You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current strategy dealing with effective fee removal for overlapping strategies works, but there are some issues that need to be address in the long run
The optimal solution depends on the size of the fee (or bid/offer; the two are the same): the smaller the fee, the worse the Optimizer is affected, and the less impact on the overall solution it has to remove it; the bigger the fee, the more benign it is for the Optimizer (eg, 10% fee is definitely not a problem) and the worse it gets when we remove it; so we need to work on the threshold between removing the fee or not
Where those adjustments are to be made is not clear; a priori that is a numerical issue that is better understood on the Optimizer / CPC side (and definitely not on the Node side); however, there are admin issues to consider as well, eg regarding CIDs
In the long run, we should probably provide a Carbon strategy constructor that encapsulates the current Carbon order constructor because then the CPC has the information it needs to understand whether or not to make adjustments (at the moment it just sees different orders and it does not connect them); we discussed the issues around CIDs of this approach with Nick, and possibly the move to a CID / Sub CID approach
Relatedly we noticed that the current CPC code creates curves that are optimized for the MargPOptimizer (notably, enforcing minimum curvature; the new code modifying strategies in place would fall into the same category); whilst this kinda makes sense -- we really only use MargPOptimizer -- this is not clean: the curves should be created as they are, and modified when reading into the Optimizer; however, this is a major change and it may not in practice be worth the effort.
Originally posted by @sklbancor in #594 (comment)
The text was updated successfully, but these errors were encountered: