EA31337 released for free | Page 7 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

No Free Lunch but all the Free Coffee you can drink 736 replies

SWAP free, rollover free Broker? 15 replies

Hi...best time to place order after news released 0 replies

Flyer! Version 2.00 is released. 0 replies

Cad News Released 1 Min Before... ! 43 replies

Attachments: EA31337 released for free
Exit Attachments

EA31337 released for free

  • #121
  • Edited 2:49pm Apr 24, 2018 1:35pm | Edited 2:49pm
Quoting imwim
{quote} I have checked the value of envelopes[CURR][MODE_MAIN] during the backtesting process and 0 (zero) is a valid value which comes during the test, and non zero values comes also ... MODE_MAIN have not any significance for the iEnvelopes indicator, and only MODE_UPPER and MODE_LOWER are used ... Source: https://www.mql5.com/en/articles/1496 Also, MQLQ4 Reference from MetaEditor mode [in] Indicator line index. It can be any of Indicators line identifiers enumeration value (0 - MODE_MAIN, 1 - MODE_UPPER, 2 - MODE_LOWER). MODE_MAIN...
Given this TestEnvelopes.mq4 test which I've created, I've run iEnvelopes() during OnInit() in the container on M15 and M30 timeframes with all modes: MODE_LOWER, MODE_MAIN, MODE_UPPER, using MetaTrader 4 build 1010, and all return positive values (which I believe is correct). It is possible that MODE_MAIN was removed in MT5, however, I think it could the side effect after you've replaced the MT4 terminal.exe with MT5. Replacing just metaeditor.exe only for compilation purposes is enough. Although I haven't tested this in the latest build, but I think this should be back compatible as I don't think they can just remove MODE_MAIN from MQL4's iEnvelopes(). And I think the mentioned article is mistaken about not including MODE_MAIN.

Result of printing MODE_LOWER/MODE_MAIN/MODE_UPPER (for M15, and same for M30):
2016.01.05 01:00 TestEnvelopes EURUSD,M15: Envelopes M1 : 1.08147/1.08255/1.08363 => CORRECT
2016.01.05 01:00 TestEnvelopes EURUSD,M15: Envelopes M5 : 1.08168/1.08277/1.08385 => CORRECT
2016.01.05 01:00 TestEnvelopes EURUSD,M15: Envelopes M15 : 1.08184/1.08292/1.084 => CORRECT
2016.01.05 01:00 TestEnvelopes EURUSD,M15: Envelopes M30 : 1.08109/1.08217/1.08325 => CORRECT
2016.01.05 01:00 TestEnvelopes EURUSD,M15: Envelopes M60 : 1.08485/1.08594/1.08702 => CORRECT
  • #122
  • Apr 24, 2018 3:17pm Apr 24, 2018 3:17pm
hi, can someone help me? the ea doesnt open any order. the smiley face is on. i guess its not mt4 settings. it has to do with the ea itself. please help
Attached Image (click to enlarge)
Click to Enlarge Name: doesnt open order.png Size: 150 KB
  • #123
  • Apr 24, 2018 3:56pm Apr 24, 2018 3:56pm
Quoting keyo
hi, can someone help me? the ea doesn't open any order. the smiley face is on. i guess its not mt4 settings. it has to do with the ea itself. please help {image}
Difficult to say, as EA is in active mode and it should work. Not sure how long you're running it, but the market pace is slow (40ticks/min), and not much is going on, so nothing is triggered. Another thing is to check, how many strategies were enabled during startup in your Journey log, and make sure some weren't disabled due to some initial problems. Also, as far as I remember, v1.075 has fewer strategies enabled by default than previous versions or some of them could have too many bitwise filtering layers configured (such as SignalMethod and OpenCondition, but if you set towards 0 or 0 - it'll trigger more frequent trades) and it may not work for the current market conditions. This is due that I couldn't find the ideal optimization settings for some after some fixes and logic changes (as it takes a lot of effort), so I left for others to decide which they want to enable. If that's the case, enable more strategies (or play with SignalMethod/OpenCondition), however, it's worth to backtest them first. Otherwise, previous versions were configured differently.
  • #124
  • Apr 24, 2018 4:09pm Apr 24, 2018 4:09pm
Quoting kenorb
{quote} Difficult to say, as EA is in active mode and it should work. Not sure how long you're running it, but the market pace is slow (40ticks/min), and not much is going on, so nothing is triggered. Another thing is to check, how many strategies were enabled during startup in your Journey log, and make sure some weren't disabled due to some initial problems. Also, as far as I remember, v1.075 has fewer strategies enabled by default than previous versions or some of them could have too many bitwise filtering layers configured (such as SignalMethod...
hi mr kenorb. i ran it for more than 8 hours. is it normal for this ea to never open order in this time period?
  • #125
  • Edited 5:00pm Apr 24, 2018 4:48pm | Edited 5:00pm
Quoting keyo
{quote} hi mr kenorb. i ran it for more than 8 hours. is it normal for this ea to never open order in this time period?
Not normal for this EA when it's running in the healthy state, however, I already explained this in the simplest way as I could in post #123, but I think you didn't read it, as you're asking the same question in a different way. I've suggested you some actions to take, but it seems you come back with no further details. Have you checked the Journal Log, how many strategies were activated, and so on, as I suggested? What else have you tried? Without full troubleshooting of the problem which may take some time, I cannot help you. Secondly, the project is provided as it is, however you are free to configure it according to your needs. Alternatively, you can always download the previous version's in EX4 format and try them instead and check the difference.
  • #126
  • Apr 24, 2018 4:54pm Apr 24, 2018 4:54pm
Quoting kenorb
{quote} Given this TestEnvelopes.mq4 test which I've created, I've run iEnvelopes() during OnInit() in the container on M15 and M30 timeframes with all modes: MODE_LOWER, MODE_MAIN, MODE_UPPER, using MetaTrader 4 build 1010, and all return positive values (which...
The TestEnvelopes.mq4 is ok only for PERIOD_M30, for the other TFs two parameters changed, (int) (Envelopes_MA_Period * ratio) and Envelopes_Deviation * ratio2 ...

