Forex Factory (https://www.forexfactory.com/)
-   Platform Tech (https://www.forexfactory.com/forum/69-platform-tech)
-   -   Is $10K -> $1M in 4 years impressive? ( My First EA ) (https://www.forexfactory.com/thread/1009318-is-10k-1m-in-4-years-impressive)

kasinath Jun 30, 2020 5:28am | Post# 1

Is $10K -> $1M in 4 years impressive? ( My First EA )
 
1 Attachment(s)
My EA is giving these backtest results, but it's leaving a LOT more money on the table (see attached chart). I posted about it asking for help but i've got no response. Silly me I expected people on the forum to be as excited as I was about the results, but now I'm wondering maybe this isn't a big deal after al.


Some context: I'm a long time software developer new to Forex (about 3 months in) and I've coded an EA that backtests with great profit. i've shared my methods in this other post where I ask for guidance.

I've seen people say that backtests results mean *nothing*. Is this true? Should I quit now because these results mean nothing?
I've also forward tested on a demo account for about a week with 100% win rate, but apparently demo testing also means nothing? ugh.
Click to Enlarge

Name: super-aggressive-gator-rsi.png
Size: 14 KB

braintheboss Jun 30, 2020 6:26am | Post# 2

My EA is giving these backtest results, but it's leaving a LOT more money on the table (see attached chart). I posted about it asking for help but i've got no response. Silly me I expected people on the forum to be as excited as I was about the results, but now I'm wondering maybe this isn't a big deal after al. Some context: I'm a long time software developer new to Forex (about 3 months in) and I've coded an EA that backtests with great profit. i've shared my methods in this other post...
Short answer:

1) MT4 backtest is useless
2) Demo account test is reliable depending broker. Some brokers haven't difference or very little differences between demo/real. In these brokers demo accounts are fine. For example I run system in demo account and copy to real accounts without any issue. Results difference are very little but due copier delay. Really are irrelevant

Then only way for be sure your system works is run it in different brokers with different leverages and see if results are same and green at the end

PS: You will not be rich trading forex even your system is good. Trading costs make very difficult get good net profit along time ( because you always will have bad moments for to be compensate )

Cnvrsn Jun 30, 2020 6:50am | Post# 3

My EA is giving these backtest results, but it's leaving a LOT more money on the table (see attached chart). I posted about it asking for help but i've got no response. Silly me I expected people on the forum to be as excited as I was about the results, but now I'm wondering maybe this isn't a big deal after al. Some context: I'm a long time software developer new to Forex (about 3 months in) and I've coded an EA that backtests with great profit. i've shared my methods in this other post...
Share the EA. I'll be happy to foward test and share results

emmzett Jun 30, 2020 7:15am | Post# 4

1 Attachment(s)
Is $10K -> $1M in 4 years impressive?
To answer your thread's title question: An average gain of 10.1% per month is not impressive, it's nice. However it would be impressive if you were indeed capable to consistently run and execute that strategy over a timespan of 4 years. That's not impossible but very very difficult and a complete different ball-park.

...I posted about it asking for help but i've got no response. Silly me I expected people on the forum to be as excited as I was...
You got no response because the content of your post is not useful. You posted a meaningless picture and a blackbox. Without source code, without a reproducable backtest (meaning inluded settings), without documentation of how your strategy is supposed to work you will not achieve much. You wrote about the directions you think to improve on but you provided no meaningful context.

Source code is a definite precondition and on that criterion alone you can measure the seriousness of everyone here. Each FF member saying otherwise - without a single exception - follows a (hidden?) agenda. If you think somebody can take your strategy and "steal your idea" you are far from where you want to get. The market is huge and the issues replicating a taken strategy in a profitable way are so numerous and not only related to trading that it could fill a book. As if it was enough to download a small text file with instructions from the internet to get rich.

...I've coded an EA that backtests with great profit...
Profit is not important, context of the profit is important. Average profit per trade in pip, profit to max. drawdown ratio, Sortino ratio, time under water etc. would be numbers providing context. The reason people say MT4 backtests are useless is because MT4/MT5 doesn't provide those numbers. Without it you are gambling and the odds are against you. And if you talk about timespans of years your chances of success go against zero.

