-
Notifications
You must be signed in to change notification settings - Fork 46
/
freqtradegym.py
275 lines (224 loc) · 8.41 KB
/
freqtradegym.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
import random
import json
import logging
import copy
from datetime import datetime, timedelta
from typing import Any, Dict, List, NamedTuple, Optional
from pandas import DataFrame
import gym
from gym import spaces
import pandas as pd
import numpy as np
import talib.abstract as ta
from freqtrade.data import history
from freqtrade.data.converter import trim_dataframe
from freqtrade.configuration import (TimeRange, remove_credentials,
validate_config_consistency)
from freqtrade.resolvers import ExchangeResolver, StrategyResolver
from freqtrade.exchange import timeframe_to_minutes, timeframe_to_seconds
from freqtrade.persistence import Trade
from freqtrade.strategy.interface import IStrategy, SellCheckTuple, SellType
logger = logging.getLogger(__name__)
class TradingEnv(gym.Env):
"""A trading environment for OpenAI gym"""
metadata = {'render.modes': ['human', 'system', 'none']}
def __init__(self, config):
super(TradingEnv, self).__init__()
self.config = config
self.config['strategy'] = self.config['gym_parameters']['indicator_strategy']
self.strategy = StrategyResolver.load_strategy(self.config)
self.fee = self.config['gym_parameters']['fee']
self.timeframe = str(config.get('ticker_interval'))
self.timeframe_min = timeframe_to_minutes(self.timeframe)
self.required_startup = self.strategy.startup_candle_count
data, timerange = self.load_bt_data()
# need to reprocess data every time to populate signals
preprocessed = self.strategy.ohlcvdata_to_dataframe(data)
del data
# Trim startup period from analyzed dataframe
dfs = []
for pair, df in preprocessed.items():
dfs.append(trim_dataframe(df, timerange))
del preprocessed
self.rest_idx = set()
idx = 0
for d in dfs:
idx += d.shape[0]
self.rest_idx.add(idx)
print(self.rest_idx)
df = pd.concat(dfs, ignore_index=True)
del dfs
# setting
df = df.dropna()
self.pair = pair
self.ticker = self._get_ticker(df)
del df
self.lookback_window_size = 40
# start
logger.info('Using stake_currency: %s ...', self.config['stake_currency'])
logger.info('Data Length: %s ...', len(self.ticker))
self.stake_amount = self.config['stake_amount']
self.reward_decay = 0.0005
self.not_complete_trade_decay = 0.5
self.game_loss = -0.5
self.game_win = 1.0
self.simulate_length = self.config['gym_parameters']['simulate_length']
# Actions
self.action_space = spaces.Discrete(3)
self.observation_space = spaces.Box(
low=np.full(24, -np.inf), high=np.full(24, np.inf), dtype=np.float)
def _next_observation(self):
row = self.ticker[self.index]
trad_status = 0
if self.trade != None:
trad_status = self.trade.calc_profit_ratio(rate=row.open)
obs = np.array([
# row.open,
# row.high,
# row.low,
# row.close,
# row.volume,
row.adx,
row.plus_dm,
row.plus_di,
row.minus_dm,
row.minus_di,
row.aroonup,
row.aroondown,
row.aroonosc,
row.ao,
# row.kc_percent,
# row.kc_width,
row.uo,
row.cci,
row.rsi,
row.fisher_rsi,
row.slowd,
row.slowk,
row.fastd,
row.fastk,
row.fastd_rsi,
row.fastk_rsi,
row.macd,
row.macdsignal,
row.macdhist,
row.mfi,
row.roc,
# row.bb_percent,
# row.bb_width,
# row.wbb_percent,
# row.wbb_width,
# row.htsine,
# row.htleadsine,
# row.CDLHAMMER,
# row.CDLINVERTEDHAMMER,
# row.CDLDRAGONFLYDOJI,
# row.CDLPIERCING,
# row.CDLMORNINGSTAR,
# row.CDL3WHITESOLDIERS,
# row.CDLHANGINGMAN,
# row.CDLSHOOTINGSTAR,
# row.CDLGRAVESTONEDOJI,
# row.CDLDARKCLOUDCOVER,
# row.CDLEVENINGDOJISTAR,
# row.CDLEVENINGSTAR,
# row.CDL3LINESTRIKE,
# row.CDLSPINNINGTOP,
# row.CDLENGULFING,
# row.CDLHARAMI,
# row.CDL3OUTSIDE,
# row.CDL3INSIDE,
], dtype=np.float)
self.status = copy.deepcopy(row)
return obs
def _take_action(self, action):
# Hold
if action == 0:
return
# Buy
if action == 1:
if self.trade == None:
self.trade = Trade(
pair=self.pair,
open_rate=self.status.open,
open_date=self.status.date,
stake_amount=self.stake_amount,
amount=self.stake_amount / self.status.open,
fee_open=self.fee,
fee_close=self.fee,
is_open=True,
)
self.trades.append({
"step": self.index,
"type": 'buy',
"total": self.status.open
})
logger.debug("{} - Backtesting emulates creation of new trade: {}.".format(
self.pair, self.trade))
# Sell
if action == 2:
if self.trade != None:
profit_percent = self.trade.calc_profit_ratio(rate=self.status.open)
profit_abs = self.trade.calc_profit(rate=self.status.open)
self.money += profit_abs
self.trade = None
self._reward = profit_percent
self.trades.append({
"step": self.index,
"type": 'sell',
"total": self.status.open
})
def step(self, action):
# Execute one time step within the environment
self._reward = 0
self._take_action(action)
self.index += 1
if self._reward > 1.5:
self._reward = 0
if self.index >= len(self.ticker):
self.index = 0
self.steps += 1
self.total_reward += self._reward
# done = (self._reward < self.game_loss) # or (self.steps > self.day_step)
# done = (self.total_reward < self.game_loss) or (self.total_reward > self.game_win) or (self.steps > self.day_step)
done = self.steps > self.simulate_length
obs = self._next_observation()
return obs, self._reward, done, {}
def reset(self):
# Reset the state of the environment to an initial state
self.steps = 0
self.index = random.randint(0, len(self.ticker)-1)
self.trade = None
self.trades = []
self._reward = 0
self.total_reward = 0
self.money = 0
self.visualization = None
return self._next_observation()
def render(self, mode='live', close=False):
# Render the environment to the screen
print(f'Step: {self.index}')
print(f'Reward: {self._reward}')
def load_bt_data(self):
timerange = TimeRange.parse_timerange(self.config['gym_parameters']['timerange'])
data = history.load_data(
datadir=self.config['datadir'],
pairs=self.config['exchange']['pair_whitelist'],
timeframe=self.timeframe,
timerange=timerange,
startup_candles=self.required_startup,
fail_without_data=True,
data_format=self.config.get('dataformat_ohlcv', 'json'),
)
min_date, max_date = history.get_timerange(data)
logger.info(
'Loading data from %s up to %s (%s days)..',
min_date.isoformat(), max_date.isoformat(), (max_date - min_date).days
)
# Adjust startts forward if not enough data is available
timerange.adjust_start_if_necessary(timeframe_to_seconds(self.timeframe),
self.required_startup, min_date)
return data, timerange
def _get_ticker(self, processed: DataFrame) -> List:
processed.drop(processed.head(1).index, inplace=True)
return [x for x in processed.itertuples()]