EA31337 released for free | Page 19 | 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

  • #361
  • May 30, 2018 7:33am May 30, 2018 7:33am
Where is the set file for 1.073?

I couldn't find it in github.

https://github.com/EA31337/EA31337-L...EURUSD/default

contains the set for 1.072 if I understand correctly.
  • #362
  • May 30, 2018 7:51am May 30, 2018 7:51am
Quoting chrisdk2017
Where is the set file for 1.073? I couldn't find it in github. https://github.com/EA31337/EA31337-L...EURUSD/default contains the set for 1.072 if I understand correctly.
All release files, like v1.073 got all the latest optimized params embedded into the source code, so the compiled version comes with the default best-known params out-of-box. So I don't have any extra/better SET files on top of it, than the default params.
  • #363
  • May 30, 2018 7:55am May 30, 2018 7:55am
Quoting kenorb
{quote} All release files, like v1.073 got all the latest optimized params embedded into the source code, so the compiled version comes with the default best-known params out-of-box.
Isnt latest version 1.075 ? Why do you recommend 1.073 over latest?

Also the back tests you ran were those on Mt4 ?

What settings you used in your backrests was it standard or did you change a single thing then we must know it to test it.

Thanks
Game taught me the game. And it didnt spare me the rod while teaching
  • #364
  • Edited 8:56am May 30, 2018 8:04am | Edited 8:56am
Quoting Singh1
{quote} Isnt latest version 1.075 ? Why do you recommend 1.073 over latest? Also the back tests you ran were those on Mt4 ? What settings you used in your backrests was it standard or did you change a single thing then we must know it to test it. Thanks
The v1.075 is trading even lesser, as it only few strategies enabled as it was aiming for the tests with even higher spreads (like 40, but then optimization system disabled half of the strategies), and people start complaining it doesn't trade for hours, secondly, it got a small startup init issue so people are a bit confused how to use it, so probably it needs more work. On the other hand, since few first posts, people got excited about 1.072, so moving to 1.073 is the less drastic change for now, so we can compare the results, understand the weaknesses and move forward. I don't know the versions beyond that (can't find the full backtest for 1.074), I'll have to dig the tests or re-run, but further versions weren't optimized after the logic changes/improvements, so they may not work well as good as the previous one with the defaults. I'll provide soon the backtests for all the versions so they can be compared.
  • #365
  • May 30, 2018 8:52am May 30, 2018 8:52am
Thanks for the update and clarification kenorb, will test v1.073 now as 1.075 only trades once in 2 days ish xD

May green pips be with you sir.
  • #366
  • Edited 10:29am May 30, 2018 9:07am | Edited 10:29am
Furthermore, it seems v1.074 got some issues, as only 3 type of indicators are enabled by default, and out of 5 strategies (Envelopes M15&M30, MACD M30, RSI M15&M30), for some reason only one (MA M15) was profitable. The results are still good, but something went wrong during the process. So I'm attaching backtest for v1.074 for 2015 which I've found, as can't find any for the other years. Of course, you can re-enable the other ones manually using params which you like and test.

Attached Image (click to enlarge)
Click to Enlarge Name: EURUSD-2000USD-2015year-20spread-M30-DS-test.gif Size: 17 KB


EA31337-Lite-Backtest-v1.074
Period 30 Minutes (M30) 2015.01.06 00:00 - 2015.12.30 23:30 (2015.01.01 - 2015.12.31)
Bars in test 12404 Ticks modelled 80271172 Modelling quality 89.28%
Initial deposit 2000.00 Spread 20
Total net profit 1089.81 Gross profit 3992.04 Gross loss -2902.22
Profit factor 1.38 Expected payoff 4.10
Absolute drawdown 565.96 Maximal drawdown 782.01 (35.29%) Relative drawdown 35.29% (782.01)
Total trades 266 Short positions (won %) 87 (29.89%) Long positions (won %) 179 (33.52%)
Profit trades (% of total) 86 (32.33%) Loss trades (% of total) 180 (67.67%)

Still isn't bad as for only one strategy (other 4 not profitable), so I'm not sure exactly what went wrong. It probably requires optimization of all the disabled strategies and re-enabling them. Please also note that the above test is for a low deposit of 2k USD, not 10k as previous ones (so it got less flexibility), and high spread of 20 points.

Then v1.075 got only logic improvements and fixes, no optimization has been done, probably that's why it doesn't trade often (as per above). So it's on the current todo list.
Attached File(s)
File Type: txt EURUSD-2000USD-2015.01.06-2015.12.30-20spread-(5)-M30-report.txt   4 KB | 274 downloads
File Type: txt EURUSD-2000USD-2015year-20spread-M30-DS-test.txt   8 KB | 282 downloads
  • #367
  • Edited 11:58am May 30, 2018 11:39am | Edited 11:58am
Quoting Singh1
{quote}Also the backtests you ran were those on Mt4?
Yes, these tests were run using MT4 platform run under Linux Ubuntu using automated scripts. Exact details of the dataset which was used to run the tests are also embedded onto image (period, how many ticks/bars and the result it-self).

