zwpy2bt

Overview

Porting strategies from zwpython to Backtrader. Original strategies are located in zwStrategy.py in zwpython.

Strategy list:

Tim0Strategy (tim0Trad)
SmaStrategy (SMA_sta)
CmaStrategy (CMA_sta)
VwapStrategy (VWAP_sta)
BBandsStrategy (BBANDS_sta)
TurStrategy (tur10)
MacdV1Strategy (macd10)
MacdV2Strategy (macd20)
KdjV1Strategy (kdj10)
KdjV2Strategy (kdj20)
RsiStrategy (rsi10)

Prerequisites

Python 2.7
Python 3.2 / 3.3/ 3.4 / 3.5

Errors could occur in newer version python.

Installation

clone repo:

git clone https://github.com/maglili/zwpy2bt.git

install requirements:

pip install backtrader[plotting]

Run

python main.py

If you use 600401_yahoo.csv as sample data, then reverse parameter in datafeed should be "True", because 600401_yahoo.csv is decreased order.

    data = bt.feeds.YahooFinanceCSVData(
        dataname=datapath,
        fromdate=datetime.datetime(2015, 1, 1),
        todate=datetime.datetime(2015, 12, 31),
        reverse=True,
    )

Strategy Package

The Strategy package consists of 3 modules.

BaseStrategyFrame
utils
zwpy_sta

./stock-zwpython/
├── main.py
├── readme.md
├── result.png
├── sample_data
│   ├── 600401_yahoo.csv
│   ├── orcl-1995-2014.txt
│   └── readme.md
└── Strategy
    ├── BaseStrategyFrame.py
    ├── __init__.py
    ├── utils.py
    └── zwpy_sta.py

2 directories, 10 files

Module Description

BaseStrategyFrame: Define the base structure of Strategy class, BaseStrategyFrame is inheriting the bt.Strategy class, and all strategies in zwpy_sta.py are inheriting the BaseStrategyFrame class.

utils: Define class / function tools.

zwpy_sta: Define various strategies from zwpython. Here, strategies are sub-class inherit from BaseStrategyFrame. This can help user reduce code works because strategies are now only need fewer code.

Result Comparison

A Comparison experiment is conducted and the result (stocks value + balances) as shown on following table.

Experiment setting:

Initial cash: 10000
Data: 600401_yahoo.csv
Backtest period: 2015, 1, 1 ~
Sizer: PercentSizerInt, percents=90

Original result:

Strategy	zwpython	backtrader
Tim0Strategy	11158.60	11158.60
SmaStrategy	10892.66	11055.51
CmaStrategy	12860.46	13404.05
VwapStrategy	17102.59	16823.13
BBandsStrategy	9966.22	9966.22
TurStrategy	15193.05	15193.05
MacdV1Strategy	14405.99	15758.85
MacdV2Strategy	10557.62	11860.55
KdjV1Strategy	18550.29	20098.48
KdjV2Strategy	14807.79	16429.68
RsiStrategy	7808.42	8995.14

After handing decimal place problem:

Remove round process in backtest.

Strategy	zwpython	backtrader
Tim0Strategy	11158.60	11158.60
SmaStrategy	10892.66	10892.66
CmaStrategy	12860.46	12860.46
VwapStrategy	17102.59	16823.13
BBandsStrategy	9966.22	9966.22
TurStrategy	15193.05	15193.05
MacdV1Strategy	14405.99	14301.22
MacdV2Strategy	10557.62	10557.62
KdjV1Strategy	18550.29	20098.48
KdjV2Strategy	14807.79	16429.68
RsiStrategy	7808.42	8995.14

After handing formula problem:

To correct the VWAP formula in zwpython.

Strategy	zwpython	backtrader
VwapStrategy	17102.59	17102.59

There are some bugs in zwpython and I already fixed some of them, so you might not have a same result as the table above.

Why still a difference b/t bt and zwpy?

The difference comes from decimal place problem in technical index calculation (e.g. VWAP, MACD, KDJ,...), say RSI = 93.700 vs RSI = 96.849, will have different decision in strategy. That merely a order difference b/t a day can cause a big different result, the problem cannot solve by merely remove rounding process that in backtest in zwpython.

Which one is more accurate?

The result of backtrader might more accurate since it still maintaining by github community (6.6k star). Technical index in Backtrader are calculated by bt.indicator or TA-lib package, which offer a reliable result, while in zwpython, Technical index are calculated by functions that created by creator of zwpython, there might exist some bugs (and there are indeed some bugs in some scripts in zwpython).

webclinic017 / zwpy2bt