--- In [email protected], Mark Hike <markh...@...> wrote: > For example, where you put HandleStops() would affect how it behaves. > > I used a lot of _TRACE() in the code and figured out the things after many > experiments. >
Mark, Can the sorts of things you figured out be expressed as logical statements of fact/behavior to guide future efforts (yours and others), or do you wind up at: "After much experimentation, this particular code finally does what I intend, but heck if I really understand why! (or why half-a-dozen alternatives didn't)" -- A general comment: I can certainly imagine that TJ, with his knowledge of the internal data structures, and the specific processing done by routines he's written, can know directly when a block of code is properly formed to get a desired result from the CBT. OTOH, we in the field, with only the summary descriptions given by the docs, can easily do alot of beating around the bush to come up with a correct formulation. At a general level, I think that's fair and appropriate. I don't know of _any_ platform where the developer(s) offer a detailed white-paper tutorial on custom backtesting development. Indeed, AB excels in both CBT capabilities provided, and documentation of them. However, IMO, a clear and specific question such as Mike's, backed with a posted code example, could fairly be given a clear and specific answer, in acknowledgment of the user's effort to post and create a documentary record for the benefit of all in the community. It wouldn't matter if this answer was days or a week or two in coming (IOW, as time permits). One good answer from the source can be referred to forever-after when the same or similar question re-arises. And, in particular, if a user is trying to answer "How are the normal stats calculated (exactly), and how can I replicate them?" - that's a very reasonable question. A platform whose stats are provably correct is a superior platform. But a claim to correct stats that cannot be verified is just a claim ...
