python 3.x

README.md

@@ -1,4 +1,4 @@
-# Trading Gym
+# Trading Gym (Python 3.x)
 
 Trading Gym is an open-source project for the development of reinforcement learning algorithms in the context of trading.
 It is currently composed of a single environment and implements a generic way of feeding this trading environment different type of price data.
@@ -7,6 +7,10 @@ It is currently composed of a single environment and implements a generic way of
 
 `pip install tgym`
 
+if you clone the project to local, run
+
+`python setup.py install`
+
 We strongly recommend using virtual environments. A very good guide can be found at http://python-guide-pt-br.readthedocs.io/en/latest/dev/virtualenvs/.
 
 ## The trading environment: `SpreadTrading`
@@ -28,3 +32,16 @@ Our environments API is strongly inspired by OpenAI Gym. We aim to entirely base
 Some examples are available in `tgym/examples/`
 
 To run the `dqn_agent.py` example, you will need to also install keras with `pip install keras`. By default, the backend will be set to Theano. You can also run it with Tensorflow by installing it with `pip install tensorflow`. You then need to edit `~/.keras/keras.json` and make sure `"backend": "tensorflow"` is specified.
+
+## To Do List
+1. Python 3
+    - [x] a. print issue
+    - [x] b. module import  (need to python setup.py install/develop)
+    - [x] c. class iter issue (__next(self)__, change next(obj))
+    - [x] d. testing coverage
+2. Add more features
+    - [ ] a. pips (%)
+    - [ ] b. limit order
+    - [ ] c. sl, pt, holding postion value
+    - [ ] d. on-line data feeder
+    - [ ] e. realtime feed

examples/csv_streamer.py

@@ -1,5 +1,6 @@
-from tgym.envs import SpreadTrading
-from tgym.gens import CSVStreamer
+from __future__ import absolute_import 
+from tgym.envs.trading_spread import SpreadTrading
+from tgym.gens.csvstream import CSVStreamer
 
 generator = CSVStreamer(filename='./examples/price_2.csv')
 
@@ -11,7 +12,7 @@
 
 environment.render()
 while True:
-    action = raw_input("Action: Buy (b) / Sell (s) / Hold (enter): ")
+    action = input("Action: Buy (b) / Sell (s) / Hold (enter): ")
     if action == 'b':
         action = [0, 1, 0]
     elif action == 's':

examples/dqn_agent.py

@@ -11,7 +11,7 @@
 from keras.layers import Dense
 from keras.models import Sequential
 from keras.optimizers import Adam
-from tgym.envs import SpreadTrading
+from tgym.envs.trading_spread import SpreadTrading
 
 
 class DQNAgent:
@@ -110,16 +110,42 @@ def _get_batches(self):
 if __name__ == "__main__":
     import matplotlib.pyplot as plt
 
-    from tgym.envs import SpreadTrading
-    from tgym.gens.deterministic import WavySignal
+    from tgym.envs.trading_spread import SpreadTrading
+    from tgym.gens.deterministic import WavySignal,RandomGenerator
+    from tgym.gens.csvstream import CSVStreamer
+    from tgym.envs.trading_tick import TickTrading
     # Instantiating the environmnent
-    generator = WavySignal(period_1=25, period_2=50, epsilon=-0.5)
+    gen_type = 'C'
+    if gen_type == 'W':
+        generator = WavySignal(period_1=25, period_2=50, epsilon=-0.5,ba_spread = 0.0001 )
+    elif gen_type == 'R':
+        generator = RandomGenerator(spread = 0.0001,range_low =1.0,range_high=2.0)
+    elif gen_type == 'C':
+        filename = r'./examples/price_usdeur.csv'
+        generator = CSVStreamer(filename=filename)   
+
     episodes = 50
     episode_length = 400
     trading_fee = .2
     time_fee = 0
     history_length = 2
-    environment = SpreadTrading(spread_coefficients=[1],
+    episode_length = 200000
+
+    # history_length number of historical states in the observation vector.
+    profit_taken = 10
+    stop_loss = -5
+    render_show = False
+    trading_type = 'T'
+    if trading_type == 'T':
+        environment = TickTrading(  data_generator=generator,
+                            trading_fee=trading_fee,
+                            time_fee=time_fee,
+                            history_length=history_length,
+                            episode_length=episode_length,
+                            profit_taken = profit_taken,
+                            stop_loss = stop_loss)
+    else:
+        environment = SpreadTrading(spread_coefficients=[1],
                                 data_generator=generator,
                                 trading_fee=trading_fee,
                                 time_fee=time_fee,

