@Nicolas, do you know if the walk forward optimization operates differently from the standard optimization engine? what I mean is, does it return the values that happen to appear at the top of the list before TBT?
Today I did a standard optimization and had to manually click through the results to get a print of each one. The best was 8th from the top. I only knew that the ones above that were less good because I had done a straight backtest with an estimated value which turned out to be right. But what if I hadn’t known that? (as we usually dont) or what about the WF result where we don’t even get to see the second best, third best etc. in order to check?
As it is, the optimization engine is effectively useless – essential to manually check each combination.
Hi all,
I find it quite confusing that we have a topic “V11 bugs” with indeed all sort of bugs announced. It is too hard to track and contribute (and possibly solve).
Am I alone in this ?
Regards,
Peter
No you are not alone … it would be better to have a separate Topic for each Bug?
Also the Topics need to be in the Platform Support Forum?
essential to manually check each combination.
By above, do you mean … click on each row’s values and then check the equity curve and detailed report?
click on each row’s values and then check the equity curve and detailed report?
yes, that is what I found myself doing. the performance bore no relation to the ranking in the optimization window – best by far (like 20% better) was 8th in the list (by Gain). So the optimization is ‘helpful’ in that it gives you an idea of what the best values might be, just don’t assume that the top is the one to go for – check a few more further down the list!
I wasn’t sure where to post this. I guess it should be in Platform Support but I didn’t think that anyone would see it there.
I had the attached yesterday when playing with a strategy that accumulated positions and did partial closure. A 19.5K loss was shown as a 39.8K profit on the equity curve. I just gave up and shut the platform down before banging my head up and down on the laptop repeatedly.
Are you guys changing Timeframe before you see what you mention above, by any chance??
In v10.3 if we changed Timeframe then the Optimiser went straight into a new Optmise run.
With v11 we now have 3 clicks to do …
1. Change Timeframe
2. Click Algo Title
3 Run Optimiser
No time frame changes. Just developed the strategy on the daily time frame and when I added accumulating positions and partial closure it all went screwy and the optimisations became worthless!
A 19.5K loss was shown as a 39.8K profit on the equity curve
I thought mine was bad, but that is totally mental!
A 19.5K loss was shown as a 39.8K profit on the equity curve.
But you are sorting on that dn column ? (see mouse arrow in my attachment).
Plus I recall that what is selected formally can be set somewhere (apologies I don’t know where that sits from the top of my head). Thus I assume that the Equity Curve shows that selection and the selection is out of view.
?
Yes it is sorted on ‘dn’ but that is not relevent. There are only 8 optimisations results with dn from -1 to 6 (it is a SEB robustness test). The top result is selected and that is the equity curve for it.
Another bug : sometimes when you backtest a large historic of data (1M for example / 2010-2020), PRT runs the backtest on 2017-2020 only. Then I launch the backtest again and it runs on 2010-2020. Not a “big” problem but a loss of time.
Face the same too. It seems like a regression in v11, never have this before.
Would it be possible to create a seperate snippet link library for recognised V11 bugs as I also am struggling to sift through previously posted bugs on forum to avoid duplicating previous posts?
Just thinking in the long run it might save time on tracking and resolving them if we have a single resource to go to.
I’m happy to help going through forum and reposting them into a library if that would cut down on moderators workload.
I also am struggling to sift through previously posted bugs on forum to avoid duplicating previous posts?
My view is that unless we post direct to PRT using the Forms below then there is little chance that any corrective / improvement action (by PRT) will result.
Even if we use the Forms below, then we have no visibility after submission or tracking / ticket number etc, but at least we have brought to the attention of the Developers of the Platform.
I say above because I think some members on here think this website is part of / belongs to PRT / IT-Finance ( The Software House / PRT Developers)? As such, all they need do is to post about bugs and improvements on this website and then they will get sorted?
Issues may get sorted, if Nicolas (this website owner) presents and progresses Issues through PRT direct.
Nicolas is a very busy chap and he can’t monitor every post on this website, so it be good for us to (also) use the Issues Form below.
Improvements need to be sent direct to PRT using the Suggestions Form below.
I am happy to host a spreadsheet on my google drive as I also host the Snippet Link Library and a few others.
I will knock-up a PRT Issues Log and PRT Improvements Log today and post links on this Topic.
PRT Issues Form
https://www.prorealtime.com/en/contact
PRT Suggestions for Improvement Form
https://www.prorealtime.com/en/contact?suggestion=1
