Randomize parameter, Optimization report and System Parameter perm. —- why?

Hello!

I am really wondering if it is worth to waste time with the

– randomize parameters
– optimization report and
– system parameter permutation

I made an interesting test.

1. I generated a lot of strategies on EURUSD. IS = 2004-2008 + OOS = 2008-2017
2. I made the other market test (GBPUSD + USDCHF)
3. I made the other TF test (M30 + H4)
4. I made the mixing trades test
5. I made the complex test (slippage, spread and history data at the same time)

6. After that I did it the other way around: before I went on with the remaining tests, I tested first, which strategies were profitable on the 2017-2019? It were 29.

7. After these I just chose 4 strategies from them. I wanted to know,  how many of these profitable strategies would have been “strong”. I made next:

– randomize parameters
– optimization report and
– system Parameter permutation

….. and no one could have got an “OK” for live trading, however they had a good equity curve.

QUESTION:  should we throw away strategies just because this? I am very over-questioned….

Brg: Csaba

I think they are not profitable on demo/real account.

You can test this and share a myfxbook account

how long was your live test?

I did not make live test.
I meant, that based on those 3 tests the strategies would have not been OK for live trading, however the OOS test showed, that the strategies were profitable…

What do you all think here? What are the most important robustness tests?

ok then how long was your confirmation/true os test csaba?

ah then you ate using blocks/settingswhich are doing your live trading not profitable.  post here your config

Here you find my config.
And how do you mean “how long was your confirmation/true os test“? — do you mean how long was IS and OOS?

IS = 4 years (2004-2008)
OOS= 10 years (2008 – 2018)

your building config is strange, why are you using UTC0 data? why you are using only a few building blocks? …and this is only 2 major questions

you will not get diversified strategies from this i am afraid

Hello Hankeys

🙂 I knew, that somebody will ask this. Okay, let answer:

1. UTC >>> I use on my computer UTC + 2 data, I converted the UTC+0, it is sure

2. Why I use just a very simple set of building blocks? >>> I think, if there is a really good breakout, it is not important if the stop order was placed because the CCI was rising or not, or some indicator crossed the other. I think, the Highest, Lowest, High, Low and Close values are good choice, because there we do not speak about indicator periods etc. Just keep it simple.

For me is more important the exit if the price moves. How it will be pulled in BE or how the trailing stop works? So how to manage the position if I have already an open position. And if the stop order was placed based on Highest or Moving Average or ADX it does not matter (for me). So why not chose the more more simply way?

3. And to the diversified portfolio >>> I think the requirements of a diversified portfolio are, that we have more instruments and more timeframes. Of course I don’t want to reach a diversified portfolio with 1 instrument and 1 timeframe.
Because we measure the diversification on MONTHLY base and based on PROFIT/LOSS, the most important thing is, that the portfolio strategies of various instruments and timeframes should complement each other. Or not? 🙂

Thank you and have a nice day!

Brg, Csaba