Different results in live and demo

Jesper · 2019-03-21T12:04:55+00:00

Developed a scalping strategy in my live (IG) account that looked viable (see image). I then transfered the code to my demo environment to run simulation for a while. Problem was that backtestning didn’t even come close to what I got in live account (see image 2). Yes it’s the same settings, same instrument, timeframe and what have you. Find this rather distressing. Any ideas?

Viewing 8 posts - 1 through 8 (of 8 total)

Author

Posts
03/21/2019 at 1:04 PM #94274 quote

Jesper
Participant

Junior

Developed a scalping strategy in my live (IG) account that looked viable (see image).

I then transfered the code to my demo environment to run simulation for a while.

Problem was that backtestning didn’t even come close to what I got in live account (see image 2).

Yes it’s the same settings, same instrument, timeframe and what have you.

Find this rather distressing.

Any ideas?

2019-03-21.png 2019-03-21-2.png

03/21/2019 at 1:23 PM #94285 quote

Nicolas
Keymaster

Legend

The most recent answer to this question is here: Discrepancies between back testing on live and demo accounts on IG

03/21/2019 at 1:31 PM #94286 quote

GraHal
Participant

Legend

Jesper wrote:

and what have you.

Tick by tick selected and spread = same in both cases?

03/21/2019 at 1:58 PM #94295 quote

Jesper
Participant

Junior

GraHal > All settings are the same (tickmode, timezone) and so on.

Read the other thread and they were talking about a different engine on the demo-accounts. I’m not sure how that would make a difference since I’m not using multi-timeframe in my code.

03/21/2019 at 3:08 PM #94309 quote

Nicolas
Keymaster

Legend

It is not about only MTF support, the most recent backtest engine has also been almost completely revamped.

03/21/2019 at 3:20 PM #94315 quote

Jesper
Participant

Junior

Ok. That sort of makes sense. Well, I’m going to ask IG to switch back to the old engine and go from there.

03/21/2019 at 3:32 PM #94319 quote

Nicolas
Keymaster

Legend

BTW, in theory, the most recent engine is the most accurate. Only a precise orders analysis of the backtest should tell you.

03/22/2019 at 11:21 AM #94393 quote

mr blue
Participant

Senior

Interesting approach — develop on LIVE system and test on DEMO 😉

Just kiddin – please see my answer a couple of minutes ago in the other discussion linked above. Hope it helps!
Author

Posts