Add support for trading view data
parent
71f9b58803
commit
e577ce34d2
|
@ -0,0 +1,77 @@
|
|||
import numpy as np
|
||||
|
||||
|
||||
class AbstractTreeLearner:
|
||||
LEAF = -1
|
||||
NA = -1
|
||||
|
||||
def author(self):
|
||||
return 'felixm' # replace tb34 with your Georgia Tech username
|
||||
|
||||
def create_node(self, factor, split_value, left, right):
|
||||
return np.array([(factor, split_value, left, right), ],
|
||||
dtype='|i4, f4, i4, i4')
|
||||
|
||||
def query_point(self, point):
|
||||
node_index = 0
|
||||
while self.rel_tree[node_index][0] != self.LEAF:
|
||||
node = self.rel_tree[node_index]
|
||||
split_factor = node[0]
|
||||
split_value = node[1]
|
||||
if point[split_factor] <= split_value:
|
||||
# Recurse into left sub-tree.
|
||||
node_index += node[2]
|
||||
else:
|
||||
node_index += node[3]
|
||||
v = self.rel_tree[node_index][1]
|
||||
return v
|
||||
|
||||
def query(self, points):
|
||||
"""
|
||||
@summary: Estimate a set of test points given the model we built.
|
||||
@param points: should be a numpy array with each row corresponding to a specific query.
|
||||
@returns the estimated values according to the saved model.
|
||||
"""
|
||||
query_point = lambda p: self.query_point(p)
|
||||
r = np.apply_along_axis(query_point, 1, points)
|
||||
return r
|
||||
|
||||
def build_tree(self, xs, y):
|
||||
"""
|
||||
@summary: Build a decision tree from the training data.
|
||||
@param dataX: X values of data to add
|
||||
@param dataY: the Y training values
|
||||
"""
|
||||
assert(xs.shape[0] == y.shape[0])
|
||||
assert(xs.shape[0] > 0) # If this is 0 something went wrong.
|
||||
|
||||
if xs.shape[0] <= self.leaf_size:
|
||||
value = np.mean(y)
|
||||
if value < -0.2:
|
||||
value = -1
|
||||
elif value > 0.2:
|
||||
value = 1
|
||||
else:
|
||||
value = 0
|
||||
return self.create_node(self.LEAF, value, self.NA, self.NA)
|
||||
|
||||
if np.all(y[0] == y):
|
||||
return self.create_node(self.LEAF, y[0], self.NA, self.NA)
|
||||
|
||||
i, split_value = self.get_i_and_split_value(xs, y)
|
||||
select_l = xs[:, i] <= split_value
|
||||
select_r = xs[:, i] > split_value
|
||||
lt = self.build_tree(xs[select_l], y[select_l])
|
||||
rt = self.build_tree(xs[select_r], y[select_r])
|
||||
root = self.create_node(i, split_value, 1, lt.shape[0] + 1)
|
||||
root = np.concatenate([root, lt, rt])
|
||||
return root
|
||||
|
||||
def addEvidence(self, data_x, data_y):
|
||||
"""
|
||||
@summary: Add training data to learner
|
||||
@param dataX: X values of data to add
|
||||
@param dataY: the Y training values
|
||||
"""
|
||||
self.rel_tree = self.build_tree(data_x, data_y)
|
||||
|
|
@ -0,0 +1,36 @@
|
|||
import pandas as pd
|
||||
import util as ut
|
||||
import datetime as dt
|
||||
|
||||
|
||||
class BenchmarkStrategy:
|
||||
def __init__(self, verbose=False, impact=0.0, commission=0.0, units=1000):
|
||||
self.verbose = verbose
|
||||
self.impact = impact
|
||||
self.commission = commission
|
||||
self.units = units
|
||||
|
||||
def addEvidence(self, symbol=0, sd=0, ed=0, sv=0):
|
||||
"""Keep this so that API is valid."""
|
||||
pass
|
||||
|
||||
def testPolicy(self, symbol="IBM",
|
||||
sd=dt.datetime(2009, 1, 1),
|
||||
ed=dt.datetime(2010, 1, 1),
|
||||
sv=10000):
|
||||
"""Benchmark is to buy 1000 shares and hold."""
