B131 Major result discrepancies when comparing M1 Mac build to Intel Mac Retest

With the two licences I have I set up my Intel Mac and M1 Mac with the exact same build, template and data.

However when I transfer strategies from the M1 Mac to the Intel Mac (full license) I noticed that when I carry out an additional 2 years OOS test the results from the original build period are then calculated differently.

From the screenshot the two columns for the profit from the original build and IS in the retest should be identical as they are exactly the same date range. They are identical for only one of the strategies.

Original build date range: 1 Jan 2010 to 31 Dec 2017 (Mac M1)
Retest: 1 Jan 2010 to 31 Dec 2019 (In sample 1 Jan 2010 to 31 Dec 2017) (Intel mac)

The data used was Dukas H1 GBPUSD cloned to London UTC with DST.

Attachments

Ichimoku Template

(4.91 KiB)

Screenshot 2021-05-07 at 152208.png

(700.80 KiB)

Ichimoku Build

(51.23 KiB)

Strategy 042.sqx

(41.45 KiB)

Strategy 0121.sqx

(61.64 KiB)

Strategy 0265.sqx

(57.78 KiB)

Strategy 079.sqx

(59.76 KiB)

Strategy 0231.sqx

(42.65 KiB)

Strategy 0201.sqx

(50.47 KiB)

Strategy 0239.sqx

(40.99 KiB)

Votes +1
Project StrategyQuant X
Type Bug
Status Fixed
Priority Normal

History

Jabezz

07.05.2021 10:33

Task created

Tomas Brynda

26.05.2021 09:27

Status changed from New to Fixed

Hi Jabezz,

there should be no difference if you add an OOS starting after your original backtest data range.

I have tested it and get matching results.

Make sure you have up to date data in both your SQX installations and when setting up the OOS part, set the start date one day later than your original test.

For example:

original backtest without OOS: 2010.01.01 - 2017.12.31

backtest with OOS: IS = 2010.01.01 - 2017.12.31, OOS = 2018.01.01 - 2018.12.31

Best regards,

Tomas

Chris G

27.05.2021 00:06

Voted for this task.

Votes: +1