ratio = tf == 30 ? 1.0 : pow(Envelopes_MA_Period_Ratio, fabs(chart.TfToIndex(PERIOD_M30) - chart.TfToIndex(tf) + 1));
ratio2 = tf == 30 ? 1.0 : pow(Envelopes_Deviation_Ratio, fabs(chart.TfToIndex(PERIOD_M30) - chart.TfToIndex(tf) + 1));

ratio = ratio2 = 0.01 for PERIOD_M15, only 1% of the 1.0 used for PERIOD_M30 ...


and iEnvelopes returns 0 (zero), a kind of bad parameters used ...
I am who I am
  • #127
  • Edited 7:29pm Apr 24, 2018 5:19pm | Edited 7:29pm
Quoting imwim
{quote} The TestEnvelopes.mq4 is ok only for PERIOD_M30, for the other TFs two parameters changed, (int) (Envelopes_MA_Period * ratio) and Envelopes_Deviation * ratio2 ... ratio = tf == 30 ? 1.0 : pow(Envelopes_MA_Period_Ratio, fabs(chart.TfToIndex(PERIOD_M30) - chart.TfToIndex(tf) + 1)); ratio2 = tf == 30 ? 1.0 : pow(Envelopes_Deviation_Ratio, fabs(chart.TfToIndex(PERIOD_M30) - chart.TfToIndex(tf) + 1)); ratio = ratio2 = 0.01 for PERIOD_M15, only 1% of the 1.0 used for PERIOD_M30 ... and iEnvelopes returns 0 (zero), a kind of bad parameters used...
The ratio system uses M30 timeframe as the baseline, and it calculates different periods for different timeframes according to the provided ratio parameters (such as Envelopes_MA_Period_Ratio & Envelopes_Deviation_Ratio), one reason, to avoid an overkill in the number of params (different period per timeframe). In order to fix the problem and use the same ratio for all timeframes, change Envelopes_MA_Period_Ratio & Envelopes_Deviation_Ratio to 1.0 (healthy range is 0.5-1.5). The reason why it has been set so low (0.1), was the result of the optimization system which for some reason "decided" it's better to set it so low, but couldn't predict the problems with the EA initialization. There could be two reason why the optimization system set extreme ratio values of ratio params for some strategies, either that configuration worked the best, or the opposite, to suppress the low-timeframe strategies from trading due to not able to find the ideal configuration. So if low ratios as 0.1 for Envelopes breaks the initialization process, the fix needs to be applied to the rules files used by the optimization system such as Envelopes_MA_Period_Ratio.rule and Envelopes_Deviation_Ratio.rule (these are the rules used by other scripts). I'll apply the fix for the next time (e.g. to not go lower than 0.5 for Envelopes), when I'll be running the optimization process. I hope that makes sense. To summarize, you need to reset these ratios to 1.0 and re-run optimization which ratios would work the best for you.
  • #128
  • Apr 24, 2018 5:28pm Apr 24, 2018 5:28pm
Quoting keyo
{quote} hi mr kenorb. i ran it for more than 8 hours. is it normal for this ea to never open order in this time period?
Just do a quick back-test for the current month for M30 time frame. I have 64 trades in 17 trading days, 64 / 17 = 3.76 trades / day ...

Ok kenorb, good to know to have an eye opened for the optimization process ...
I am who I am
  • #129
  • Apr 24, 2018 7:08pm Apr 24, 2018 7:08pm
Quoting imwim
{quote} Just do a quick back-test for the current month for M30 time frame. I have 64 trades in 17 trading days, 64 / 17 = 3.76 trades / day ... Ok kenorb, good to know to have an eye opened for the optimization process ...
what set file do you use? can you upload it here? im using the default setting.
  • #130
  • Edited 8:06pm Apr 24, 2018 7:44pm | Edited 8:06pm
Quoting imwim
{quote} Just do a quick back-test for the current month for M30 time frame. I have 64 trades in 17 trading days, 64 / 17 = 3.76 trades / day ... Ok kenorb, good to know to have an eye opened for the optimization process ...
It is still a low number of trades compared to previous versions. In the latest version (branch), only 11 strategies have been activated out of hundreds, because the optimization system couldn't find the settings optimal for the disabled one. So this is something that needs work. Not necessary the logic (which needs to be refactored soon), but the challenge is to find the best optimal settings (I'm using EURUSD 2015-2017 with spread 20 as the baseline)

Unfortunately, I can't provide the best settings for the latest version (the default are the best to my knowledge and aren't perfect), as each change to EA's logic, all the tests need to be re-run and it requires enormous processing power to find its best configuration (at least few weeks with hundreds of instances running against different components/strategies, then running dozens of final-stage tests on top of that). For the optimization system, the following rule files are used, around 200 rules, each testing different components of EA, sometimes running for days in a row, sometimes some needs to be re-run multiple times to re-adjust with other params.
  • #131
  • Apr 24, 2018 11:06pm Apr 24, 2018 11:06pm
1.072 is making lots of trades in my demo account. 2% plus in one day so far.

In the backtest it only loses - equity curve going down to zero - which I don't understand.
  • #132
  • Edited 1:26am Apr 25, 2018 1:15am | Edited 1:26am
Quoting keyo
{quote} what set file do you use? can you upload it here? im using the default setting.
I'm using the default settings too ...

Quoting chrisdk2017
1.072 is making lots of trades in my demo account. 2% plus in one day so far. In the backtest it only loses - equity curve going down to zero - which I don't understand.
Idem. More good results in back-testing when boost is off (fixed sizes 0.01 lots for example) ...

The back-testing run with fixed spread, but the real/demo run win dynamic spread, so the results will be always different ...

