Can you optimize this backtest (speed comparison)

Viewing 15 posts - 16 through 30 (of 85 total)

← 1 2 3 4 5 6 →

Author

Posts
06/30/2020 at 2:07 PM #137711 quote

GraHal
Participant

Master

zilliq wrote:

But I don’t understand why it takes only 2 minutes for 60 secondes on the nicola’s computer

Switch off and then log back in again and do the optimise backtest straight away as a first task and tell us if it is any faster?

I say this because I notice that when the RAM used by PRT gets to 2.5 GB ish then it starts getting jittery etc.

06/30/2020 at 2:29 PM #137715 quote

zilliq
Participant

Master

I did and it’s the same thing @Grahal

Java takes 2,52 Go in RAM (on 16 Go), so I can’t have less

Can you test with the MTF code even in 10.3, to see, Thanks

06/30/2020 at 2:40 PM #137725 quote

GraHal
Participant

Master

I will later yes

06/30/2020 at 2:56 PM #137732 quote

zilliq
Participant

Master

88 % in 47 minutes

2020-06-30_15h55_44.jpg

06/30/2020 at 3:07 PM #137736 quote

Nicolas
Keymaster

Master

wow that code is not really optimized! Same indicator is called multiple times, better to use only 1 variable. You have also used many times the TIMEFRAME instruction to call the same timeframe, also a better idea to organize all the same timeframe codes below only 1 timeframe selection if possible.

Anyway i’m testing it that way and i’m at 5% completed in less than 3 minutes.

06/30/2020 at 3:35 PM #137752 quote

Nicolas
Keymaster

Master

31 minutes for 50%.

zilliq thanked this post

06/30/2020 at 4:00 PM #137761 quote

zilliq
Participant

Master

Thanks @Nicolas.

Ps: I know it’s not optimized, it’s not the purpose, it’s only to compare time of optimization. And we see MultiFrame slow clearly the optimization (It’s the same code but with MTF)

06/30/2020 at 4:06 PM #137762 quote

Nicolas
Keymaster

Master

zilliq wrote:

And we see MultiFrame slow clearly the optimization (It’s the same code but with MTF)

Obviously since the code is way more complex! Also, by using MTF you force X times more data to be loaded (OHLC history in X different timeframes).

zilliq thanked this post

06/30/2020 at 4:07 PM #137763 quote

GraHal
Participant

Master

Less than 3 minutes for 8,000 combos over 100,000 bars on GBP/USD 15 min TF on PRT v10.3.

06/30/2020 at 5:41 PM #137783 quote

zilliq
Participant

Master

With the MTF code ? because it seems very very fast @Grahal

06/30/2020 at 6:08 PM #137786 quote

Zigo
Participant

Master

Test

21% after 1′

48% after2′

75% after 3 ‘

100% after3’ 55″

Knipsel-test.png

06/30/2020 at 6:53 PM #137795 quote

zilliq
Participant

Master

Thanks @zigo

It’s with the MTF version ?, because as Grahal it’s very very fast

I see 100 unit on your picture ?

I’m at 98 % after 1h08 with the MTF version ….EUR/USD 1 mn 100 000 units 8000 combinations

2020-06-30_19h52_27.jpg

06/30/2020 at 6:55 PM #137797 quote

GraHal
Participant

Master

No I guess not … I tested the 1st .itf you posted.

I’ve just scrolled through and seen the 2nd .itf you posted … I will do that one later and report back.

06/30/2020 at 7:35 PM #137799 quote

Paul
Participant

Master

mtf version, demo 10.3, mac, 100k bars 1min tf, eurusd, 3m35s

06/30/2020 at 7:48 PM #137803 quote

Paul
Participant

Master

and full version 10.3, 100k bars 1m35s

Screenshot-2020-06-30-at-20.48.28.jpg
Author

Posts