examples/price_2.csv

@@ -0,0 +1,6 @@
+10.0,21.0,11.0,22.0
+12.0,22.0,12.0,24.0
+12.0,22.0,13.0,24.0
+12.1,22.1,15.1,27.1
+12.5,22.5,16.1,26.1
+11.5,21.5,17.1,21.1

examples/random_generator.py

@@ -2,20 +2,17 @@
 In this example we show how a random generator is coded.
 All generators inherit from the DataGenerator class
 The class yields tuple (bid_price,ask_price)
+spread = ask_price - bid_price
+limit range between range_low, range_high
+better format for return value
 """
+from __future__ import absolute_import 
 import numpy as np
 from tgym.core import DataGenerator
+from tgym.gens.deterministic import RandomGenerator
 
-
-class RandomGenerator(DataGenerator):
-    @staticmethod
-    def _generator(ba_spread=0):
-        while True:
-            val = np.random.randn()
-            yield val, val + ba_spread
-
-
-time_series_length = 10
-mygen = RandomGenerator()
-prices_time_series = [mygen.next() for _ in range(time_series_length)]
-print prices_time_series
+if __name__ == "__main__":
+    time_series_length = 100
+    mygen = RandomGenerator(spread = 0.01,range_low =1.0,range_high=2.0)
+    prices_time_series = [next(mygen) for _ in range(time_series_length)]
+    print (prices_time_series)

examples/testing.py

@@ -0,0 +1,46 @@
+def add_position(action,pr,ls):
+    if action in ('b','s'): 
+        bt = (action,(_tick_buy if action == 'b' else _tick_sell), pr, ls)
+        _holding_position.append(bt)
+
+def close_position(position):
+    if position[0] == 'b':
+        pr = (_tick_buy - position[1]) * 10000.0
+        st = (position[1] - _tick_sell) * 10000.0
+    elif position[0] == 's':
+        pr = (position[1] - _tick_sell ) * 10000.0
+        st = (_tick_buy - position[1]  ) * 10000.0
+
+    if pr >= position[2]:
+        print(position, pr, 'pr')
+        return pr, True
+    elif st <= position[3]:
+        print(position, st,'st')
+        return st, True
+    else :
+        print(position, 'nothing')
+        return 0, False
+
+
+
+def calculate_position():
+    rewards = 0
+    result = []
+    for position in _holding_position:
+        reward, remove = close_position(position)
+        rewards += reward
+        if remove:
+            result.append(position)
+    return rewards, result
+
+
+#testing script 
+_tick_buy = 1.3330
+_tick_sell = 1.3329
+_holding_position = [('b',1.3333,100,50),('b',1.3343,300,600),('b',1.3383,90,20),('s',1.3333,100,50)]
+add_position('s',100,501)
+add_position('b',100,502)
+rewards, new_position = calculate_position()
+
+print (rewards)
+print (new_position)

