When developing strategies, I often have many parameters (e.g., switches for calculation methods, filters, etc.) for testing. I would like to request a feature that allows multi-selection of parameters for block enabling/disabling. It would be nice time saver.
Rename
Can you develop the idea more? It's not clear to me how "block enabling/disabling" is any different than what's already available to select optimization parameters, i.e., the checkboxes next to the Parameter(s) in the Standard view of Optimization Settings. One way or another, you need to define the "block".
Selecting more than one row at a time then enabling or disabling the checkboxes in one shot.
How much time do you think it would save?
I think it would actually add more time. Instead of clicking each row to select, and then click AGAIN to enable/disable, right now to can enable/disable with one click. I'm failing to see how the proposed change would benefit the workflow in any way, but I might be missing something?
Enabling/disbling many parameters with one click, or menu action, would be much quicker. I have over 50 parameters but only optimize less than 5 typically. Currently, I need to disable the majority (e.g. 50 clicks .vs. a few clicks)
I suppose with the extended multi-select (shift click to select many items) it would save time in that case.
Wow, 50 parameters sounds like a lot of degrees of freedom.
QUOTE:
Enabling/disbling many parameters with one click, or menu action, would be much quicker. I have over 50 parameters buy only optimize less than 5 typically. Currently, I need to disable the majority (e.g. 50 clicks .vs. a few clicks)
@bengh80503
There are not so many traders who are not afraid of parameters). Definitely, you are one of them. So am I. Would be great if you share some experience about dealing with parameters as people-who-are-afraid-of-parameters's experience usually is not interesting. I can share mine.
QUOTE:
Eugene: Wow, 50 parameters sounds like a lot of degrees of freedom.
QUOTE:
Replikant_m: There are not so many traders who are not afraid of parameters). Definitely, you are one of them. So am I. Would be great if you share some experience about dealing with parameters as people-who-are-afraid-of-parameters's experience usually is not interesting.
As someone with an educational background in Applied Mathematics, Statistics and Computer Science, I would have to agree that is many degrees of freedom.
Here is some background info to chew upon...
I am assembling a multi-setup system with quite a few trading setups, each using various adaptive indicators, and multiple exit methods. Each setup has parameters for enabling/disabling the setup, buy/sell thresholds, as well as associated indicator parameters. Since the indicators adapt to market behavior (e.g., dominant cycle, trend mode, volatility, etc.), they have parameters dealing with calculation methods and sources of data upon which to calculate the indicator. And so forth. I typically enable one setup and exit at a time for testing. Once robust parameter sets have been achieved, I run some optimizations that find a nice combination of systems. Eventually, setup selection will become automated.
I know it sounds ambitious, but I am a retired software engineer by trade. I have also developed a trading platform from scratch in C#. I have also worked many trading systems using other platforms. I also find it to be an enjoyable activity for me during retirement.
I am looking forward to participating in the evolution and growth of WL7!
Be sure to vote for your own Feature Request by viewing the WishList!
https://www.wealth-lab.com/wishlist
https://www.wealth-lab.com/wishlist
@bengh80503
Thanks for sharing the idea behind this.
I use many parameters per strategy to increase the amount of data and... variability. Backtesting and interpreting results is a whole research for me, I'm confident in an idea of the way of interpreting backtesting results beeing a real reason for overfitting, but not the amount of parameters.
Thanks for sharing the idea behind this.
I use many parameters per strategy to increase the amount of data and... variability. Backtesting and interpreting results is a whole research for me, I'm confident in an idea of the way of interpreting backtesting results beeing a real reason for overfitting, but not the amount of parameters.
Optimization is a tricky thing. I wish people wold convert to my preferred term: "Parameter Stability Test" (PST).
Also I would suggest to use the great Walk Forward Optimization feature within WL7. Not (only) for finding the optimum parameter, but to get more realistic historical results. Check the right mouse click on the WFO result page.
Also I would suggest to use the great Walk Forward Optimization feature within WL7. Not (only) for finding the optimum parameter, but to get more realistic historical results. Check the right mouse click on the WFO result page.
QUOTE:
Optimization is a tricky thing. I wish people wold convert to my preferred term: "Parameter Stability Test" (PST).
Also I would suggest to use the great Walk Forward Optimization feature within WL7. Not (only) for finding the optimum parameter, but to get more realistic historical results. Check the right mouse click on the WFO result page.
