<- Back to Tasks

WalkForward Matrix is not working

Hi,

Because of the lack of documentation I've been testing by trial & error several optimizations and several strategies and there's still a HUGE MESS on the WalkForward Matrix.

First, in SQX, all the optimizations I've run are showing same results for the Strategy and for the WF Matrix strategy. See Screen1. Weird! But lets move on.

After the Optimization run I just focus on the ratio WF Net Profit Stability > 60%. That's the setup I use on SQ3. If I run the test on SQ3 I get a nice 88% - See Screen2: All good!
BUT when I take the same strategy and run the Optimization on SQX I get only 12%. See Screen3.
I must say that the parameters 'WF Net Profit Stability' in SQ3 and 'Net Profit (WalkForward Optimization, WF Stability)' in SQX are those who come by the Default WF conditions in both cases. I've followed your instructions.

I've used the same parameters in SQ3 and SQX. See Screen4, Screen5

I attach the Strategy I've been testing in both formats; SQ3 and SQX. The strategy was generated in SQ3 and retested and resaved/reloaded in SQX.
The period I've been working for all the tests and optimizations was 01/01/2013 to 31/12/2017

I don't understand how something that worked so smoothly in SQ3 is so confuse and wrecked in SQX
Attachments
Screen3.jpg
Screen3.jpg
(135.86 KiB)
Strategy 431.sqx
Strategy 431.sqx
(23.88 KiB)
Screen4.jpg
Screen4.jpg
(132.19 KiB)
Strategy 431.str
Strategy 431.str
(385.70 KiB)
Screen2.jpg
Screen2.jpg
(165.78 KiB)
Screen5.jpg
Screen5.jpg
(112.57 KiB)
Screen1.jpg
Screen1.jpg
(47.30 KiB)
Optimize.cfx
Optimize.cfx
(2.31 KiB)
  • Votes +16
  • Project StrategyQuant X
  • Type Bug
  • Status Fixed
  • Priority Normal

History

o
#1

Enric

01.12.2018 20:03

Task created

Dw
#2

Diwi

02.12.2018 07:00
Voted for this task.
DR
#3

mentaledge

02.12.2018 11:54
Voted for this task.
N
#4

nathan

02.12.2018 17:19
Voted for this task.
m
#5

mabi

03.12.2018 08:15
Voted for this task.
g
#6

geffect

03.12.2018 12:10
Voted for this task.
RL
#7

rickliao

03.12.2018 14:39
Voted for this task.
YJ
#8

extreme

04.12.2018 16:33
Voted for this task.
rg
#9

Robert Gergelyi

04.12.2018 16:50
Voted for this task.
a
#10

rnd_c3de51959aaa

09.12.2018 13:43
Voted for this task.
g
#11

Gecko

11.12.2018 06:51
Voted for this task.
gg
#12

geagea

13.12.2018 22:25
Voted for this task.
FS
#13

prometteusBR

18.12.2018 18:41
Voted for this task.
mp
#14

Michele

23.12.2018 14:22
Voted for this task.
b
#15

bentra

26.12.2018 17:48
IF bugs existed in SQ3 such as including in-sample data for some of the stats then we can expect some differences as those bugs get fixed in SQX and It should be correctly tougher to pass in SQX. Hopefully we can get the SQX and SQ3 results a little more comparable though so a new video does not have to be made.
o
#16

Enric

27.12.2018 20:59
Hey Bentra. Are you saying that SQ3 is who calculates bad WFM? Suposedly SQ3 was Ok
e
#17

eowithrand

28.12.2018 01:07
Voted for this task.
AC
#18

AC1962

28.12.2018 10:07
Voted for this task.
b
#19

bentra

28.12.2018 23:22
@Enric: I'm not sure but I think SQ3 might have been including in-sample data in some walk forward stats which is fundamentally incorrect. Here are some related bugs for SQX:
https://roadmap.strategyquant.com/tasks/sq4_3766
https://roadmap.strategyquant.com/tasks/sq4_3720

https://roadmap.strategyquant.com/tasks/sq4_3256



o
#20

Enric

30.12.2018 13:00
Voted for this task.
o
#21

Enric

30.12.2018 13:07
@Bentra. Not sure if SQ3, SQX, or both of them are bugged. Either way WFM is currently useless. I voted on the related bugs BTW
MF
#22

Mark Fric

02.01.2019 12:07

Status changed from New to Fixed

Votes: +16

Drop files to upload

or

choose files

Max size: 5MB

Not allowed: exe, msi, application, reg, php, js, htaccess, htpasswd, gitignore

...
Wait please