There are situations where I am *not* that much interested in a realistic simulation but a reproducible instead.
For example in the earlier stages of strategy development it is crucial to check all the steps the algorithm does. During these runs I can't have changing results from run to run (due to randomly choosen NSF positions) because I can never tell if a change in code has the desired effect - and so on...
I'd suggest to add an (optional) field "Seed for Random Number Generator" (which is probably visible in expert mode only) which makes the NFS positions always the same.
For example in the earlier stages of strategy development it is crucial to check all the steps the algorithm does. During these runs I can't have changing results from run to run (due to randomly choosen NSF positions) because I can never tell if a change in code has the desired effect - and so on...
I'd suggest to add an (optional) field "Seed for Random Number Generator" (which is probably visible in expert mode only) which makes the NFS positions always the same.
Rename
Sounds like a good idea. And since it's for experts, maybe it would be best not to expose it as a preference but just as a new property in UserStrategyBase, like StartIndex.
You can always use smaller position size (or larger capital) to prevent NSF. Or if having NSF is the part of the strategy you can set weights withing your strategy.
I kind of agree with @Replikant_m. Why the need in feature creep if we can simply reduce the size and use Transaction Weight by the book?
Let's keep in mind that adding too many single-user options causes GUI bloat that could turn away new users from the product.
Let's keep in mind that adding too many single-user options causes GUI bloat that could turn away new users from the product.
QUOTE:
You can always use smaller position size
... but if the bug I am chasing depends on position size and probably NSF positions and so on, then the bug will go away with smaller position sizes.
... and if make position sizes larger again I have different results in each run...
You can already use a Transaction Weight to achieve the same result. I don’t want to bog WL7 down with even more minutiae. Let’s use the solution that’s already possible?
d'accord
Your Response
Post
Edit Post
Login is required