Could you please unveil some more details). Your words sound interesting but I can't get the idea - maybe because of the lack of details, maybe because of my English)).
Well, if you optimize the period from the first day of the launch of a stock until today you will very likely get good results. Problem is, that 5, 10 or twelve years ago you did not know the best value - hence the backtesting results are "biased".
If you use the great WFO features of WL7 and use expanding window, you basically backtest and optimize as if you started X years ago.
You get a much better real backtest result, more realistic.
If you use the great WFO features of WL7 and use expanding window, you basically backtest and optimize as if you started X years ago.
You get a much better real backtest result, more realistic.
QUOTE:
#6 I suppose with the extended multi-select (shift click to select many items) it would save time in that case.
Alternate #1: Place a checkbox next to the column heading to enable/disable the whole list. Currently all list entries are enabled. For those who have large parameter sets, this would help since they could disable the whole list then enable just those that they are testing. For those with sma;; parameter sets, there would be little or no impact.
Alternate #2: Disable all initially. Currently all list entries are enabled. For those who have large parameter sets, this would help since they could then enable just those that they are testing. For those with small parameter sets, it would add a few more clicks.
I don't have 50 parameters, but I do have more than my computer can handle in a reasonable time. I too would find the multi-select feature to be handy.
I'll throw my two cents into how I would implement this.
1) Selecting parameter rows should be handled similar to File Explorer file selection so we're using familiar key shortcuts.
2) Clicking the space bar should toggle the check/uncheck of the check box.
For similar behavior to File Explore (details view format), these are some thoughts:
* CNTRL-A should select all rows.
* SHIFT key would select a range of parameter rows. Click on one parameter to select only the single row. SHIFT-Click should select this row and all those between the previous clicked. For example, click row 2 then SHIFT-Click row 4 should cause 2, 3, and 4 rows to be selected.
* CNTRL-CLICK should select/deselect a single row without deselecting any others. For example, click row 2 then CNTRL-Click row 4 should cause 2 and 4 rows to be selected. If a row is already selected, it should deselect it. For example, if you hit CNTRL-A and then CNTRL-Click on row 2 then all rows will be selected except for row 2.
The SHIFT-Click and CNTRL-Click seem to already work this way to select/deselect, but CNTRL-A would be a new feature.
Using the space bar to check and uncheck should toggle the check/uncheck, but all boxes should be in the same state. You can use the first row selected to determine the start state. If the first selected row is checked and you hit the space bar, all the rows selected should be unchecked. Hit the space bar again and all selected rows should be checked.
Hopefully this is a reasonable summary.
I'll throw my two cents into how I would implement this.
1) Selecting parameter rows should be handled similar to File Explorer file selection so we're using familiar key shortcuts.
2) Clicking the space bar should toggle the check/uncheck of the check box.
For similar behavior to File Explore (details view format), these are some thoughts:
* CNTRL-A should select all rows.
* SHIFT key would select a range of parameter rows. Click on one parameter to select only the single row. SHIFT-Click should select this row and all those between the previous clicked. For example, click row 2 then SHIFT-Click row 4 should cause 2, 3, and 4 rows to be selected.
* CNTRL-CLICK should select/deselect a single row without deselecting any others. For example, click row 2 then CNTRL-Click row 4 should cause 2 and 4 rows to be selected. If a row is already selected, it should deselect it. For example, if you hit CNTRL-A and then CNTRL-Click on row 2 then all rows will be selected except for row 2.
The SHIFT-Click and CNTRL-Click seem to already work this way to select/deselect, but CNTRL-A would be a new feature.
Using the space bar to check and uncheck should toggle the check/uncheck, but all boxes should be in the same state. You can use the first row selected to determine the start state. If the first selected row is checked and you hit the space bar, all the rows selected should be unchecked. Hit the space bar again and all selected rows should be checked.
Hopefully this is a reasonable summary.
QUOTE:
When developing strategies, I often have many parameters (e.g., switches for calculation methods, filters, etc.) for testing.
You may be interested in my idea (which was declined unfortunately) if you want to use the strategy parameter as a switch but not used for optimization.
https://www.wealth-lab.com/Discussion/Separate-Strategy-Parameters-and-Optimization-Parameters-7238
Your Response
Post
Edit Post
Login is required