In the Asian session we have a huge spread, a case ignored in back-tests ...
I am who I am
  • #133
  • Edited 10:04am Apr 25, 2018 4:35am | Edited 10:04am
Quoting chrisdk2017
1.072 is making lots of trades in my demo account. 2% plus in one day so far. In the backtest it only loses - equity curve going down to zero - which I don't understand.
To my understanding, backtests are very not reliable in MT4 platform by design. I think it's just broken on many levels and cannot be trusted. It requires a lot of hacks, but still, I couldn't find the reliable method.
First, MT4 doesn't support dynamic spread by design (there are some hacks, but they may not work). For example, backtest result will get you smooth huge profits when the market is violate (that's the reason why this EA closes trades after 24h by default as for workaround to not rely on any advantages/lucks), in real trading, the broker will kill your profits with very high spreads, or will keep re-quoting your orders during that "lucky" period.
Secondly, when you set the current spread which is selected by default (it's not the same as dynamic, but fixed along the test), you'll get different results at the different time when you run it, so set it to 10 or 20 instead as for workaround.
For the third, I still don't understand the significance of HST files during the backtesting, if this data is re-calculated during the test as the ticks flow (related post #110), but I believe they're not. If the technical data during backtesting is fed from HST files (technical indicator data) for the current bars (for "performance" reasons), that would be the worse (that's why I've implemented param to ignore some ticks to do some experimentation). In real trading, technical data changes as the ticks go.
For backtesting and optimization, I'm using my own scripts for backtesting which aims to run backtest using more reliable methods, but still could be not reliable. It still needs some debugging to get more facts whether MT4 backtesting can be trusted or not at all.
  • #134
  • Apr 25, 2018 5:09am Apr 25, 2018 5:09am
Quoting kenorb
{quote} To my understanding, backtests are very not reliable by design in MT4. I think it's just broken on many levels and cannot be trusted. It requires a lot of hacks, but still, I couldn't find the reliable method. First, MT4 doesn't support dynamic spread by default (there are some hacks, but they may not work). Secondly, when you set the current spread, you'll get different results every time you run it, so set it to 10 or 20 instead. For the third, I still don't understand the significance of HST files during the backtesting, if this data...
Thanks.

I know Steve Hopwood and friends make profitable EAs with no backtesting at all, they only use forward testing.

Your EA with the set for 1.072 seems profitable so far on EURUSD.

I have seen so called perfect backtests for EAs with 99.9% data and when put live they lose all the time.
  • #135
  • Apr 25, 2018 5:24am Apr 25, 2018 5:24am
unfortunately the EA does not work for me yet.

But before I continue my search for errors ...

Has the EA already been tested for several days on a demo / live ??
How is the result?

As already said, backtests can be a lie! I already had EA, which ran perfectly in the backtest. She had not survived a week on the demo ....

Thank you for your response!
  • #136
  • Apr 25, 2018 5:37am Apr 25, 2018 5:37am
Quoting alexf82
unfortunately the EA does not work for me yet. But before I continue my search for errors ... Has the EA already been tested for several days on a demo / live ?? How is the result? As already said, backtests can be a lie! I already had EA, which ran perfectly in the backtest. She had not survived a week on the demo .... Thank you for your response!
To quote myself:

"1.072 is making lots of trades in my demo account. 2% plus in one day so far. "
  • #137
  • Apr 25, 2018 6:13am Apr 25, 2018 6:13am
Quoting chrisdk2017
{quote} To quote myself: "1.072 is making lots of trades in my demo account. 2% plus in one day so far. "
Can you post your preset file please?
  • #138
  • Apr 25, 2018 7:03am Apr 25, 2018 7:03am
Quoting shiva
{quote} Can you post your preset file please?
It is the preset file you posted in https://www.forexfactory.com/showthr...8#post10984688
  • #139
  • Apr 25, 2018 10:32am Apr 25, 2018 10:32am
Quoting chrisdk2017
{quote} It is the preset file you posted in https://www.forexfactory.com/showthr...8#post10984688
LOL
  • #140
  • Apr 26, 2018 5:51am Apr 26, 2018 5:51am
+1.96% on demo today, very good
0 traders viewing now