2. Core Concepts at a Glance

This chapter organizes the core concepts in qteasy in tables and brief notes, making it easy to quickly look them up when reading subsequent chapters or using the API. For a more complete structure and relationships, see Overall Architecture and Design Approach.

2.1. 1. 概念表（名称 / 含义 / 在文档中的位置）

Concepts	Meaning	Where It Fits in This Series
DataSource	An abstraction of local data storage that manages historical data tables in files or databases, and does not proactively pull data.	Data Acquisition, Storage, and Data Types
Data Table	Tables in the DataSource that store market, financial, macro, and other data in a unified structure.	Data Acquisition, Storage, and Data Types
DataType	A type description of “information that a strategy can reference,” determined by name, freq, asset_type, etc.	Data Retrieval, Storage, and Data Types, How Strategies Declare and Use Data
htype	Historical data type name (e.g., close, open), often used together with freq and asset_type to identify a kind of data	Data Acquisition, Storage, and Data Types
dtype_id	The unique identifier of a DataType, in the form name_assettype_freq (e.g., close_E_d)	See “Composition and Examples of dtype_id” below.
Operator	The strategy container and execution entry point; holds a list of Groups and the group_timing_table, and provides run(config)	Operator and Group
Group	A collection of strategies with the same run_freq and run_timing, with a signal_type and a blender	Operator and Group
Strategy	The base Strategy class and its subclasses; declares data_types/window_length and Parameters, and implements realize()	How a Strategy Runs
run_freq	The execution frequency of a strategy (group), e.g., 'd' for daily, 'm' for monthly.	Operator and Group
run_timing	When a strategy (group) runs within a period, e.g., 'close' at close, 'open' at open.	Operator and Group
group_timing_table	A Time Step × Group table that marks which Groups run at each time step.	Operator and Group
blender	How signals from multiple strategies within the same Group are combined (expression or default rules)	Operator and Group
signal_type	Signal types: PT (target position), PS (proportional buy/sell), VS (quantity buy/sell)	Below: “Three signal types PT/PS/VS”
Parameter	Definition of strategy tunable parameters (name, type, value range, etc.)	How a Strategy Runs
data_types	Required data types declared by the strategy (DataType or list)	How a strategy declares and uses data
window_length	Window length of historical data required by the strategy (e.g., 20 days)	How a strategy declares and uses data
realize()	Strategy logic entry point; takes no parameters; fetches data via get_pars/get_data and returns signals	How a Strategy Runs
get_data / get_pars	Within realize(), fetch data by dtype_id and fetch tunable parameters by parameter name	How a strategy declares and uses data
Backtester	Backtest executor: drives the Operator according to group_timing_table and simulates fills	Backtesting, Live Trading, and Optimization
Optimizer	Optimizer: searches the parameter space and runs backtests multiple times, aggregating the objective function	Backtesting, Live Trading, and Optimization
Trader	Live trading management: triggers the Operator according to schedule and coordinates the Broker	Backtesting, Live Trading, and Optimization
Broker	Abstract interface for order execution and fill reports	Backtesting, Live Trading, and Optimization
config	Run configuration (asset pool, time range, costs, capital plan, etc.), shared by backtesting/live trading/optimization	Backtesting, Live Trading, and Optimization
qt.run	Unified entry point: qt.run(op, mode=…, **kwargs) → op.run(config, datasource, logger)	Backtesting, Live Trading, and Optimization

2.2. 2. dtype_id 的构成与示例

dtype_id is composed of three parts concatenated together: name_assettype_freq.

• name: data type name (e.g., close, open, total_mv, pe).

• assettype: asset type, such as E (stocks), IDX (index), ANY (any).

• freq: data frequency, e.g., d (daily), w (weekly), m (monthly), q (quarterly).

Example:

dtype_id	Meaning
close_E_d	Daily closing price of stocks
close_IDX_d	Daily closing price of an index
close_ANY_d	Daily closing price of any asset
total_mv_E_q	Quarterly total market capitalization of stocks
pe_E_d	Daily stock P/E ratio (if available)

When the strategy calls get_data(dtype_id), the id used must match the dtype_id generated by the DataType declared by the strategy (usually viewable via the strategy’s data_type_ids property).

2.3. 3. 三种信号类型 PT/PS/VS 的语义对比表

Type	Full name	Meaning (one sentence)	Example
PT	Position Target	Target position ratio: the signal indicates the desired long/short position ratio to reach.	0.5 indicates a target 50% long position.
PS	Proportion Signal	Proportional trading: the signal indicates what proportion of capital or holdings to buy/sell.	0.5 indicates using 50% of capital to buy, or selling 50% of the position.
VS	Volume Signal	Quantity-based trading: the signal indicates the quantity to buy/sell (shares or units).	100 indicates buying 100 shares; -50 indicates selling 50 shares.

The same numeric value will be interpreted as different order intents under different signal_type; a Group’s signal_type is determined when adding the strategy, and strategies within the same group output the same type of signal, which is then blended by the blender.

2.4. 4. 三种策略基类的输入/输出对比表

Base class	Typical use cases	Inputs (get_data, etc.)	Outputs (return value of realize)
RuleIterator	Market timing, single-rule iteration	Single-asset or multi-asset data window (by dtype_id)	Scalar signal (e.g., 1 / -1 / 0); the same rule applies to all assets
FactorSorter	Factor-based stock selection	Multi-asset cross-sectional data (e.g., factor values for multiple stocks)	1D factor array (one value per asset); the engine selects stocks by factor ranking and filtering rules
GeneralStg	General-purpose multi-asset strategy	Multi-asset, multi-data-type data window	1D signal array (one target position or buy/sell ratio per asset)

For more details, see How Strategy Runs: Timing, Data, and Parameters.