 8:24pm Jun 13, 2007 8:24pm
Quoting achandrasekaran99
Then run the backtest for the same period that gave problem. If you (or anyone) can report back here whether this stopped the problem or not that will be useful (I don't seem to have the data for this).

-chandra
(usually I dont look at zip files, so I did not read the previous report by mry2001)
I used your modified EA and it was a big improvement. Both of the attached backtests would not even have worked before. The EURO has the exact same parameters as the former backtests I posted. This time it did not have 14 pages of small loses even though it was for more weeks. There does still seem to be a slight problem. If you look at the reports there are a few examples on each pair where it opened and closed multiple trades in a minute. I no longer know if this will show up in forward tests or if it is a real problem. Unusual things happen in back tests. Thanks for your work on this.

A note to others forward testing the EA: Follow the time frames recommended by the developers of the EA to start with. Don't follow my time frames or currency pairs. They may be poor performing combinations. I try to optimize the settings for a lot of currency pairs on many different time frames. Then I compare the results and start several forward tests. Both of the attached tests are probably over optimized to the last 2 month's curve.
Attached File(s)
File Type: pdf Euro Daily 5 months.pdf   49 KB | 480 downloads
File Type: pdf GU 30min 5 months.pdf   87 KB | 473 downloads
  • #578
  • Edited Jun 14, 2007 1:22pm Jun 13, 2007 9:08pm | Edited Jun 14, 2007 1:22pm
Quoting Snobol1

Chandra I sent you an e-mail re buy/sell algorithm. As planned, the buy and sell algorithms are identical except for the < and> expressions which are necessary in ANY logic equation. Now the parameters are something different. Some parameters are in pips and others in bars. Take a look at my explanation of how it works in a previous post (547). Correct parameters are a must and it seems that not all pairs give as good results as GBPUSD.

Chandra says that a delay in confirmation of a buy might cause a sell to be triggered. I cannot see that happening, because for a buy a bottom must appear and for a sell a top must exist, I don't think a bottom AND a top can exist at the same ( or nearly the same) time. Maybe am wrong, so I'll keep testing.

Cheers
Pip2
Hi Pip2,

I did not say the second order will be a sell. It will be the repeat of whatever order was placed before (most likely a Buy because that comes first in the code).

I think it is possible for a price to meet buy and sell conditions when different parameters are used as we do here. A down-move (or down-correction) with recovery (a buy signal), can be viewed as a down-move and correction (a sell signal), even with same parameters (but with same parameters it will be only one or other at any point even if it exchanges by one or two pip moves). But with different parameters the possibility of overlap of opposite signals becomes real.

It seems my fix has reduced the problem drastically as nvwine has reported. A few repeat orders are still there, I don't know what caused those. May be that will need the observation of it happening in realtime along with the circumstances.

-chandra
  • #579
  • Edited Jun 14, 2007 10:05am Jun 13, 2007 9:28pm | Edited Jun 14, 2007 10:05am
Simple indicator . Copy to Expert\Indicator Folder , load it into any chart and wait for a buy or sell signal alert. Suitable for all pair TF30. Maybe this one can help us.

New revision SimpleV8.2 own color change plus addition filter last update June 14 2007 10.45 pm
Regards,

GoldDigger
Attached File(s)
File Type: ex4 SimpleV8.ex4   12 KB | 659 downloads
  • #580
  • Jun 13, 2007 10:59pm Jun 13, 2007 10:59pm
Quoting achandrasekaran99
Hi Pip2,

I did not say the second order will be a sell. It will be the repeat of whatever order was placed before (most likely a Buy because that comes first in the code).

I think it is possible for a price to meet buy and sell conditions when different parameters are used as we do here. A down-move (or down-correction) with recovery (a buy signal), can be viewed as a down-move and correction (a sell signal), even with same parameters (but with same parameters it will be only one or other at any point even if it exchanges by one or two pip moves). But with different parameters the possibility of overlap of opposite signals becomes real.

It seems my fix has reduced the problem drastically as nvwine has reported. A few repeat orders are still there, but that will need the observation of it happening in realtime along with the circumstances (such as volatility).

-chandra
OK Chandra, I misquoted you. I think if conditions are set up for a buy, a buy will occur on the tick that caused it, but if conditions to buy are still there on the next tick then another buy ( or sell) could happen. I think that is what was going on.

With Chandra's fix (thank you) when once a buy is triggered, ( i.e. buysig=1) another buy or sell is inhibited. Somehow I have not experienced this at all, so I have no experience with it. I backtested Chandra's fix on GBPUSD H1 and got the same results as before. So the fix was not detrimental to the overall operation of the EA.

I don't think there is any EA out there with absolutely no problems. So please don't expect the perfect EA to appear anywhere anytime soon. I think all EA's are a work in progress and improvements could come at any time. On this thread, at least, constructive criticisms are welcome and are productive. Jam, Chandra, Isto, Oleg77,and others have made this thread a good one to learn and contribute and above all get rich!lol.

A word about Time Periods. If you change the time period e.g. from 1 month to 2 months you MUST re-optimize and update the set file. Failure to do so will result in unpredictable results. If done on a live account you will lose your shirt, and maybe your pants too. Please don't lament poor performance if you have not optimized for the TF and period you are testing. Optimize ALL parameters at one time, because even though the B/S algorithms are the same, their parameters are not!

Happy testing.
Cheers,
Pip2
0 traders viewing now