However not all is lost. Very few here are capable to write software, so you don't need to rely on others to provide you with solutions to the afore mentioned issues. You may have a look at this book which covers many and more of what I mentioned. It may be helpful on your way:

Evaluation and Optimization of Trading Strategies [Pardo, 2008].pdf

Mahrgell Jun 30, 2020 8:30am | Post# 5

It is rather easy to write an EA that when backtesting makes those profits during the last 2 weeks.
(e.g. aim at German stocks, declare "if daily loss >10%, doom short that thing, if biweekly loss >98% reverse into long")

Does this mean anything about this being a smart algorithm? Does this backtesting result tell me anything about whether this algorithm should be used in the future on live trading?
No, not a damn thing.

So the main task after backtesting should always be to figure out, if you really found some "general" rules, or just overfitted to your existing data.

kasinath Jun 30, 2020 10:20am | Post# 6

{quote} To answer your thread's title question: An average gain of 10.1% per month is not impressive, it's nice. However it would be impressive if you were indeed capable to consistently run and execute that strategy over a timespan of 4 years. That's not impossible but very very difficult and a complete different ball-park. {quote} You got no response because the content of your post is not useful. You posted a meaningless picture and a blackbox. Without source code, without a reproducable backtest (meaning inluded settings), without documentation...
Great response! Thanks a ton for the context, and for the book. Going to read and soak it up. Some targeted responses:

  1. I did include a reproducable backtest by the way, with a settings file; it's in the zip. I mentioned this in my post.
  2. I did provide context on the strategy. Alligator + RSI.
  3. Sharing the source code is a sensitive topic as I have worked on it with friends' IP, so isn't entirely 100% mine to share even though I wrote it. I will speak with them though.
  4. Fantastic break down on how to properly evaluate the promise of a strategy! I never even heard of Sortino ratio. Much to learn.
  5. I appreciate the encouraging words re:my edge on writing software. Will use it as best I can.

Thanks again!


kasinath Jun 30, 2020 10:27am | Post# 7

It is rather easy to write an EA that when backtesting makes those profits during the last 2 weeks. (e.g. aim at German stocks, declare "if daily loss >10%, doom short that thing, if biweekly loss >98% reverse into long") Does this mean anything about this being a smart algorithm? Does this backtesting result tell me anything about whether this algorithm should be used in the future on live trading? No, not a damn thing. So the main task after backtesting should always be to figure out, if you really found some "general" rules, or just overfitted...
Thanks for your response. This is very true. I have seen backtests that just overfit to data.

Fortunately mine does not, it uses general rules based on commonly used indicator strategies.

I realize that is not immediately clear in the topic of this thread, but do see the thread referenced in my original post.

Here is the link again.
https://www.forexfactory.com/showthread.php?p=13034161

kasinath Jun 30, 2020 10:30am | Post# 8

{quote} Share the EA. I'll be happy to foward test and share results
Thanks so much!
I shared the EA. It is attached to the original thread. Here is the link again. Please reply to that original thread with any feedback.

https://www.forexfactory.com/showthread.php?p=13034161

kasinath Jun 30, 2020 10:33am | Post# 9

{quote} .o...way for be sure your system works is run it in different brokers with different leverages and see if results are same and green at the end.
...
You rich will not be rich trading forex even your system is good. Trading costs make very difficult get good net profit along time ( because you always will have bad moments for to be compensate )
These are gems! Thanks a ton. I will try this out.

If you don't mind please share with me the broker you use, if it's not against the forum rules to do so. Otherwise I would appreciate if you shared with me in a direct message.

laikos Jun 30, 2020 11:19am | Post# 10

I'm not meant to pour cold water on your EA. Just try it on cent account live and you will quickly realize you have missed something. I've been through this all too many times

braintheboss Jun 30, 2020 11:27am | Post# 11

{quote} These are gems! Thanks a ton. I will try this out. If you don't mind please share with me the broker you use, if it's not against the forum rules to do so. Otherwise I would appreciate if you shared with me in a direct message.
IC Markets, HotForex and Darwinex at this moment. But live accounts are in IC Markets due European leverage

kasinath Jun 30, 2020 11:55am | Post# 12

{quote} IC Markets, HotForex and Darwinex at this moment. But live accounts are in IC Markets due European leverage
Nice one. Will check them out. Thanks!

kasinath Jun 30, 2020 11:55am | Post# 13

I'm not meant to pour cold water on your EA. Just try it on cent account live and you will quickly realize you have missed something. I've been through this all too many times
That's great advice. I didn't even know cent accounts existed. Will give this a try.

LDFX Jul 1, 2020 7:12am | Post# 14

{quote} That's great advice. I didn't even know cent accounts existed. Will give this a try.

FXTM offers cent account.

kasinath Jul 1, 2020 10:51am | Post# 15

{quote} FXTM offers cent account.
Thanks! I ended up getting one with Alpari. It's testing now.

