Read the book a quantitative investment, notes the following.
Title: Open quantitative investment black box
Author: Rishi K. Narang
Translator: Guo Jianguang
Publisher: Machinery Industry Press
Edition: March 2012 First Edition first brush
preface
quantitative trading is human after rigorous research to get deals policy, and then delivered to the system to implement. The main difference between trading strategy with subjective judgment type is how to generate strategies and how to implement.
Alpha strategy to return on the portfolio size by adjusting the time and select different positions. Beta policy replication index, or slightly exceed the performance of the index.
The first part of the world's quantitative trading
Chapter 1 Why do we focus on quantitative trading
can be learned from wide body passenger: Force yourself to make more in-depth reflection on the various aspects of the investment strategy adopted. Strict adherence to discipline.
Chapter 2 Introduction to quantitative trading
difference between good and mediocre quants is that whether quants have good judgment. When there was information model can not handle driving changes in the market, the need for manual intervention policy enforcement.
How quantitative trading system is composed of?
The Alpha model, risk control model, cost control model portfolio construction model form, and form execution model. Building and execution model must also have high-quality data.
The second part of the black box to open the
Chapter 3 α model: how quants make money?
α refers to investors and market fluctuations unrelated to return. It refers to the quants to profit and manifested in the investment process skills. The pursuit of alpha-return strategies, asset allocation decision is decision time portfolio and set its position size optional nature. The core idea: not always good or always poor financial products. The key factor is to decide when to hold more and holding empty. The idea behind value investing is overvalued and sell when the stock to buy when the stock undervalued. Any effective alpha model has its scope of application. Understand quants most critical point is to understand how they think scientifically. Alpha model theory driven and data-driven sub-categories. The vast majority of theory-driven quants can be divided into five categories: trend type, recovery type, value / benefit type, growth, quality type. Price data into relevant data and fundamental data. It assumes that the data is data-driven model can imply what is going to happen, and by means of analysis techniques can identify some of the market trend. Advantage is know not why, fewer people use. The disadvantage is that the calculation amount may exceed the acceptable range, the other dependent on the historical data, requires constant adjustment. Noisy input data might generate an error signal. Differences quants implementation strategies including targets (in the end what model to predict), the investment period (high frequency, neutral, long-term), betting structure, investment scope, set the model, the operating frequency. Various strategies alpha mixed linear model is often used method, non-linear model and machine learning models.
Chapter 4 Risk control model
of risk management is to improve the quality and sustainability of returns, and purposeful selection and control of the exposure scale implementation. A great advantage is the ability to measure quantitative trading of various exposures, and for the selection of these exposures are purposeful. Which it wishes to recognize the systemic risk may assume, measure the size of the exposure of each portfolio and then decide whether to accept these risks.
Chapter 5 transaction cost model
The idea behind: Any transaction must spend costs. Transaction costs accounted for 20% -50% return. Two reasons for the transaction: improve returns and reduce losses. But minor modifications to its transaction, it may not cover the transaction costs. Transaction cost model is not to minimize transaction costs, but are used to determine the required transaction cost model. The composition of transaction costs: commissions, slippage (refer to the traders decided to start trading in order to actually execute price changes between the two periods), market impact (impact of the order traders on the market). Trends follow the strategy is sensitive to slippage.
Type of transaction cost model: cost model constants transaction, transaction cost model linear, piecewise linear model transaction costs, transaction costs quadratic model.
Chapter 6 portfolio construction model
goal is to determine the quants to be held by the portfolio. There are two types of ways, one is rule-based, and the other using optimization tools.
Rule-based portfolio construction model: equal positions weights equal risk weight, alpha-driven weights, weights tree. Their common challenge is how to explain the principles of economic rationality and driving rules behind them.
Portfolio optimization tools: need to enter the expected returns, expected volatility, expected correlation. The most commonly used optimization tools: no constraint optimization, constrained optimization condition, Blake - Letterman optimization method, Gerry Nord - Kahn Methods: portfolio optimization factors, resampling, most based on data mining Optimization. Output target portfolio.
Chapter 7 Execution Model
Transactions: Electronic trading or manually.
Order execution algorithm has two types of market orders and limit orders. There are different ways of execution orders. To decide whether passive or active initiative to act. Large orders can be split into small order execution.
Chapter 8 data
input system determines what you can use this system to do from nature. The basic data types are divided into price-related data and fundamental data. Handling missing data: the known data using the most recent, or insertion reasonable value. Processing error data, such as binary errors (when cells are components), the use of the singular point filters. Front view of the deviation, i.e., the use of future data.
