awesome-archive / trading-gym

A Trading environment base on Gym

Trading-Gym

install

$ pip install trading-gym

Creating features with ta-lib is suggested, that will improve the performance of agent and make it easy to learn. You should install ta-lib before it. Take Ubuntu x64 for example.

$ wget http://prdownloads.sourceforge.net/ta-lib/ta-lib-0.4.0-src.tar.gz 
$ tar -zxvf ta-lib-0.4.0-src.tar.gz
$ cd ta-lib/
$ ./configure --prefix=$PREFIX
$ make install

$ export TA_LIBRARY_PATH=$PREFIX/lib
$ export TA_INCLUDE_PATH=$PREFIX/include

$ pip install TA-Lib

See more.

input format

[
    {
        "open": 10.0,
        "close": 10.0,
        "high": 10.0,
        "low": 10.0,
        "volume": 10.0,
        "date": "2019-01-01 09:59"
    },
    {
        "open": 10.1,
        "close": 10.1,
        "high": 10.1,
        "low": 10.1,
        "volume": 10.1,
        "date": "2019-01-01 10:00"
    }
]

actions

Action	Value
PUT	0
HOLD	1
PUSH	2

Examples

quick start

from trading_gym.env import TradeEnv
import random


env = TradeEnv(data_path='./data/test_exchange.json')
done = False
obs = env.reset()
for i in range(500):
    action = random.sample([0, 1, 2], 1)[0]
    obs, reward, done, info = env.step(action)
    env.render()
    if done:
        break

A sample train with stable-baselines

from trading_gym.env import TradeEnv
from stable_baselines.common.vec_env import DummyVecEnv
from stable_baselines import DQN
from stable_baselines.deepq.policies import MlpPolicy


data_path = './data/fake_sin_data.json'
env = TradeEnv(data_path=data_path, unit=50000, use_ta=True)
env = DummyVecEnv([lambda: env])

model = DQN(MlpPolicy, env, verbose=2, learning_rate=1e-5)
model.learn(200000)


obs = env.reset()
for i in range(8000):
    action, _states = model.predict(obs)
    obs, rewards, done, info = env.step(action)
    env.render()
    if done:
        break

observation

• native obs: shape=(*, 51, 6), return 51 history data with OCHL

env = TradeEnv(data_path=data_path)

• obs with ta: shape=(*, 10), return obs using talib.
• • default feature: ['ema', 'wma', 'sma', 'sar', 'apo', 'macd', 'macdsignal', 'macdhist', 'adosc', 'obv']

env = TradeEnv(data_path=data_path, use_ta=True)

• custom obs: return custom obs

def custom_obs_features_func(history, info):
    obs = []
    for observation in history:
        obs.append[observation.close]
    return obs


env = TradeEnv(data_path=data_path,
               get_obs_features_func=custom_obs_features_func,
               ops_shape=(1))

Reward

• reward = fixed_profit
• profit = fixed_profit + floating_profit
• floating_profit = (latest_price - avg_price) * unit
• unit = int(nav / buy_in_price)
• avg_price = ((buy_in_price * unit) + charge) / unit
• fixed_profit = SUM([every floating_profit after close position])