Optimise an EA and get the best of the optimised ones. - page 28

 

Current results on favourites - on the previous page.

Another TC for overoptimisation is the CADCHF pair:

CADCHF ChnFlatDTS.mq5
CADCHF ChnFlatRTS.mq5
CADCHF ChnFlatSAR.mq5
CADCHF ChnFlatSP.mq5
CADCHF ChnTrendDTS.mq5
CADCHF ChnTrendRTS.mq5
CADCHF ChnTrendSAR.mq5
CADCHF ChnTrendSP.mq5
CADCHF EMAFIatDTS.mq5
CADCHF EMAT rendRTS.mq5

That is, all of them are channel-based and the other two are MA-based. (I have already optimized the others).

Experts to optimise should have a date of 07.04.18

If they are earlier - you need to download the latest versions from Yandex-disk

EALeague
EALeague
  • yadi.sk
View and download from Yandex.Disk
 
Georgiy Merts:

Current results on favourites - on the previous page.

Another TC for overoptimisation is the CADCHF pair:

CADCHF ChnFlatDTS.mq5
CADCHF ChnFlatRTS.mq5
CADCHF ChnFlatSAR.mq5
CADCHF ChnFlatSP.mq5
CADCHF ChnTrendDTS.mq5
CADCHF ChnTrendRTS.mq5
CADCHF ChnTrendSAR.mq5
CADCHF ChnTrendSP.mq5
CADCHF EMAFIatDTS.mq5
CADCHF EMAT rendRTS.mq5

That is, all of them are channel-based and the other two are MA-based. (I have already optimized the others).

Experts to optimise must have a date of 07.04.18

If they are earlier - you need to download the latest versions from Yandex-disk

I will make this"CADCHF ChnFlatDTS.mq5".

 
Aleksey Vyazmikin:

Making this"CADCHF ChnFlatDTS.mq5"

 
Aleksey Vyazmikin:

Checked, entered into the league.

You have 30 regcodes.

Only, Alexey, it is not very critical, you don't need to change anything, but still, next time - correct the beginning of the forward.

You have it on 31.08.17 and five months after April 15 is 15.09.17.

So, the current testing period should be 15.04.17-15.04.18, with forward from 15.09.17.


Prickle.

CADCHF ChnFlatDTS showed a good graph, but the quality is somehow - out of the blue... 50%. Surprised, looked into it...

It turns out that the quality is lowered due to too infrequent trades.)

In a year it turns out only 33 pieces. Not surprising - we are working on H4, rebound from the channel boundary is not such a common event after all.

 
Georgiy Merts:

Checked, entered into the league.

You have 30 regcodes.

Only, Alexey, it is not very critical, no need to change anything, but still, next time - correct the beginning of the forward.

You have it on 31.08.17 and five months after April 15 is 15.09.17.

I.e. at the current moment the testing period should be 15.04.17-15.04.18, with forward from 15.09.17


Prickle.

CADCHF ChnFlatDTS showed a good graph, but the quality is somehow - out of the blue... 50%. Surprised, looked into it...

It turns out that the quality is lowered because of too rare trades :)

It only works out at 33 per year. It's not surprising - working on H4, a bounce from the channel boundary is, after all, not such a frequent event.

I'd forgotten the logic behind it :) Tweaked it, maybe I'll put something else in for the night.

I think that rare deals are good too.

 
Aleksey Vyazmikin:

I think rare deals are also a good thing.

Rare deals are bad in that they don't show enough rich statistics.

I believe that a "good" indicator is 150 deals a year. If there are more than that, it doesn't matter, it doesn't affect the quality score. If it is less than 20, the quality score drops to zero. And 33 is not zero, but very little. The calculation gives only 10% of the quality. But, at the expense of other components, which are quite good - the overall quality score rises to 50%

 
Georgiy Merts:


CADCHF ChnFlatRTS.mq5

 

taken from

CADCHF ChnTrendSAR.mq5

CADCHF ChnTrendSP.mq5

CADCHF EMATrendRTS.mq5

add

I have not read the whole thread,

There are errors in the tester log

2018.04.16 09:25:33.283 Core 5 pass (0, 25) tested with error "incorrect input parameters" at 0:00:00.078

9871293 - SAR

9871294 - SP

9871295 - RTS


at RTS:

2018.04.16 10:40:09.228 Core 6 genetic pass (2, 510) tested with error "critical runtime error 517 in OnTick function (module -1 exception 0xc0000005)" in 0:00:15.625

MT 1795 Build

 

Aleksey Vyazmikin, Fast235- I will process by mid-day. Regcodes - I will take them into account. If I need them I will write them.

Fast235, error "incorrect input parameters" is quite correct. It means that the current parameter combination does not make sense (for example the Breakeven trigger is less than its level).

Error 517 is strange. I will try to contact the developers.

 

Alexei - it's OK, you've got 31 regcod.

Fast235,

unfortunately, you are the one with the XML file not from the "Forward Optimisation" tab.

You did everything right, but you didn't include the forward period. As a result, your optimization went on a full year period without forward period. Most of these passes are too erratic and it is impossible to select the best of them.

Considering the time you have wasted - I can provide you with one regcod. Write on which magik, and on which account number to generate it - I'll write it.

The other two codes I will generate if all three files are taken from the tab "Forward Optimization".

An error - let's see. I can't reproduce it in mine, it runs fine.