Chapter 9 research
objective of the study was to examine through deliberate investment strategy. Good performance common feature of quantitative investment strategy is to insist on the scientific method in the study. The difference is quantitative strategies more unstable, need to continue to modify, update.
Source thought: Market Watch, literature, learn from other quantitative traders, lessons subjective judgment type traders. Inspection process is essential for research. In quantitative trading, the model is an approximation of the world. The first step in the inspection process is to use sample data to train the model within to find the optimal parameters. Course of the study there is a variety of pitfalls to be avoided. To test the hypothesis.
The third part of the investment guidelines quantitative trading strategy
Chapter 10 quantitative trading strategies inherent risk
model risk: modeling unsuitability, the model wrong settings, execution errors.
Change the logic of the market risk
external shocks that
the risk of proliferation or homogeneous investor risk
Chapter 11 criticism of quantitative trading - Ensure that accurate records of
transactions is an art rather than a science: human behavior can be modeled.
Due to underestimate the risks, quants led to more market turmoil: to quantify the birth of trading before the market turmoil is not small.
Quants can not handle unusual events in the market and rapid change.
Quants have the same transaction.
Only a small number of long-term large-scale quantitative funds in order to thrive: the huge amount of money management is not always a good thing.
Quantitative investment exist in data mining fault: the problem of over-fitting. When given the complexity of the model, the more available data, there have been less likely to fit the problem. Long-term investment, investment decisions using data mining is not feasible.
Chapter 12 evaluation quants and quantitative trading strategies
many quantitative trading techniques from the experience and know-how, rather than from a mathematical advantage.
To understand the evaluation of a wide customer strategy he had built as much as possible.
Investors are more advantages inherent or absolute, rather than on the performance of comparative advantage. The right way to deal with adversity, first of all have good monitoring tools.
Looking to the future Chapter 13 quantitative investment
strategy
Summary: The "black box" quantitative trading model is divided into alpha, risk control, cost control three parts. Alpha is a quantitative investment model can make money simply divided into theory driven and data-driven type, the former type is divided into trends, recovery type, value / benefit type, growth, quality type. It assumes that the data is data-driven model can imply what is going to happen, and by means of analysis techniques can identify some of the market trend. A great advantage is the ability to measure quantitative trading of various exposures, and for the selection of these exposures are purposeful. Transaction costs accounted for 20% -50% return. The composition of transaction costs: commissions, slippage (refer to the traders decided to start trading in order to actually execute price changes between the two periods), market impact (impact of the order traders on the market). Traders will quantify these three parts together to form their own quantitative trading model, after backtesting test, put into practice. Quantify traders through their own research, to other peer learning, quantitative trading strategies to get the idea of learning from the literature. Quantitative trading is not a panacea, there are risks and problems of its own.
In addition to continue to experiment a bit traded pyalgotrade frame function, there marketOrder / limitOrder and enterLong / enterShort / enterLongLimit / enterShortLimig categories. enterLong market price (Bar next opening price) to buy, enterLongLimit when the market (under the opening price of a Bar) equal to or lower than a specified price when buying, enterShort and enterShortLimit are sold. Is beginning to enter more top approach is recommended. goodTillCanceled To adapt to the real disk interface, the interface may limit a firm offer the previous day's orders will not be executed, so set goodTillCanceled = True to ensure that the next day or more after the time, the order is still in effect until canceled manually.
(Refer to the above: https://blog.51cto.com/youerning/2162751 would like to thank!)
To do the test example pyalgotrade documentation of homework, results are as follows:
The former is a single under marketOrder, after a single use is under enterLong, only orders in __init__, and then has been held, we can see two single-mode final earnings or nuances of. The benefits of using enterLong can return to a position of the object, then you can output a variety of transaction information. Also looked at the raw data, it is Adj.Open, that is, the opening price after the recovery of the right to buy. Could it be trading with the closing price of the class?
Next, I intend to find this book of practical operation, according to practice it.
The code https://github.com/zwdnet/MyQuant/tree/master/06
我发文章的三个地方，欢迎大家在朋友圈等地方分享，欢迎点“在看”。
我的个人博客地址：https://zwdnet.github.io
我的知乎文章首页： https://www.zhihu.com/people/zhao-you-min/posts
我的博客园博客地址： https://www.cnblogs.com/zwdnet/
我的微信个人订阅号：赵瑜敏的口腔医学学习园地