abu: Redefine Quantitative Trading System

Abu seeks the intelligent-strategy which be able to explained by of a word.

Abu can help users to improve the strategy automatically, take the initiative to analyze the behavior of the orders generated by the strategy to prevent the losing-money transaction.

Right now, we still writting code by hand, abu is designed to be running complete-automatically in the future, including the entire work-process and strategy itself.

Our expectations : abu users only need to provide some seed strategy, on the basis of these seeds, computer continue to self-learning, self-growth, to build a new strategy which can adjust its parameters with the time series data.

Index

Content	Path
Abu Quantitative Trading System	./abupy
Abu Quantitative Trading Tutorial	./abupy_lecture
《量化交易之路》 (The Road of Quantitative Trading) example code	./ipython and ./python
《机器学习之路》 (The Road of Machine Learning) example code	https://github.com/maxmon/abu_ml

Feature

Optimizing strategies by a variety of machine learning techniques
Guiding traders in real trading, improving the profit of strategy, to beat the market

Supported investment markets:

US stocks, A stocks, Hong Kong stocks
Futures market, Options Market
BTC(bitcoin)，LTC(Litecoin)

Project Design Goal：

Separate basic strategy and strategy optimization module
Improve flexibility and adaptability

Install

Deploy

Recommended to use Anaconda to deploy the Python environment, see here

Test

import abupy

UI operation（Non programming）

More examples of UI operations

Document

1: Develop Trading Strategy

Section 1 UI operation tutorial

Trading strategy decide when to invest, backtesting tell us the simulation of profit about this strategy in the historical data.

coding buy factor
backtesting factor step by step
coding sell factor

2: Optimize Trading Strategy

Through stop loss and profit cap to keep profit generated by the strategy, lower risk.

basic stop loss and profit cap strategy
stop loss strategy
profit cap strategy

3: Slippage and Transaction Costs

Consider slippage and transaction costs on applying the strategy

implement slippage strategy
custom transaction costs

type	date	symbol	commission
buy	20150423	usTSLA	8.22
buy	20150428	usTSLA	7.53
sell	20150622	usTSLA	8.22
buy	20150624	usTSLA	7.53
sell	20150706	usTSLA	7.53
sell	20150708	usTSLA	7.53
buy	20151230	usTSLA	7.22
sell	20160105	usTSLA	7.22
buy	20160315	usTSLA	5.57
sell	20160429	usTSLA	5.57

4: Multi-stock Backtesting and Position Control

Backtesting on multiple stocks, control position to lower risk.

multiple stocks with same factor
custom position-control strategy
multiple stocks with different factor
faster running with multi-processing

5: Develop Stock-picking Strategy

A good trading strategy needs a good stock.

coding stock-picking factor
run multiple stock-picking factor
faster running with multi-processing

6: Metrics on Backtesting

Good metric give you right direction.

basic useage about metric
visualization on metrics
expand self-custom metric

7: Seeking the best parameter for Strategy and Scoring

By customizable scoring, seek the best parameter for strategy.Like:how many days should be on MA?

parameter range
using grid search to seek the best parameter
metric on scoring
scoring with different weight
custom scoring by yourself

8: Backtesting on A-Stock

backtesting on A-Stock example
dealing with price limit
analyze multiple trading result

9: Backtesting on HK-Stock

backtesting on HK-Stock example
optimize strategy, improve the stability of the system
encapsulate the "strategy" of optimizing the strategy as a class decorator

10: Backtesting on Bitcoin and LiteCoin

analyze bitcoin and litecoin market trend
visualization analysis on bitcoin and litecoin market trend
backtesting on Bitcoin and LiteCoin market

bitcoin loss10: [-26.895, -3.284] , top10：(4.182, 38.786]
bitcoin recent 1 year risk lower：loss10: [-16.273, -2.783], top10: (3.948, 15.22]
litecoin loss10: [-28.48, -4.1]， top10: (4.405, 41.083]
litecoin recent 1 year risk lowerloss10: [-22.823, -3.229] 高收益top10: (5.0606, 37.505]

btcchange	btc365change	ltcchange	ltc365change
[-26.895, -3.284]	[-16.273, -2.783]	[-28.48, -4.1]	[-22.823, -3.229]
(-3.284, -1.547]	(-2.783, -1.056]	(-4.1, -2.022]	(-3.229, -1.375]
(-1.547, -0.8]	(-1.056, -0.424]	(-2.022, -0.922]	(-1.375, -0.655]
(-0.8, -0.224]	(-0.424, -0.071]	(-0.922, -0.389]	(-0.655, -0.226]
(-0.224, 0.143]	(-0.071, 0.272]	(-0.389, 0]	(-0.226, 0.078]
(0.143, 0.568]	(0.272, 0.698]	(0, 0.413]	(0.078, 0.453]
(0.568, 1.108]	(0.698, 1.316]	(0.413, 0.977]	(0.453, 0.913]
(1.108, 2.171]	(1.316, 2.334]	(0.977, 1.889]	(0.913, 1.957]
(2.171, 4.182]	(2.334, 3.948]	(1.889, 4.405]	(1.957, 5.0606]
(4.182, 38.786]	(3.948, 15.22]	(4.405, 41.083]	(5.0606, 37.505]

11: Backtesting on Futures Market

features of futures market
backtest bullish contract
backtest bearish contract
optimize strategy by displacement ratio

12: Bitcoin with Machine Learning Example

How to use machine learning technology correctly in quantitative trading of investment goods?

extraction of bitcoin features
abu built-in machine learning module
verification and unbalanced technology of test set
inherits AbuMLPd to encapsulate data processing

13: Quantitative Technology Analysis and Applications

Technical analysis is based on three assumptions:1. The market discounts everything.2. Price moves in trends.3. History tends to repeat itself.

resistance line, support line automatically drawn
analysis of gap
analysis of traditional technical metric

14: Quantitative Correlation Analysis Application

Behind similar investment trend, it is often with similar investment groups.

relevant similarity measure
distance measurement and similarity
application of similarity interface
natural correlation

15: Quantitative Trading and Search Engines

Search and analyze failed orders generated by strategy, intercept possible failing orders by the ump .

backtest splitted set
analyze transaction manually
concept of referee system
angle referee
give a natural and reasonable explanation
optimal classification-cluster selection

16: UMP Main Referee System

gap main-referee
price main-referee
fluctuation main-referee
verify whether the main-referee works well
organize referees to make more complex comprehensive decisions
let the referee learn how to cooperate with their own to make the most correct judgments

17: UMP Edge Referee System

gap edge-referee
price edge-referee
fluctuation edge-referee
comprehensive edge-referee
verify whether the edge-referee works well
open the edge-referee mode

18: Custom Referee System

train new main-referee from different perspectives
train new edge-referee from different perspectives
add a new perspective to record the game (record backtesting feature)
main-referee with the new perspective
edge-referee with the new perspective

The design goals of ump module are:

no need to hard-code strategy
mo need to manually set the threshold
separate the strategy and optimize-monitor module to improve flexibility and adaptability
discover issues hidden in strategy
auto-learn new transaction data

19: Data Source

Abu support stock, futures, digital coins and other financial investment. Abu support quotes query and transactions, and also offer a high degree of customization.

switch data mode
switch data storage
switch data source
update the whole market data
access to external data sources：stock data sources
access to external data sources：futures data sources
access to external data sources：bitcoin and litecoin data sources

Contact us

More abu quantitative tutorial please pay attention to our WeChat public number: abu_quant

Also any questions, please contact my personal WeChat number:

License

GPL