154 lines
5.0 KiB
Python
154 lines
5.0 KiB
Python
import time
|
|
from datetime import timedelta
|
|
import logging
|
|
import arrow
|
|
import requests
|
|
from pandas.io.json import json_normalize
|
|
from pandas import DataFrame
|
|
import talib.abstract as ta
|
|
|
|
|
|
logging.basicConfig(level=logging.DEBUG,
|
|
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
def get_ticker_dataframe(pair: str) -> DataFrame:
|
|
"""
|
|
Analyses the trend for the given pair
|
|
:param pair: pair as str in format BTC_ETH or BTC-ETH
|
|
:return: DataFrame
|
|
"""
|
|
minimum_date = arrow.now() - timedelta(hours=6)
|
|
url = 'https://bittrex.com/Api/v2.0/pub/market/GetTicks'
|
|
headers = {
|
|
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.36',
|
|
}
|
|
params = {
|
|
'marketName': pair.replace('_', '-'),
|
|
'tickInterval': 'OneMin',
|
|
'_': minimum_date.timestamp * 1000
|
|
}
|
|
data = requests.get(url, params=params, headers=headers).json()
|
|
if not data['success']:
|
|
raise RuntimeError('BITTREX: {}'.format(data['message']))
|
|
|
|
data = [{
|
|
'close': t['C'],
|
|
'volume': t['V'],
|
|
'open': t['O'],
|
|
'high': t['H'],
|
|
'low': t['L'],
|
|
'date': t['T'],
|
|
} for t in sorted(data['result'], key=lambda k: k['T']) if arrow.get(t['T']) > minimum_date]
|
|
dataframe = DataFrame(json_normalize(data))
|
|
|
|
dataframe['sar'] = ta.SAR(dataframe, 0.02, 0.2)
|
|
|
|
# calculate StochRSI
|
|
stochrsi = ta.STOCHRSI(dataframe)
|
|
dataframe['stochrsi'] = stochrsi['fastd'] # values between 0-100, not 0-1
|
|
|
|
macd = ta.MACD(dataframe)
|
|
dataframe['macd'] = macd['macd']
|
|
dataframe['macds'] = macd['macdsignal']
|
|
dataframe['macdh'] = macd['macdhist']
|
|
|
|
return dataframe
|
|
|
|
|
|
def populate_trends(dataframe: DataFrame) -> DataFrame:
|
|
"""
|
|
Populates the trends for the given dataframe
|
|
:param dataframe: DataFrame
|
|
:return: DataFrame with populated trends
|
|
"""
|
|
"""
|
|
dataframe.loc[
|
|
(dataframe['stochrsi'] < 20)
|
|
& (dataframe['close_30_ema'] > (1 + 0.0025) * dataframe['close_60_ema']),
|
|
'underpriced'
|
|
] = 1
|
|
"""
|
|
dataframe.loc[
|
|
(dataframe['stochrsi'] < 20)
|
|
& (dataframe['macd'] > dataframe['macds'])
|
|
& (dataframe['close'] > dataframe['sar']),
|
|
'underpriced'
|
|
] = 1
|
|
dataframe.loc[dataframe['underpriced'] == 1, 'buy'] = dataframe['close']
|
|
return dataframe
|
|
|
|
|
|
def get_buy_signal(pair: str) -> bool:
|
|
"""
|
|
Calculates a buy signal based on StochRSI indicator
|
|
:param pair: pair in format BTC_ANT or BTC-ANT
|
|
:return: True if pair is underpriced, False otherwise
|
|
"""
|
|
dataframe = get_ticker_dataframe(pair)
|
|
dataframe = populate_trends(dataframe)
|
|
latest = dataframe.iloc[-1]
|
|
|
|
# Check if dataframe is out of date
|
|
signal_date = arrow.get(latest['date'])
|
|
if signal_date < arrow.now() - timedelta(minutes=10):
|
|
return False
|
|
|
|
signal = latest['underpriced'] == 1
|
|
logger.debug('buy_trigger: %s (pair=%s, signal=%s)', latest['date'], pair, signal)
|
|
return signal
|
|
|
|
|
|
def plot_dataframe(dataframe: DataFrame, pair: str) -> None:
|
|
"""
|
|
Plots the given dataframe
|
|
:param dataframe: DataFrame
|
|
:param pair: pair as str
|
|
:return: None
|
|
"""
|
|
|
|
import matplotlib
|
|
|
|
matplotlib.use("Qt5Agg")
|
|
import matplotlib.pyplot as plt
|
|
|
|
# Three subplots sharing x axe
|
|
fig, (ax1, ax2, ax3) = plt.subplots(3, sharex=True)
|
|
fig.suptitle(pair, fontsize=14, fontweight='bold')
|
|
ax1.plot(dataframe.index.values, dataframe['close'], label='close')
|
|
# ax1.plot(dataframe.index.values, dataframe['close_30_ema'], label='EMA(60)')
|
|
# ax1.plot(dataframe.index.values, dataframe['close_90_ema'], label='EMA(120)')
|
|
ax1.plot(dataframe.index.values, dataframe['sar'], 'rx', label='SAR')
|
|
# ax1.plot(dataframe.index.values, dataframe['sell'], 'ro', label='sell')
|
|
ax1.plot(dataframe.index.values, dataframe['buy'], 'bo', label='buy')
|
|
ax1.legend()
|
|
|
|
ax2.plot(dataframe.index.values, dataframe['macd'], label='MACD')
|
|
ax2.plot(dataframe.index.values, dataframe['macds'], label='MACDS')
|
|
ax2.plot(dataframe.index.values, dataframe['macdh'], label='MACD Histogram')
|
|
ax2.plot(dataframe.index.values, [0] * len(dataframe.index.values))
|
|
ax2.legend()
|
|
|
|
ax3.plot(dataframe.index.values, dataframe['stochrsi'], label='StochRSI')
|
|
ax3.plot(dataframe.index.values, [80] * len(dataframe.index.values))
|
|
ax3.plot(dataframe.index.values, [20] * len(dataframe.index.values))
|
|
ax3.legend()
|
|
|
|
# Fine-tune figure; make subplots close to each other and hide x ticks for
|
|
# all but bottom plot.
|
|
fig.subplots_adjust(hspace=0)
|
|
plt.setp([a.get_xticklabels() for a in fig.axes[:-1]], visible=False)
|
|
plt.show()
|
|
|
|
|
|
if __name__ == '__main__':
|
|
while True:
|
|
pair = 'BTC_ANT'
|
|
#for pair in ['BTC_ANT', 'BTC_ETH', 'BTC_GNT', 'BTC_ETC']:
|
|
# get_buy_signal(pair)
|
|
dataframe = get_ticker_dataframe(pair)
|
|
dataframe = populate_trends(dataframe)
|
|
plot_dataframe(dataframe, pair)
|
|
time.sleep(60)
|