(MCX Comedy..Errr. Commodity) - How to Fake your Trade

[vijkris]

U clicked on "statistics" to generate the loosers/winners. In that line only u will have "charts",trades",forumula etc. so click on trades, there each trade details will be there. this has nothing to do with afl.

@vijkris : Vijay: How to get the timestamp of the signal?

TimeNum() ?

 

[Tuna]

U clicked on "statistics" to generate the loosers/winners. In that line only u will have "charts",trades",forumula etc. so click on trades, there each trade details will be there. this has nothing to do with afl.

Verfied, looks Ok. so some 9 instances of the trades are screwed up. Puri 350+ line ki code pe sirf STI ki code copy kiya tha, usi kutte main bug hain. Marne do. it wont matter, I can re-verify the STI trigger from KITE once signal is generated.

 

[Tuna]

...deleted

 

[Tuna]

AFAIK, it means to give birth to a new arrow, it needs all past data as well 9 future data/quotes. each and everytime.

I may be wrong as well.. this i what my Einstenbuddhi understands...

You were right. It was 9 future quote for each candle. But the lucky escape for me (considering the amount of the work I have put here) - it has no bearing on buy and sells signal.

All of you know Ichimoku cloud, I have something similar to it. Now the leading span lead the price. AmiB was unable to paint that and was upset for that reason LoL.

Bar painting is fine, no issue we Buy and Sell signal. Verified same for 200 odd pairs from the backtest timestamp, all ok.

 

[Tuna]

@Happy_Singh @monkeybusiness @vijkris : Guys, do any of you perform forward walk to keep the system variables uptodate with changing market conditions? If yes, what is the cutoff period you generally use? I guess you understand what I am saying.

1. You built a system, perform backtest - looks good
2. With enough sample data in place, you perform optimization and fine tune the variables
--done
3. Now in actual deployement, it is advised to perform a walk forward (with the baseline variables -as derived on step 2), to apply the final touch up.

This is where the question is. So, what is the look back and forward market test we generally use. say (1 year minus last 3 months look back) + (3 months - till date) forward test to fine tune it and repeat this every quarter.

Is there any mathematical model for determining this ratio?

 

[manojborle]

http://www.marketcalls.in/amibroker/walk-forward-testing-explained-in-simple-terms.html

Tuna ji it seems its given here

 

[marimuthu13]

..................

 

[Tuna]

http://www.marketcalls.in/amibroker/walk-forward-testing-explained-in-simple-terms.html

Tuna ji it seems its given here

Read it, this is more about spliting each year into segment and walkforwarding, but what I am interested is the final walk forward segment, cause where we trade is the current market condition. (9+3 of 2013) will not be same as (9+3 of 2016). Anyways, I guess there can not be any straight answer to this.

But the kind of torture i am doing on my Mcbook, pata pada have not taken a single trade in this mech system and puri system burnout ho gaya. When left home it was still running. Phew.

 

[manojborle]

Tuna ji it seems other links also suggesting same (3:1 ratio)

http://www.automated-trading-system.com/walk-forward-testing/

 

[monkeybusiness]

@Happy_Singh @monkeybusiness @vijkris : Guys, do any of you perform forward walk to keep the system variables uptodate with changing market conditions? If yes, what is the cutoff period you generally use? I guess you understand what I am saying.

I am not good with auto testings, I test it on bar replay, writing each trade in excel with SL, and out come of trade.
1. that gives me much needed screen time
2. tunes my mind to draw downs & potentials of system
3. helps in identifying setups which can be used to hedge going forward in trade, for reducing SL(loss due to sl can be reduced), and maximizing winnings from winners.
4. Made me aware of break points where I will goof up trade.

etc etc.