-
1. Optimizing across all instruments in a chart should try to even out the result for robustness. IOW, if 1 out of 10 issues have a 1000% bull run, the optimization should not try to maximize that one on the expense of 9 other losers just because net profit is greater.
2. Optimizing across all instruments in a chart should show and optimize based on the value of the chosen fitness level for the total portfolio.
3. Optimizing across all instruments in a chart should have option to use Median instead of Average for fitness.
4. Position sizing: If I have 10 issues in a chart, it should be an option where I could say the total account is for example $100k and each position should be trading with 10% of the total account. This seems to work for Kelly, Optimal f and Secure F, but not for say Fixed Percentage.
5. There should be an option to skip the indicator submenus and be taken straight to all indicators to save a step in the workflow, if that is how you prefer to work.
6. On my dual-CPU Xeon with 44 cores/88 threads, the initialization takes and sometimes computation takes longer than on my old two-core laptop! There must be a more efficient way of dealing with distributed job allocations and related overheads.
7. An option to define slippage in basis points. Points is too crude for chart with multiple issues with an order of magnitude difference in Price.
8. Ability to have an indicator’s constant be another indicator, i.e. a variable/adaptive instead.
9. An easier way to do recursive indicators.
10. More than 30,000 tickers.
11. 3D plot over optimization space.
12. Log Y scale doesn’t always work.
13. Ability to optimize forecast horizon also in normal Prediction Wizard (and not only in add-ons like AT2).
14. Ability to forecast training set also in normal Prediction Wizard.
15. Ability to train on newer data and back test/paper trade on older data (non-chronological) since make mor sense to train on latest for going forward than the other way around.
16. Ability to not have to use consecutive data or ability to exclude certain data/dates from training and/or back-testing (e.g. the COVID-19 dip) which “distorts” the typical patterns.
17. Ability to stop optimizing ALL remaining chart pages without losing the results so far (as you inevitably do when you abort). As it is now, if I have a chart with 100 tickers but want to stop optimizing after 30, I either need to manually hit “Stop Optimizing” 70 times (with several minutes in between each as NST prepares to optimize the next ticker). A “Stop Optimizing Remaining” button would solve this.
18. Ability to vary position size as a function of signal “strength”. For example, if I am predicting the %Change in Price, I may want to go in with say twice the amount if the predicted value is twice as high (e.g. 2% instead of 1%). Moreover, I do not own Adaptive Nets but I have read the manual online and if I were to buy them, I’d like the positioning to depend on the strength of the probability being output. Perhaps a solution could be to have an option under position size where a fixed amount is multiplied by the output of a prediction or indicator (e.g. $10,000 x 0.87).
19. Similarly, I don’t yet own the AT2 add-on but from reading the manual, it would be nice to be able to toggle a few different objective functions, including traditional statistical error and correlation ones. Or I may want to maximize the percentage the sign right to be used for a specific function in a greater trading strategy.
20. When you make your own indicators, it would be good to be able to lock certain values (constants) and define min and max permissible values. Often I copy or port an obscure indicator from Tradingview or TASC and (as usual since they never work) put it aside and forget it. But when I pick it up again and want to insert it into a prediction or trading strategy, it is almost impossible to remember what all the parameters are and how to set them for the optimizer (static, logical ranges etc). Either that or some way to click on it and have my own notes next to each parameter.
21. Ability to test co-integration (as opposed to correlation) between issues (“Other instrument data”) for pairs trading, preferably also the Dickey Fuller test in one handy indicator (e.g. https://blog.quantinsti.com/augmented-dickey-fuller-adf-test-for-a-pairs-trading-strategy/).
22. A way of quickly locking all the parameters of an indicator to their default values in the optimizer instead of having to expand them one-by-one and manually having to make up to 100 clicks just to padlock them individually.
You must be logged in to reply to this topic.