Quoting Singh1
{quote}What settings you used in your backrests was it standard or did you change a single thing then we must know it to test it.
The provided backtest results are just running EX4 it-self (no any external SET files were loaded) after the final optimization process, in other words, the default params were used for the given release file which are basically the best-known optimized params embedded into the binary.
  • #368
  • May 30, 2018 12:04pm May 30, 2018 12:04pm
[quote=kenorb;11110541]{quote} That means you're lack of proper data for these dates. Testing on the broker data isn't reliable, as there could be too many gaps and low-quality data. You set testing from 01.03, but your data starts from 12.04, you want to end on 31.05, but your tick data stops on 24.05, so you can't rely on this. {quote}


what must i do about it sir?
  • #369
  • May 30, 2018 12:09pm May 30, 2018 12:09pm
[quote=Tabary;11116458]
Quoting kenorb
{quote} That means you're lack of proper data for these dates. Testing on the broker data isn't reliable, as there could be too many gaps and low-quality data. You set testing from 01.03, but your data starts from 12.04, you want to end on 31.05, but your tick data stops on 24.05, so you can't rely on this. {quote} what must i do about it sir?
I've provided the backtesting steps here. Otherwise, if you install Docker for Windows with Linux container support, I can show the alternative, quicker way.
  • #370
  • May 30, 2018 12:14pm May 30, 2018 12:14pm
[quote=kenorb;11116487]
Quoting Tabary
{quote} I've provided the backtesting steps here. Otherwise, if you install Docker for Windows, I can show the alternative, quicker way.

got it, i'am on my way

thank you sir
  • #371
  • May 30, 2018 12:21pm May 30, 2018 12:21pm
Quoting zeramlakireg
{quote} I've already shared the set file for EURUSD only TE. Just look it up in the attahments (paperclip icon on top of the page). The mql market has a lot of EAs doing 1,000% daily in backtests but will blow your account in a couple of days (even hours)! So, don't trust your backtest/optimization if you are not kenorb. XD Just forward test the EA. Also, don't expect other account's performance to be replicated in yours. You will start forward test in a different market condition and your account history (where the EA bases its lot sizing) are...

sir, i think what Mr singh1 meant is to share the .set file to be copy it in the preset folder inside our mq4 folder

i would like that too if you don't mind sharing
  • #372
  • May 30, 2018 12:52pm May 30, 2018 12:52pm
hi all
seems different broker have different trades
I test old version. 1.072, use default setting (on EURUSD M15)

at the same time. this week.
a) on SVSFX demo account, when I enable the EA, it open the trade Sell and Buy on my MT4
b) on IronFX demo account , when I enable the EA, it only open the SELL
c) on FXTM-cent demo account , it only open Sell trade.
does someone have the same problem?
I will try other version...
  • #373
  • Edited 6:07pm May 30, 2018 4:00pm | Edited 6:07pm
Quoting oocat0227
hi all seems different broker have different trades I test old version. 1.072, use default setting (on EURUSD M15) at the same time. this week. a) on SVSFX demo account, when I enable the EA, it open the trade Sell and Buy on my MT4 b) on IronFX demo account , when I enable the EA, it only open the SELL c) on FXTM-cent demo account , it only open Sell trade. does someone have the same problem? I will try other version...
Some strategies are set to trade with trend, some aren't. Check the trend detection description in post #178 and #180. So possibly different strategies are triggered on different brokers, but you'll have to compare the magic numbers, which are probably different. There are hundreds of factors taken into the account (like conditions and filtering systems in place) which are affecting how EA opens the trades, like slight price differences between brokers can trigger different conditions, also account balance/stats, broker spreads/leverage, existing orders, calculated lot size, free margins, slight delays, and many other factors could trigger different trades. You can also take the screenshots of the screen and compare the differences, or compare the initialization and trade logs which can give you more information. Another way is to enable debug messages.
Basically, each run is unique and difficult to predict why certain trades are triggered, and other not on different brokers. This will require a new feature focusing on it and logging that kind of things. Normally without any filters, EA could potentially receive hundreds of buy/sell signals per minute, and logging such data, such as why certain orders weren't triggered, it can only slow the EA and generate very huge logs.
  • #374
  • May 31, 2018 2:06am May 31, 2018 2:06am
from 5.1 to 5.30 one month on demo account.
Attached Image (click to enlarge)
Click to Enlarge Name: QQ图片20180531150431.png Size: 24 KB
  • #375
  • May 31, 2018 2:22am May 31, 2018 2:22am
[quote=kenorb;11116487]
Quoting Tabary
{quote} I've provided the backtesting steps here. Otherwise, if you install Docker for Windows with Linux container support, I can show the alternative, quicker way.
Could you show us the Docker way?

I guess this is for the latest version, 1.075?
  • #376
  • May 31, 2018 3:27am May 31, 2018 3:27am
Quoting kenorb
Furthermore, it seems v1.074 got some issues, as only 3 type of indicators are enabled by default, and out of 5 strategies (Envelopes M15&M30, MACD M30, RSI M15&M30), for some reason only one (MA M15) was profitable. The results are still good, but something went wrong during the process. So I'm attaching backtest for v1.074 for 2015 which I've found, as can't find any for the other years. Of course, you can re-enable the other ones manually using params which you like and test. {image} EA31337-Lite-Backtest-v1.074 Period 30 Minutes (M30) 2015.01.06...
Hy.....I've start with 1.073 EUR/USD TF 30 since yesterday afternoon but no any trade till now..... No any errors.. Is it normal?
Thanks
Regards
  • #377
  • Edited 5:22am May 31, 2018 4:29am | Edited 5:22am
Quoting flashlj2008
from 5.1 to 5.30 one month on demo account. {image}
Very pity. This version of EA was optimized for 2014-2015, so 3 years behind, as it was released in December 2016. So it doesn't take into the account bad situation happening in 2017 and 2018, that's probably why. It needs to be fully re-optimized. I can help with the optimization process.

Quoting cescof
{quote} Hy.....I've start with 1.073 EUR/USD TF 30 since yesterday afternoon but no any trade till now..... No any errors.. Is it normal? Thanks Regards
Probably similar to above, it was only backtested for 2014-2016. It needs re-optimization or debugging the reason. I'll try to run the backtest for 2017 and see what's going on. The backtest result was showing around 1 trade per day on average, so maybe you need to wait a bit longer, I'm not sure.
  • #378
  • May 31, 2018 6:02am May 31, 2018 6:02am
This EA is really nicely written. You have put a lot of effort into it. The code is very easy to understand and is nicely structured.

What it lacks, is some "brains" that would allow the EA to dynamically adjust based on market conditions. Please don't get me wrong. It's a really nice EA, and the foundation for making something amazing out of it is already there.

Thanks for sharing the code with us mortals.
  • #379
  • Edited 6:41am May 31, 2018 6:13am | Edited 6:41am
[quote=chrisdk2017;11118744]
Quoting kenorb
{quote} Could you show us the Docker way? I guess this is for the latest version, 1.075?
I've published Docker container for backtesting and supported containers can be pulled from Docker's hub. The main container (latest) doesn't have any backtest data by default, so it needs to download each time, but in the next days/weeks I can provide tagged containers per each pair and year, so they'll come with data ready for testing, so all files needs to be downloaded only once (so once you've got container downloaded, backtest or optimization test can be started in few seconds with a single command). Here is the list of supported historical data (I'm still populating the data, DS means Dukascopy feed). These repositories consist of raw CSV files which are converted to MT4 HST/FXT format by the scripts for all the timeframes (including custom ones). On request, any source/broker data can be added, including cryptocurrencies, so the new data can be used for backtesting/optimization.

Once Docker is installed (on Windows, Linux or macOS), you can pull the container by (run in the terminal/cmd):
docker pull ea31337/fx-mt-vm

Then for help run: docker run ea31337/fx-mt-vm help
For backtesting param, run: docker run ea31337/fx-mt-vm run_backtest

To run the backtest, you can use supported params as displayed by above help command.

Here is the example command for backtesting v1.075 for 2017.01-02 on EURUSD with spread 10:
docker run -e BOOT_CODE='file_get https://github.com/EA31337/EA31337/r...ull-v1.075.ex4' ea31337/fx-mt-vm run_backtest -e EA31337 -y 2017 -m 1-2 -p EURUSD -s 10 -v

To test the different version, change it in two places in above command.

There are many custom params, so backtest or optimization tests can be fully customized. In the next weeks, I'll be improving and simplifying backtesting processes for Docker, so any version of EA can be re-optimized. So if you need any particular feature, please let me know.
  • #380
  • Edited 6:44am May 31, 2018 6:27am | Edited 6:44am
Quoting SlicEnDicE
What it lacks, is some "brains" that would allow the EA to dynamically adjust based on market conditions.
I was thinking exactly the same way. I was attempting to add RSI_DynamicPeriod param (some logic in RSI_CheckPeriod()), but I don't know any equation which allows detecting the current best period (without trials and errors), and the optimization process only finds the average one, so it's something to work on (it's the limitation of all EAs based on technical indicators). The larger optimization period, the values are actually more flatten, so it's safer to use, but less flexible and profitable at the same time.

I'll be working on the refactor soon, I hope, so I was planning to add dummy trades, which aims to be the simulation of the real trades, but only happening in the EA's memory, so based on the dummy/hidden profit happening in the memory (like opening dummy trades for different periods and comparing the profits), EA can find the best params (such as current best periods and other params), for the current dataset, in order to use it before actually starting to trade, then opening only certain type of trades after they were tested and proven in the memory.

Another method is to tap the params into machine learning algorithm and calculate the best params along as the EA trades.
0 traders viewing now