examples/trading_spread_env.py

@@ -0,0 +1,52 @@
+"""
+The aim of this file is to give a standalone example of how an environment  runs.
+"""
+import os
+import numpy as np
+from tgym.core import DataGenerator
+from tgym.envs.trading_tick import TickTrading
+from tgym.gens.deterministic import WavySignal, RandomGenerator
+
+
+gen_type = 'W'
+if gen_type == 'W':
+    generator = WavySignal(period_1=25, period_2=50, epsilon=-0.5,ba_spread = 0.0001 )
+elif gen_type == 'R':
+    generator = RandomGenerator(spread = 0.0001,range_low =1.0,range_high=2.0)
+
+
+episode_length = 200
+trading_fee = 0.2
+time_fee = 0
+# history_length number of historical states in the observation vector.
+history_length = 2
+profit_taken = 6000
+stop_loss = -7000
+
+environment = TickTrading(  data_generator=generator,
+                            trading_fee=trading_fee,
+                            time_fee=time_fee,
+                            history_length=history_length,
+                            episode_length=episode_length,
+                            profit_taken = profit_taken,
+                            stop_loss = stop_loss)
+
+environment.render()
+i = 0
+while True:
+    #action = input("Action: Buy (b) / Sell (s) / Hold (enter): ")
+
+    # if action == 'b':
+    #     action = [0, 1, 0]
+    # elif action == 's':
+    #     action = [0, 0, 1]
+    # else:
+    #     action = [1, 0, 0]
+    '''
+    kind of random action
+    '''
+    action = [0, 1, 0] if i%7 == 0 else ([0, 0, 1] if i%13 == 0 else [1, 0, 0])
+
+    environment.step(action)
+    environment.render()
+    i += 1

examples/trading_tick_env.py

@@ -0,0 +1,56 @@
+"""
+The aim of this file is to give a standalone example of how an environment  runs.
+"""
+import os
+import numpy as np
+from tgym.core import DataGenerator
+from tgym.envs.trading_tick import TickTrading
+from tgym.gens.deterministic import WavySignal, RandomGenerator
+from tgym.gens.csvstream import CSVStreamer
+
+gen_type = 'C'
+if gen_type == 'W':
+    generator = WavySignal(period_1=25, period_2=50, epsilon=-0.5,ba_spread = 0.0001 )
+elif gen_type == 'R':
+    generator = RandomGenerator(spread = 0.0001,range_low =1.0,range_high=2.0)
+elif gen_type == 'C':
+    filename = r'./examples/price_usdeur.csv'
+    generator = CSVStreamer(filename=filename)   
+
+episode_length = 200000
+trading_fee = 0.2
+time_fee = 0
+# history_length number of historical states in the observation vector.
+history_length = 2
+profit_taken = 10
+stop_loss = -5
+render_show = False
+environment = TickTrading(  data_generator=generator,
+                            trading_fee=trading_fee,
+                            time_fee=time_fee,
+                            history_length=history_length,
+                            episode_length=episode_length,
+                            profit_taken = profit_taken,
+                            stop_loss = stop_loss)
+
+if render_show :
+    environment.render()
+i = 0
+while True:
+    #action = input("Action: Buy (b) / Sell (s) / Hold (enter): ")
+
+    # if action == 'b':
+    #     action = [0, 1, 0]
+    # elif action == 's':
+    #     action = [0, 0, 1]
+    # else:
+    #     action = [1, 0, 0]
+    '''
+    kind of random action
+    '''
+    action = [0, 1, 0] if i%7 == 0 else ([0, 0, 1] if i%13 == 0 else [1, 0, 0])
+
+    environment.step(action)
+    if render_show :
+        environment.render()
+    i += 1

setup.py

@@ -12,7 +12,7 @@
     url='https://github.com/prediction-machines/tgym',
     packages=find_packages(),
     install_requires=[
-        'matplotlib==2.0.2'
+        #'matplotlib>=2.0.2'
     ],
     license="MIT license",
     zip_safe=False,

tests/envs/test_trading.py

@@ -1,6 +1,6 @@
 import numpy as np
-from tgym.envs import SpreadTrading
-from tgym.gens import WavySignal
+from tgym.envs.trading import SpreadTrading
+from tgym.gens.deterministic import WavySignal
 
 
 class TestSpreadTrading(object):

tests/gens/test_csvstream.py

@@ -1,8 +1,7 @@
 import csv
 
 import numpy as np
-from tgym.gens import CSVStreamer
-
+from tgym.gens.csvstream import CSVStreamer
 
 def test_csv_streamer():
     with open('test.csv', 'w+') as csvfile:

tests/gens/test_random.py

@@ -1,5 +1,5 @@
 import numpy as np
-from tgym.gens import AR1, RandomWalk
+from tgym.gens.random import AR1, RandomWalk
 
 
 def test_random_walk():