Gamestudio Links
Zorro Links
Newest Posts
CSV to Tick import problem
by pcz
Today at 16:40
An Introduction to Algorithmic Trading: Basic to Advanced Strate
by yosoytrader
Today at 16:25
Square Root Rule & Additional Investment
by DdlV
Today at 15:15
What are you working on?
by painkiller
Today at 12:55
Support and Resistance Zones/Cluster
by sdh309795gaas
Today at 05:22
AUM Magazine
Latest Screens
Triton Wing
ZeaL
Computer Repair Simulator
Collider (working title)
Space Rider
Who's Online
15 registered (Denn15, painkiller, MatPed, Fiber, alibaba, firecrest, 3dgamelight, Iglarion, Matt_Aufderheide, pcz), 465 Guests and 5 Spiders online.
Key: Admin, Global Mod, Mod
Newest Members
Marce, pascalx, SpaceNL, onlinejack, johnnyp
17716 Registered Users
Topic Options
Rate This Topic
#464998 - 03/24/17 00:32 Question on TMF and WFO training
jyd Offline
Newbie

Registered: 03/09/17
Posts: 12
Hi,

Just wonder whether we need to include the TMF in WFO training or not. I have been testing using the TMF (a modified TrailingStopLong from the menu) in a modified workshop 5 implementation. After train without TMF enabled, I can get the following result:
Code:
Monte Carlo Analysis... Median AR 103%
Profit 3803$  MI 46$  DD 461$  Capital 643$
Trades 690  Win 65.9%  Avg +66.8p  Bars 86
AR 85%  PF 1.99  SR 1.19  UI 4%  R2 0.79



Then test again using optimized parameters and with TMF enabled:
Code:
Monte Carlo Analysis... Median AR 115%
Profit 4109$  MI 49$  DD 450$  Capital 636$
Trades 690  Win 67.2%  Avg +72.3p  Bars 92
AR 93%  PF 2.14  SR 1.20  UI 3%  R2 0.86



But if I trained with TMF enabled, then I get this:
Code:
Monte Carlo Analysis... Median AR 88%
Profit 3572$  MI 43$  DD 793$  Capital 822$
Trades 651  Win 62.4%  Avg +65.7p  Bars 95
AR 63%  PF 2.02  SR 0.97  UI 6%  R2 0.52



It seems when trained with TMF enabled, I get a worst result than without. So my question is whether I have done some wrong or we actually need to exclude the TMF during training, but enable it for testing and live trade?


Edited by jyd (03/24/17 00:35)

Top
#465019 - 03/27/17 11:21 Re: Question on TMF and WFO training [Re: jyd]
jcl Offline

Chief Engineer

Registered: 07/22/00
Posts: 25655
Loc: Frankfurt
If the TMF affects the trade, then you should include it in training, and accept the worse result as more realistic.

Top
#465027 - 03/28/17 08:51 Re: Question on TMF and WFO training [Re: jcl]
jyd Offline
Newbie

Registered: 03/09/17
Posts: 12
thanks jcl

Top
#465028 - 03/28/17 11:34 Re: Question on TMF and WFO training [Re: jyd]
Ger1 Offline
Newbie

Registered: 03/28/17
Posts: 16
Hi jcl

I was wondering exactly the same. Could you please clarify?

Also, in the manual it is recommended:

When your exit system is complex, optimize entry parameters with a simplified exit such as a simple stop. For using a simple exit while the entry parameters are optimized, evaluate the current parameter number (ParCycle), f.i. if(Train && ParCycle <= NumParameters-3) setExitSimple();.

Could you please advise on how and when to implement this recommendation from the manual?


Edited by Ger1 (03/28/17 11:35)

Top
#465077 - 03/29/17 15:31 Re: Question on TMF and WFO training [Re: Ger1]
jcl Offline

Chief Engineer

Registered: 07/22/00
Posts: 25655
Loc: Frankfurt
"complex" is indeed not a very clear formulation. Better is "when the return strongly depends on the exit system".

Entry parameters should not be strongly affected by your exit algorithm, since this is one of the causes of overfitting.

Top



Gamestudio download | chip programmers | Zorro platform | shop | resources | magazine | manual | support faq | bugs | beta features

oP group Germany GmbH | Birkenstr. 25-27 | 63549 Ronneburg / Germany | info (at) 3dgamestudio.net