-
Add OptionStrategyMatcher (#4924)
frostbyte_neo 发布于
2020-12-02 21:42:24 +00:00 | 2693 次提交 在此版本后已推送到 master- Reformat/cleanup OptionStrategies
This file was breaking pretty much every style convention in LEAN.
There are other things that should be addressed in here that weren't,
such as passing non-argument names as argument names for ArgumentException,
as well as preferring constructors over property initializer syntax, but
such changes aren't being made to keep this commit strictly reformatting
instead of refactoring.Added braces and reformatted long lines to make code more legible.
- Add abstract base class for OptionStrategy Option/UnderlyingLegData
This allows us to create either or and later use the Invoke method to push it
into the appropriate list on OptionStrategy.- Replace O(n) option contract search with 2 O(1) TryGetValue calls
A better improvement would be resolving the correct symbol in the strategy, but
this immediate change is instead just focused on removing the O(n) search inside
a loop.- Add BinaryComparison and supporting methods in ExpressionBuilder
We're going to use these binary comparisons to make it possible to create
ad-hoc queries against a collection of symbols. Using these expressions,
along with type supporting composition of these expression, we'll be able
to define predicates that can declaratively define how to match an option
strategy with an algorithms current holdings.- Make GetValueOrDefault defaultValue optional
Was receiving ambiguous invocations leading to neading to invoke this
method explicitly (LinqExtensions.GetValueOrDefault) instead of being
able to use it as an extension method. Making the default value optional
seems to have resolved this ambiguity, leading to cleaner code in the
OptionPositionCollection (forthcoming)- Add OptionPosition and OptionPositionCollection
OptionPositionCollection aims to provide a single coherent interface
for querying an algorithm's option contract positions and the underlying
equity's position in a performant, immutable way. The immutability of
the type is necessary for how the options matcher will operate. We need
to recursively evaluate potential matches, each step down the stack removing
positions from the collection consumed by each leg matched. This will enable
parallelism of the solution as well as simplifying the mental model for
understanding due to not needing to track mutations to the collection
instance.-
Add Option test class for easily creating option symbol objects
-
Add OptionStrategyLegPredicate and OptionStrategyLegDefinition
The definition is a composition of predicates, and each predicate supports
matching against a set of pre-existing legs and a current position being
checked for the next leg (this leg). In addition to the matching functionality,
it also supports filtering the OptionPositionCollection, which is where much
of the work for resolving potential option strategies is done. By successively
filtering the OptionPositionCollection through successive application of predicates,
we wil end up with a small set of remaining positions that can be individually
evaluated for best margin impacts.All of this effectively unrolls into a giant evaluation tree. Because of this
inherent structure, common in combinatorial optimization, the OptionPositionCollection
is an immutable type to support concurrent evaluations of different branches of
the tree. For large position collections this will dramatically improve strategy
resolution times. Finally, the interface between the predicate and the positions
collection is purposefully thin and provides a target for future optimizations.- Add OptionStrategyDefinition and OptionStrategyDefinitions pre-defined definitions
The OptionStrategyDefinition is a definitional object provided a template and functions
used to match algorithm holdings (via OptionPositionCollection) to this definition. The
definition defines a particular way in which option positions can be combined in order to
achieve a more favorable margin requirement, thereby allowing the algorithm to hold more
positions than otherwise possible. This ties into the existing OptionStrategy classes and
the end result of the matching process will be OptionStrategy instances definiing all
strategies matched according to the provided definitions.- Add OptionStrategyMatcher and Options class, w/ supporting types
OptionStrategyMatcherOptions aims to provide some knobs and dials to control how
the matcher behaves, and more importantly, which positions get prioritized when
matching. Prioritization is controlled via two different enumerators, one controller
which definitions are matched first and the other controller which positions are
matched first. Still unimplemented, is computing multiple solutions and running the
provided objective function to determine the best match. When this gets implemented,
we'll also want to implement the timer. For anyone looking to implement these features,
please talk with Michael Handschuh as there's a particular way of representing these
types of combinatorial solutions (a 3D tree) that can be used as a variation of the
linear simplex method for optimizing combinatorial problems.-
OptionStrategyMatcher: Address PR review comments
-
Ensure created OptionStrategy legs all have the same multiplier
Each leg definition match gets it's own multiplier which indicates the
maximum number of times we matched that particular leg. When we finish
matching all legs, we pick the smallest multiplier from all the legs in
the definition and use that as the definition's multiplier. When we go
to create the OptionStrategy object we MUST make sure we're using the
multiplier from the definition and not from the individual legs.This change fixes this issue and also provides a guard clause to ensure
that we're not trying to use a multiplier larger than what was matched.- Add XML docs for OptionStrategyDefinitions from OptionStrategies
下载附件