Skip Useless Information a good idea?

New comment
 

By checking the Skip Useless information option in Strategy Tester it hides negative results.

But is it a good idea to see the worst case scenarios Optimization as these results may happen if the Optimization is not set correctly in the EA?

 
It says a lot if not all about your system. By seeing whole picture of P|L ratio realted to DD and no of trades across wide spectrum of Ea input parameters will tell how much "luck" is there. What u looking for is nice Distribution curve, spikes are purly related to luck rather then strong backgroud of your strategy. There is however a theory how such random spikes could be turn into stady growth, But its another story.
 
  1. nicwaznego, your comment has nothing to do with the OPs question. He was talking about the optimization passes, not the equity curve of the final result. Once the optimization finishes THEN choosing which result and wither it's even useful comes in to play.
  2. It's blindly going through 10K passes to find the best combinations most of which will not be useful. Supposed it optimizing the SL and the best profit was 2.5 ATR and the worst was 0.1 ATR. Why would you care about the 0.1 or even want to see it.
  3. I run with skip unchecked so I can see the progress of the tester, am I at 2k/10k or 4k/10k The actual results I don't care. Suppress them if you don't care about its progress.
 
WHRoeder:
  1. nicwaznego, your comment has nothing to do with the OPs question. He was talking about the optimization passes, not the equity curve of the final result. Once the optimization finishes THEN choosing which result and wither it's even useful comes in to play.
  2. It's blindly going through 10K passes to find the best combinations most of which will not be useful. Supposed it optimizing the SL and the best profit was 2.5 ATR and the worst was 0.1 ATR. Why would you care about the 0.1 or even want to see it.
  3. I run with skip unchecked so I can see the progress of the tester, am I at 2k/10k or 4k/10k The actual results I don't care. Suppress them if you don't care about its progress.

i didn't mean equity curve, its less relevant /on some systems /slow but steady// can be adjusted by lot size. a curve of optimized parameter BUT /must/ in correlation with DD p/l and no trades. other words how smoothly its distribute of its peak point.

obviously testing beyond the point when "slop" becoming flat ish is most likely irrelevant so u right,

all matter = your strategy is worth as much as its weakest point. ///there is some modelling method which I am unable to explain due to my crappy English skill, which allow u to predict how well your strategy will cope /very short future/ if condition of what strategy was optimized on change dramaticly. for clarification = u can predict weather /1 day for almost sure, 2 days half of previous result, 3rd day half of second... thats why 4,5,6... become irrelevant, luck factor overtaking, thats why again in real world any weather prediction beyond third day in terms of science is just bu... sh...

Ignorance of basic principles is responsible for 100000s "amazing" systems with superior equity curve, hmmm what a pity it was working only in 2010 from jan to jun!!! by employing self adjusting factor u can constantly stay near main stream. and once thats why /at lest me/ i need to know them negative results to some extend. cos my curve peak parameters today might be shifted significantly on the left or right side of the axis. and if new curve there still remain quite wide and "smooth". its an indication of bullet proof strategy.

and since market /forex/ data is quite easy manageable by statistical means and rules its possible to create successful trading systems.

the only problem with this approach is time time time thats why am here by the way.... wanna employ EA to do what takes me quite a time on paper, i wont give up till i be able to code my 2 main strategy myself ... long way to go.


 
nicwaznego:


Good input - thanks.

The holy grail is: " is nice Distribution curve."

My EAs have a wide distribution of Optimization results; almost to the point of random.

I am concerned that if EA inputs are not correct for the market then I may get a surprise bad result as: "your strategy is worth as much as its weakest point" so I probably should be looking at raw Optimization results as well.

I see the point about the 0.1 ATR setting being outside the profit likelihood and I am looking for a way to improve confidence in the range of Optimization settings.

I am interested in a Self Adjusting EA but for now I manually eyeball the chart for the best timeframe fit and then optimize and update daily.

ZUP looks very interesting.

 
couta:


Good input - thanks.

The holy grail is: " is nice Distribution curve."

My EAs have a wide distribution of Optimization results; almost to the point of random.

I am concerned that if EA inputs are not correct for the market then I may get a surprise bad result as: "your strategy is worth as much as its weakest point" so I probably should be looking at raw Optimization results as well.

I see the point about the 0.1 ATR setting being outside the profit likelihood and I am looking for a way to improve confidence in the range of Optimization settings.

I am interested in a Self Adjusting EA but for now I manually eyeball the chart for the best timeframe fit and then optimize and update daily.

ZUP looks very interesting.

Are u mql_able??? if so, i can show u some code which utilize self adjustment, its not written in mql, logic is pretty straight forward, i do use that logic for trading manually, not forex however but stocks, it might not give good result on it own so what i do is import data /ohlc/ of my portfolio to excel, than it goes through script i wrote mirroring logic of that code and than once again through another script based on my very own "conclusions". its a very simple code, don't expect any thing too sophisticated, its probably or might be useless for Forex however it will give u a picture of what am talkin about. the amazing thing about it is that it survived market crush, not exactly, however self adjusting have very rare "idiosyncrasy" i would say. Amplify lets say average day to day market condition but smooth-en "crashes"... isnt all about trading.... otherwise u would be just better of with wide portfolio of buy and hold portfolio. What is silly since SHORT is allowed.

 
nicwaznego:

Are u mql_able??? if so, i can show u some code which utilize self adjustment, its not written in mql, logic is pretty straight forward, i do use that logic for trading manually, not forex however but stocks, it might not give good result on it own so what i do is import data /ohlc/ of my portfolio to excel, than it goes through script i wrote mirroring logic of that code and than once again through another script based on my very own "conclusions". its a very simple code, don't expect any thing too sophisticated, its probably or might be useless for Forex however it will give u a picture of what am talkin about. the amazing thing about it is that it survived market crush, not exactly, however self adjusting have very rare "idiosyncrasy" i would say. Amplify lets say average day to day market condition but smooth-en "crashes"... isnt all about trading.... otherwise u would be just better of with wide portfolio of buy and hold portfolio. What is silly since SHORT is allowed.


Thanks for the offer of the code although there was a self adjusting script on MQL4.com / forum which was on mql script.

My EAs are very simple as I want to avoid the idiosyncratic scenarios.

 
couta:


Thanks for the offer of the code although there was a self adjusting script on MQL4.com / forum which was on mql script.

My EAs are very simple as I want to avoid the idiosyncratic scenarios.

:)
 

step by step example.... because of my lack programming know-how it isn't going to be anything sophisticated however i am so immersed how much how quickly u can do using laptop and right piece of software....

any way i did manage to write substitute of what i use to trade manually, just basic principles cos as i have mentioned b4... lack of skill... so don't be to critical...

for simplicity, speed of test and practicality when trade manually method is based on so called ONCE_PER_BAR check however all test are done using every tick, 1M resolution from 2000.01.01 till 2day, lot fixed 0.1.

ok first what we do is a parameter "storm" with quite low resolution as what we want is an indication, than successful set-ups are narrowed...

after second filtration, result as expected above x -axis. plenty of solid blue is when line touch 0 / no trade due to "impossible condition"/ ...


than what u do is pick up result based on parameter u particularly like /low dd, equity curve, no of trades, whatever/

to avoid being to judgemental i took just best result in terms of equity from the top of curve.... run test... result.....

result isn't to good 14k - 10k deposit = 4k nett out of around 270 trades /11 years/!!!!!!!!! we gonna starve!

so i did combined another result from test, random best result 500 results away from 1st. result....

ok bit better : 20800 nett -10k deposit... we end up with 10.8k thx to incised no of trades around 810... way below 1k a year... still need to get up early morning....

just one more time to avoid having u bored... lets add 3 random best parameter /from test result/ same, random best 1000 from 1st one. result....

equity 26k - 10 deposit= 16k/11 years ... still to much to die to little to live...

its out of 1700 trades... so i think there might be something in my methodology, but whether its practical ... its hard to say.

one more conclusion... i getting absolutely mad knowing what i was missing... i couldn't even dream about analysing 1%of that data.... its time to brush up some old ideas

thx..

 

1 remark... i ve forgot to mention, its crucial to run optimization also for short and long positions separately... this example speaks for it self...

/breakout trading 1h, check on every new bar only, 2011- till 2day, 0.1lot/

quick optimization for long and short combined...

now lets split signals....

SHORT ONLY....

and Long only...

418 trades total but actually if there is someone who worked hard its gonna be SHORT

269 pos long - waste of space - 0.9k

149 pos short - 3.5k

that's why what i am always aiming at is totally independent control of short and long pos.

 

> ..totally independent control of short and long pos..

Scalping aside, its not unusual to have different settings for short & long - even if it is a smaller ATR take profit and larger ATR stop loss on longs compared to shorts

-BB-

12
New comment
Reason: