* Improve information tracked in regression's {algorithm}.{lang}.details.log
The details.log file aims at providing a diff-able document that quickly and
easily provides actionable information. Since many regression algorithms use
the algorithm's debug/error messaging facilities to log various pieces of algo
state. This document also support a configuration option: regression-high-fidelity-logging'
that logs EVERY piece of data, again, with the aim of providing an easily diff-able
documenbt to quickly highlight actionable information. I may have missed omse key
pieces of information here, but now that the entire QC knows about this regression
tool, if additional information is required then hopefully it's easy enough at this
point to extend the RegressionResultHandler to suit our needs.
The RegressionResultHandler was initially implemented to provide a concise log of
all orders. This was achieved by simply using the Order.ToString method. While
testing/investigating OptionExerciseOrder behavior, it became evident that more
information was required to properly identify the source of potential failures or
differences between previous regression test runs. This change adds logging for
almost every IResultHandler method and additionally attempts to capture the
actual portfolio impact of every OrderEvent. This is accomplished by logging
the portfolio's TotalPortfolioValue, Cash properties and the security's
SecurityHolding.Quantity property.
This change also standardizes the timestamps used to folloow the ISO-8601 format.
When using the RegressionResultHandler, it is highly recommeded to also disable
'forward-console-message' configuration option to ensure algorithm Debug/Error
message logging is done synchronously to ensure correct ordering with respect to
log messages via Log.Debug/Trace/Error.
* Fix typo in options OrderTests test case name
* Update SymbolRepresentation.GenerationOptionTickerOSI to extension method
Far more convenient as an extension method
* Improve R# default code formatting rules
Many of these rule changes focus on improving the readability of code,
with a particular emphasis on multi-line constructs, chained method calls
and multi-line method invocations/declarations.
* Add braces, use string interpolation and limit long lines
* Refactor OptionExerciseOrder.Quantity to indicate change in #contracts
For all other order types, the Order.Quantity indicates the change in the algorithm's
holdings upon order execution for the order's symbol. For OptionExerciseOrder, this
convention was broken. It appears as though only exercise was initially implemented,
in which case only long positions were supported and a code comment indicated that
only positive values of quantity were acceptable, indicating the number of contracts
to exercise. At a later date, assignment simulation was added and utilized a negative
order quantity. This caused some major inconsistencies in how models view exercise
orders compared to all other order types. This change brings OptionExerciseOrder.Quantity
into alignment with the other order types by making it represent the change in holdings
quantity upon order execution.
This change was originally going to be much larger, but in order to minimize risks and to
make for an easier review experience, the additional changes will be committed separately
and pushed in their own PR. Some of the issues identified include:
* Manual Exercise (especially for OTM) is not covered
* Margin Calculations (in particular taking into account opposing contracts held)
* IBrokerage.OptionPositionAssigned is raised for exercise (later filtered by tx handler)
Fixes OptionPortfolioModelTests to use exercise model to properly model exercise of
non-account quote currency option contract.
* Add internal subscription manager
- Add InternalSubscriptionManager that will handle internal
Subscription. Replaces the realtime updates
- Fix thread race condition in the TimeTriggeredUniverseSubscription, we
have one thread injecting data points, the main algorithm thread, and
the base exchange is pulling from it
- Fixes for FakeDataQueue
- Adding unit tests
* Address reviews and fixes
- Internal subscription will use extended market hours
- Only sample charts accordingly
- Get api-url once
- Refactor ZipDataCacheProvider to avoid value types thread issues, it
will now use a timer. Add missing dispose calls.
- Synchronizer won't share the `SubscriptionFrontierTimeProvider`
instance since it's not thread safe and shouldn't be called
- Fix BacktestingResultHandler `_daysProcessed` thread safety
- Fixing `RealTimeScheduleEventService` thread safety
- Reduce ResultHandler amount of packets sent and their size.
- Only send order and orderEvents packet if there is actually data in them.
- Don't send Holdings fields if value is 0
- Don't send AlphaRuntimeStatistics fields if default values
- Algorithms logs of the same type will be batched together when sent
- Increasing backtesting update to 3 seconds same as LiveTrading
- Increasing insight sending interval from 1 second to 3
- Reducing code duplication for retrieving and sending algorithm logs
- Disable order event streaming in backtest
- Add new OrderEvent IsUpdate flag to be set by the different brokerage
implementations
- Update regression test stats after rebase
- Live and backtesting will send delta order events updates
- Live will store order events every 10 minutes per day
- Backtesting will store last 100 order events on every update and will
store all order events in the end of the backtest
- Update logic in `Run` method will be executed by an internal Task
handled by the `BaseResultHandler`
- Call `DataFeed.Exit` even if algorithm initialization failed
* Changed variable names of protected members in BaseResultHandler to
match existing variable naming convention
* Changed AlgorithmRunner return type
* Remove AlgorithmResults dictionary from AlgorithmRunner
* Create AlgorithmRunnerResults container class
* Modify Relative Sampling test to accept failure cases
* Misc. updates as a result of changing AlgorithmRunner return type
* Get rid of `previousTime` and use `time` instead in AlgorithmManager
* Refactor variable names in Backtesting and Live IResultHandler impls
* Moves shared variables to BaseResultHandler
* Modifies BacktestNodePacketTests statistics to get tests passing
* Adds new StatisticsBuilder tests
* Modifies BacktestingResultHandler tests to make them passing
- Regarding these tests, the decision was made to get them
passing so that if any behavior changes, we will know immediately.
Next commit will contain regression test changes for easy rollback.
* Removed Sample[a-zA-Z]+ methods from IResultHandler definition
* Converted Sample[a-zA-Z]+ methods from public to protected
* Updated inheritors of BaseResultHandler to use new accessibility
modifiers
* Removes useless code in ResolutionSwitchingAlgorithm
* Refactors AlgorithmManager loop
* Refactors StatisticsBuilder methods and strategy for series alignment
* Move sampling logic to the corresponding IResultHandler
* Changes benchmark resolution to Resolution.Hour
* Modifies IResultHandler to enable external sampling
* Adds BacktestResultHandler unit tests
* Adds ResolutionSwitchingAlgorithm to test misalignment
* Adds support to AlgorithmRunner to store algorithm IResultHandler
Warning: this commit breaks accurate calculations for algorithms that
only make use of `Daily` resolution data. Previously, because
the benchmark was added in Daily resolution in backtesting, any
algorithm that only made use of daily data would have an accurate
calculation for beta and various other statistics.
These changes serve to fix the statistics calculations of non-daily
resolution algorithms, with daily resolution to be revisited at a later
time.
- BacktestingResultHandler will send a maximum of 50 orders per update
packet and will check `LastFillTime` and `LastUpdateTime` too
- Fix invalid linked file
`AccumulativeInsightPortfolioRegressionAlgorithm`
- Reduce code duplication between live and backtesting
- `LiveTradingResultHandler` will store logs progressively, appending to
the file
- `BacktestingResultHandler` stores logs at end
- Remove chart subscription logic. Will stream all chart updates if any
(wont stream empty updates)
- Only serialize properties which are not null
- Adding Chart and Series `IsEmtpy()` extension. Adding unit tests
- `SubscriptionDataReader` will check map file first data and adjust
start date based on it
- Adding unit test
- Reducing code duplication
- Setting up `HistoryProvider` event handling
- Add `ProbabilisticSharpeRatio` to `PortfolioStatistics`
- `Probabilistic Sharpe Ratio` will be added to the `RunTimeStatistics`
sent by the `ResultHandlers`
- Making `TradeBuilder.ClosedTrades` thread safe since its accessed by
the `ResultHandlers`
- Removing `:` from live runtime statistics
- Adding unit tests
I noticed some deltas in regression statistics after a change. Before this
change, we log all orders into an {AlgorithmName}.{Language}.orders.log.
This information turned out to be insufficient to identify the regression.
This change also adds daily logging of portfolio value and each security's
holding quantity/value, as well as the full cash book. This should help to
more quickly identify the root cause behind system regression test failures.
- `TimeSliceFactory` will avoid creating empty collections
- `ExecutionModels` will check target collection count before trying to
enumerate
- Reduce calls to .`TotalPortfolioValue`
- `SecurityValues` will only be created when required
- `TimeKeeper` will use TimeZone unique Id as dictionary key. The
TimeZone hash is expensive.
- `AlgorithmManager` will avoid calling `DateTime.UtcNow`,
`ConvertFromUtc()` and `RoundDownInTimeZone()`
- Adding `WorkerThread` class, wrapper for a worker thread that will
execute given `Actions`.
- Algorithm related code (`Construction`, `Initialization`,
`Execution` will be executed by the same `WorkerThread` instance,
this is required for `Python` debugging.
- `Alpha Assets` chart will only store last data point
- Adding new `JsonRoundingConverter` that will round to 4 (number of
digits currently used for comparing alpha statistics) fractional
digits.
- Will be used for `Insights` and `ChartPoint`
- Using `Aggregate(lambda)` vs `Sum(lambda)` since the later is slower
due to performing an extra `Select`
- For `QCAlgorithm.Framework.OnFrameworkData()` will avoid calling
`ToArray()` on empty `Enumerables` due to its cost * the number of
calls. If the `Enumerable` is the empty instance, which is static,
will create a new empty array and return it instead.
- Replacing `SecurityIdentifier` `SecurityType` and `GetHashCode`
implementations for `Lazy` versions, that are performed just once, since
these values do not change and are used multiple times.
- For the different `DataDictionary<T>` implementations adding `this[
Symbol] get; set` since existing overload `this [string]` produces an
extra round operations `Symbol->string->Symbol` with a significant
impact.
- Adding `PortfolioTargetCollection.AddRange()` overload using an array
to avoid unnecessary convertions.
- Removing `using QCAlgorithmFramework = QuantConnect.Algorithm.QCAlgorithm`
- Removing `QCAlgorithmFrameworkBridge`
- Removing `IsFrameworkAlgorithm`
- Making `EmitInsightBasedOnFill` private. Adding new
`IOrderEventProvider` exposing an `event` to which `QCAlgorithm` will
subscribe.
- `AccountType.Cash` algorithms will be allowed to manually trade and
emight insights manually or with alpha model.
Michael Handschuh