LDFX Jul 3, 2020 4:31pm | Post# 16

{quote} Thanks! I ended up getting one with Alpari. It's testing now.
Great, I wish you good luck in your testing. I didnít know that Alpari provided it also, it is a good way to test a strategy with proper position sizing without much risk.

By the way I just wanted to answer you that a EA with such return is not impressive at all, but on a live trading on 4 years that would be absolutely amazing.

Also the drawdown was very sharp/sudden and huge. Have you tested it on 10 years ?

kasinath Jul 4, 2020 11:39am | Post# 17

{quote} Great, I wish you good luck in your testing. I didnít know that Alpari provided it also, it is a good way to test a strategy with proper position sizing without much risk. By the way I just wanted to answer you that a EA with such return is not impressive at all, but on a live trading on 4 years that would be absolutely amazing. Also the drawdown was very sharp/sudden and huge. Have you tested it on 10 years ?
Thanks! I am making progress. Learning alot.

I now realize that the return in my EA is a lie haha. I say this because I have learned that my parameters are overfit! (i didnt even know what that meant some days ago).

I have not yet tested it on 10 years, but I think i need to refine my trend detection strategy; right now i have a really bad TP/SL ratio (100:3400). I now know this is a failure. I will be posting more here in this forum with questions so I'll keep you up to speed.

here's my most recent post: https://www.forexfactory.com/showthread.php?p=13041738

Thanks again for your help!

LDFX Jul 4, 2020 4:21pm | Post# 18

{quote} Thanks! I am making progress. Learning alot. I now realize that the return in my EA is a lie haha. I say this because I have learned that my parameters are overfit! (i didnt even know what that meant some days ago). I have not yet tested it on 10 years, but I think i need to refine my trend detection strategy; right now i have a really bad TP/SL ratio (100:3400). I now know this is a failure. I will be posting more here in this forum with questions so I'll keep you up to speed. here's my most recent post: https://www.forexfactory.com/showthread.php?p=13041738...
Glad you are making progress, backtesting is useful for that, choosing the right parameters, but donít get fooled by the amazing returns a EA can show.

Indeed your r:r ratio is very bad.

I asked you about the 10 years backtesting because if you had began it just prior the huge drawdown you would thought that this strategy was bad, while perhaps it is, it might also be because you were using too much leverage or bad r:r ratio.

But keep in mind that real live trading will give you different results, the conditions are not the same. Backtesting however is a great to gauge the parameters such as risk per trade, maximum total exposure, r:r ratio, etc...

Anyway I wish you good luck in your forward testing with the cents account, smart choice, try also a one year demo account if your cent account goes to zero.

Trading is a journey, innit ?


cheers,

LDFX Jul 4, 2020 4:26pm | Post# 19

{quote} Thanks! I am making progress. Learning alot. I now realize that the return in my EA is a lie haha. I say this because I have learned that my parameters are overfit! (i didnt even know what that meant some days ago). I have not yet tested it on 10 years, but I think i need to refine my trend detection strategy; right now i have a really bad TP/SL ratio (100:3400). I now know this is a failure. I will be posting more here in this forum with questions so I'll keep you up to speed. here's my most recent post: https://www.forexfactory.com/showthread.php?p=13041738...
Glad you are making progress, backtesting is useful for that, choosing the right parameters, but donít get fooled by the amazing returns a EA can show.

Indeed your r:r ratio is very bad.

I asked you about the 10 years backtesting because if you had began it just prior the huge drawdown you would have thought that this strategy was bad, while perhaps it is, it might also be because you were using too much leverage or bad r:r ratio.

But keep in mind that real live trading will give you different results, the conditions are not the same. Backtesting however is a great tool to gauge the parameters such as risk per trade, maximum total exposure, r:r ratio, etc...

Anyway I wish you good luck in your forward testing with the cents account, smart choice, try also a one year demo account if your cent account goes to zero.

Trading is a journey, innit ?


cheers,

kasinath Jul 4, 2020 5:05pm | Post# 20

{quote} Backtesting however is a great tool to gauge the parameters such as risk per trade, maximum total exposure, r:r ratio, etc...
Precisely. But to your point, i will also test broader periods. My strategy didnt do very well before 2016 (facePalm).

On the note of gauging parameters: any chance you have any tips on analysing MFE / MAE for mt4 backtests? It seems if i can plot this, it will help me pick reasonable stops.

{quote} Trading is a journey, innit ? cheers,
You said it, man.


© Forex Factory