Hello!
I wanted to ask about the flow of how others move from design to trading.
I've created a few systems, optimized to see what is good, and then found a few that also pass a walk forward test. How do you go about moving them to trading?
Initially I was getting a good result in my walk through tests, but the reporting didn't make sense. For example I had an optimize on hold days which should return an integer between one and ten, but I was getting values like 1088.47. Other values were similarly wrong. When I ran a new walk forward test to screenshot the error though I was unable to replicate it. If I get it again I'll document it, otherwise hopefully it does not reoccur.
Do people normally just take the final settings from the last Out Of Sample run in a walk forward and then trade the system using those values? Or would you run an optimize to find the settings to use for trading?
Now that I've been able to create a few systems and play with a few things I was curious to see about how others move forward into live trading.
Thanks!
I've been able to replicate the issue I'm showing.
movingaverage = Optimize( "moving average", 23, 1, 50, 1 );
indicatorone = Optimize( "indicator one", 17, 2, 30, 1 );
indicatortwo = Optimize( "indicator two", 21, 10, 30, 1 );
increment = Optimize( "increment", 19, 0, 50, 1 );
HoldDays = Optimize( "HoldDays", 1, 1, 10, 1 );
All of these parameters to optimize should be integers, mostly between 1 - 50. However, in the walk forward results they are not accurate.
Can anyone assist with what I might be doing wrong?
or should I be using a different method to solve for these parameters for trading? Should I be running an Optimize on the last In Sample period to determine parameter values instead?
It's hard to help you without knowing your code. Many things can go wrong, from one of your formula to a display format.
Optimization is always performed on in-sample, then parameters are tested on out-of-sample to check whether the system is performing as before. Optimizing out-of-sample doesn't make sense, except if you're doing continuous optimization, that is running an optimization on a rolling/moving 'window' (period of fixed length). But in that case it's not called 'out-of-sample'...
Thanks Alligator!
I had been doing some searching of the forums and believe it was a display issue. I pasted the output to excel and noticed that extra columns came through. I toggled on all column headings and it looks like the data is aligning correctly.
Perhaps I didn't explain it well, but now that I have the variable settings from the Walk Forward, when I move to live trading are those the settings I would use to start trading?
If I understand correctly the settings from the last In-Sample period of a Walk Forward would be the same as running an optimize on that time period and would be the same settings used to place actual trades. Is that correct?