|
||||
dates = pd.date_range(sd, ed)
|
||||
prices = ut.get_data([symbol], dates, addSPY=False,
|
||||
colname='close', datecol='time')
|
||||
|
||||
orders = pd.DataFrame(index=prices.index)
|
||||
orders["Symbol"] = symbol
|
||||
orders["Order"] = ""
|
||||
orders["Shares"] = 0
|
||||
orders.iloc[0] = [symbol, "BUY", self.units]
|
||||
orders.iloc[-1] = [symbol, "SELL", -self.units]
|
||||
|
||||
if self.verbose:
|
||||
print(type(orders)) # it better be a DataFrame!
|
||||
print(orders)
|
||||
return orders
|
|
@ -0,0 +1,114 @@
|
|||
import datetime as dt
|
||||
import pandas as pd
|
||||
import util
|
||||
import indicators
|
||||
|
||||
|
||||
class ManualStrategy:
|
||||
def __init__(self, verbose=False, impact=0.0, commission=0.0):
|
||||
self.verbose = verbose
|
||||
self.impact = impact
|
||||
self.commission = commission
|
||||
|
||||
# this method should create a QLearner, and train it for trading
|
||||
def addEvidence(self, symbol="IBM",
|
||||
sd=dt.datetime(2008, 1, 1),
|
||||
ed=dt.datetime(2009, 1, 1),
|
||||
sv=10000):
|
||||
|
||||
# add your code to do learning here
|
||||
|
||||
# example usage of the old backward compatible util function
|
||||
syms = [symbol]
|
||||
dates = pd.date_range(sd, ed)
|
||||
prices_all = util.get_data(syms, dates) # automatically adds SPY
|
||||
prices = prices_all[syms] # only portfolio symbols
|
||||
# prices_SPY = prices_all['SPY'] # only SPY, for comparison later
|
||||
if self.verbose:
|
||||
print(prices)
|
||||
|
||||
# example use with new colname
|
||||
# automatically adds SPY
|
||||
volume_all = util.get_data(syms, dates, colname="Volume")
|
||||
volume = volume_all[syms] # only portfolio symbols
|
||||
# volume_SPY = volume_all['SPY'] # only SPY, for comparison later
|
||||
if self.verbose:
|
||||
print(volume)
|
||||
|
||||
def macd_strat(self, macd, orders):
|
||||
"""Strategy based on MACD cross."""
|
||||
|
||||
def strat(ser):
|
||||
m = macd.loc[ser.index]
|
||||
prev_macd, prev_signal, _ = m.iloc[0]
|
||||
cur_macd, cur_signal, _ = m.iloc[1]
|
||||
shares = 0
|
||||
if cur_macd < -1 and prev_macd < prev_signal \
|
||||
and cur_macd > cur_signal:
|
||||
if self.holding == 0:
|
||||
shares = 1000
|
||||
elif self.holding == -1000:
|
||||
shares = 2000
|
||||
elif cur_macd > 1 and prev_macd > prev_signal \
|
||||
and cur_macd < cur_signal:
|
||||
if self.holding == 0:
|
||||
shares = -1000
|
||||
elif self.holding == 1000:
|
||||
shares = -2000
|
||||
self.holding += shares
|
||||
return shares
|
||||
|
||||
orders['Shares'] = orders['Shares'].rolling(2).apply(strat)
|
||||
|
||||
def three_indicator_strat(self, macd, rsi, price_sma, orders):
|
||||
"""Strategy based on three indicators. Thresholds selected based on
|
||||
scatter plots."""
|
||||
def strat(row):
|
||||
shares = 0
|
||||
_, _, macd_diff = macd.loc[row.name]
|
||||
cur_rsi = rsi.loc[row.name][0]
|
||||
cur_price_sma = price_sma.loc[row.name][0]
|
||||
if self.holding == -1000 and cur_price_sma < 0.9:
|
||||
shares = 2000
|
||||
elif self.holding == 0 and cur_price_sma < 0.9:
|
||||
shares = 1000
|
||||
elif self.holding == -1000 and cur_rsi > 80:
|
||||
shares = 2000
|
||||
elif self.holding == 0 and cur_rsi > 80:
|
||||
shares = 1000
|
||||
elif self.holding == -1000 and macd_diff < -0.5:
|
||||
shares = 2000
|
||||
elif self.holding == 0 and macd_diff < -0.5:
|
||||
shares = 1000
|
||||
elif self.holding == 1000 and cur_price_sma > 1.1:
|
||||
shares = -2000
|
||||
elif self.holding == 0 and cur_price_sma > 1.1:
|
||||
shares = -1000
|
||||
self.holding += shares
|
||||
return shares
|
||||
|
||||
orders['Shares'] = orders.apply(strat, axis=1)
|
||||
|
||||
def testPolicy(self, symbol="IBM",
|
||||
sd=dt.datetime(2009, 1, 1),
|
||||
ed=dt.datetime(2010, 1, 1),
|
||||
sv=10000, macd_strat=False):
|
||||
|
||||
self.holding = 0
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
|
||||
orders = pd.DataFrame(index=df.index)
|
||||
orders["Symbol"] = symbol
|
||||
orders["Order"] = ""
|
||||
orders["Shares"] = 0
|
||||
|
||||
macd = indicators.macd(df, symbol)
|
||||
rsi = indicators.rsi(df, symbol)
|
||||
price_sma = indicators.price_sma(df, symbol, [8])
|
||||
|
||||
if macd_strat:
|
||||
self.macd_strat(macd, orders)
|
||||
else:
|
||||
self.three_indicator_strat(macd, rsi, price_sma, orders)
|
||||
return orders
|
|
@ -0,0 +1,170 @@
|
|||
import datetime as dt
|
||||
import pandas as pd
|
||||
import util
|
||||
import indicators
|
||||
from qlearning_robot.QLearner import QLearner as Learner
|
||||
from dataclasses import dataclass
|
||||
|
||||
@dataclass
|
||||
class Holding:
|
||||
cash: int
|
||||
shares: int
|
||||
equity: int
|
||||
|
||||
|
||||
class QLearner(object):
|
||||
|
||||
def __init__(self, verbose=False, impact=0.0, units=1000, commission=0.0, testing=False, n_bins=5):
|
||||
self.verbose = verbose
|
||||
self.impact = impact
|
||||
self.commission = commission
|
||||
self.testing = testing # Decides which type of order df to return.
|
||||
self.indicators = ['macd_diff', 'rsi', 'price_sma_8']
|
||||
self.n_bins = n_bins
|
||||
self.bins = {}
|
||||
self.num_states = self.get_num_states()
|
||||
self.num_actions = 3 # buy, sell, hold
|
||||
self.learner = Learner(self.num_states, self.num_actions)
|
||||
self.units = units
|
||||
|
||||
def row_to_state(self, holding, df_row):
|
||||
"""Transforms a row into a state value."""
|
||||
holding = (holding + self.units) // self.units
|
||||
assert(holding in [0, 1, 2])
|
||||
|
||||
# For each indicator that goes into the state the interval becomes
|
||||
# smaller based on how many bins the indicator has. The first
|
||||
# 'indicator' is the information about how many shares we are currently
|
||||
# holding. So for example, if I have 450 states then the intervall (aka
|
||||
# remaining_states) is 150 because there are three values for holding:
|
||||
# holding = 0 -> state = 0 * 150 = 0
|
||||
# holding = 1 -> state = 1 * 150 = 150
|
||||
# holding = 2 -> state = 2 * 150 = 300
|
||||
remaining_states = self.num_states // 3
|
||||
state = holding * remaining_states
|
||||
|
||||
for indicator in self.indicators:
|
||||
value = df_row[indicator]
|
||||
bin_n = self.indicator_value_to_bin(indicator, value)
|
||||
remaining_states //= self.n_bins
|
||||
state += bin_n * remaining_states
|
||||
return state
|
||||
|
||||
def indicator_value_to_bin(self, indicator, value):
|
||||
for i, upper_bound in enumerate(self.bins[indicator]):
|
||||
if value < upper_bound:
|
||||
return i
|
||||
return i + 1
|
||||
|
||||
def add_indicators(self, df, symbol):
|
||||
"""Add indicators for learning to DataFrame."""
|
||||
for indicator in self.indicators:
|
||||
if indicator == "macd_diff":
|
||||
indicators.macd(df, symbol)
|
||||
df.drop(columns=["macd", "macd_signal"], inplace=True)
|
||||
elif indicator == "rsi":
|
||||
indicators.rsi(df, symbol)
|
||||
elif indicator.startswith("price_sma_"):
|
||||
period = int(indicator.replace("price_sma_", ""))
|
||||
indicators.price_sma(df, symbol, [period])
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
df.dropna(inplace=True)
|
||||
|
||||
def bin_indicators(self, df):
|
||||
"""Create bins for indicators."""
|
||||
for indicator in self.indicators:
|
||||
ser, bins = pd.qcut(df[indicator], self.n_bins, retbins=True)
|
||||
self.bins[indicator] = bins[1:self.n_bins]
|
||||
|
||||
def get_num_states(self):
|
||||
"""Return the total num of states."""
|
||||
num_states = 3 # Three states holding (1000, 0, -1000)
|
||||
for _ in self.indicators:
|
||||
num_states *= self.n_bins
|
||||
return num_states
|
||||
|
||||
def handle_order(self, action, holding, adj_closing_price):
|
||||
shares = 0
|
||||
if action == 0: # buy
|
||||
if holding.shares == 0 or holding.shares == -self.units:
|
||||
shares = self.units
|
||||
elif action == 1: # sell
|
||||
if holding.shares== 0 or holding.shares == self.units:
|
||||
shares = -self.units
|
||||
elif action == 2: # hold
|
||||
shares = 0
|
||||
|
||||
cost = shares * adj_closing_price
|
||||
if shares != 0:
|
||||
# Charge commission and deduct impact penalty
|
||||
holding.cash -= self.commission
|
||||
holding.cash -= (self.impact * adj_closing_price * abs(shares))
|
||||
holding.cash -= cost
|
||||
holding.shares += shares
|
||||
|
||||
holding.equity = holding.cash + holding.shares * adj_closing_price
|
||||
|
||||
def get_reward(self, equity, new_equity):
|
||||
if new_equity > equity:
|
||||
return 1
|
||||
return -1
|
||||
|
||||
def train(self, df, symbol, sv):
|
||||
holding = Holding(sv, 0, sv)
|
||||
|
||||
row = df.iloc[0]
|
||||
state = self.row_to_state(holding.shares, row)
|
||||
action = self.learner.querysetstate(state)
|
||||
adj_closing_price = row[symbol]
|
||||
equity = holding.equity
|
||||
self.handle_order(action, holding, adj_closing_price)
|
||||
|
||||
for index, row in df.iloc[1:].iterrows():
|
||||
adj_closing_price = row[symbol]
|
||||
new_equity = holding.cash + holding.shares * adj_closing_price
|
||||
r = self.get_reward(equity, new_equity)
|
||||
s_prime = self.row_to_state(holding.shares, row)
|
||||
a = self.learner.query(s_prime, r)
|
||||
equity = new_equity
|
||||
self.handle_order(a, holding, adj_closing_price)
|
||||
if self.verbose:
|
||||
print(f"{holding=} {s_prime=} {r=} {a=}")
|
||||
|
||||
def addEvidence(self, symbol="IBM", sd=dt.datetime(2008, 1, 1), ed=dt.datetime(2009, 1, 1), sv=10000):
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
self.add_indicators(df, symbol)
|
||||
self.bin_indicators(df)
|
||||
|
||||
for _ in range(15):
|
||||
self.train(df, symbol, sv)
|
||||
|
||||
def testPolicy(self, symbol="IBM", sd=dt.datetime(2009, 1, 1), ed=dt.datetime(2010, 1, 1), sv=10000):
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
orders = pd.DataFrame(index=df.index)
|
||||
orders["Symbol"] = symbol
|
||||
orders["Order"] = ""
|
||||
orders["Shares"] = 0
|
||||
shares = orders["Shares"]
|
||||
self.add_indicators(df, symbol)
|
||||
holding = 0
|
||||
|
||||
for index, row in df.iterrows():
|
||||
state = self.row_to_state(holding, row)
|
||||
action = self.learner.querysetstate(state)
|
||||
|
||||
if action == 0: # buy
|
||||
if holding == 0 or holding == -self.units:
|
||||
holding += self.units
|
||||
orders.loc[index, "Shares"] = self.units
|
||||
elif action == 1: # sell
|
||||
if holding == 0 or holding == self.units:
|
||||
holding -= self.units
|
||||
orders.loc[index, "Shares"] = -self.units
|
||||
elif action == 2: # hold
|
||||
pass
|
||||
|
||||
if self.testing:
|
||||
return orders
|
||||
else:
|
||||
return orders[["Shares"]]
|
||||
|
|
@ -0,0 +1,30 @@
|
|||
import numpy as np
|
||||
from AbstractTreeLearner import AbstractTreeLearner
|
||||
|
||||
|
||||
class RTLearner(AbstractTreeLearner):
|
||||
|
||||
def __init__(self, leaf_size = 1, verbose = False):
|
||||
self.leaf_size = leaf_size
|
||||
self.verbose = verbose
|
||||
|
||||
def get_i_and_split_value(self, xs, y):
|
||||
"""
|
||||
@summary: Pick a random i and split value.
|
||||
|
||||
Make sure that not all X are the same for i and also pick
|
||||
different values to average the split_value from.
|
||||
"""
|
||||
i = np.random.randint(0, xs.shape[1])
|
||||
while np.all(xs[0,i] == xs[:,i]):
|
||||
i = np.random.randint(0, xs.shape[1])
|
||||
|
||||
# I don't know about the performance of this, but at least it
|
||||
# terminates reliably. If the two elements are the same something is
|
||||
# wrong.
|
||||
a = np.array(list(set(xs[:, i])))
|
||||
r1, r2 = np.random.choice(a, size = 2, replace = False)
|
||||
assert(r1 != r2)
|
||||
split_value = (r1 + r2) / 2.0
|
||||
return i, split_value
|
||||
|
|
@ -0,0 +1,106 @@
|
|||
import datetime as dt
|
||||
import pandas as pd
|
||||
import util
|
||||
import indicators
|
||||
from RTLearner import RTLearner
|
||||
|
||||
|
||||
class StrategyLearner(object):
|
||||
|
||||
def __init__(self, verbose=False, impact=0.0, commission=0.0, testing=False):
|
||||
self.verbose = verbose
|
||||
self.impact = impact
|
||||
self.commission = commission
|
||||
self.testing = testing
|
||||
|
||||
def _get_volume(self):
|
||||
"""For reference."""
|
||||
volume_all = ut.get_data(syms, dates, colname="Volume")
|
||||
volume = volume_all[syms] # only portfolio symbols
|
||||
# volume_SPY = volume_all['SPY'] # only SPY, for comparison later
|
||||
if self.verbose:
|
||||
print(volume)
|
||||
|
||||
def _add_indicators(self, df, symbol):
|
||||
"""Add indicators for learning to DataFrame."""
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
indicators.macd(df, symbol)
|
||||
indicators.rsi(df, symbol)
|
||||
indicators.price_sma(df, symbol, [8])
|
||||
indicators.price_delta(df, symbol, 3)
|
||||
df.dropna(inplace=True)
|
||||
|
||||
def addEvidence(self, symbol="IBM",
|
||||
sd=dt.datetime(2008, 1, 1),
|
||||
ed=dt.datetime(2009, 1, 1),
|
||||
sv=10000):
|
||||
|
||||
self.y_threshold = 0.2
|
||||
self.indicators = ['macd_diff', 'rsi', 'price_sma_8']
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
self._add_indicators(df, symbol)
|
||||
|
||||
def classify_y(row):
|
||||
if row > self.y_threshold:
|
||||
return 1
|
||||
elif row < -self.y_threshold:
|
||||
return -1
|
||||
else:
|
||||
pass
|
||||
return 0
|
||||
|
||||
def set_y_threshold(pct):
|
||||
if max(pct) < 0.2:
|
||||
self.y_threshold = 0.02
|
||||
|
||||
self.learner = RTLearner(leaf_size = 5)
|
||||
# self.learner = BagLearner(RTLearner, 3, {'leaf_size': 5})
|
||||
data_x = df[self.indicators].to_numpy()
|
||||
pct = df['pct_3']
|
||||
|
||||
# This is a hack to get a low enough buy/sell threshold for the
|
||||
# cyclic the test 'ML4T-220' where the max pct_3 is 0.0268.
|
||||
set_y_threshold(pct)
|
||||
y = pct.apply(classify_y)
|
||||
|
||||
self.learner.addEvidence(data_x, y.to_numpy())
|
||||
return y
|
||||
|
||||
def strat(self, data_y, orders):
|
||||
self.holding = 0
|
||||
|
||||
def strat(row):
|
||||
y = int(data_y.loc[row.name][0])
|
||||
shares = 0
|
||||
if self.holding == 0 and y == 1:
|
||||
shares = 1000
|
||||
elif self.holding == -1000 and y == 1:
|
||||
shares = 2000
|
||||
elif self.holding == 0 and y == -1:
|
||||
shares = -1000
|
||||
elif self.holding == 1000 and y == -1:
|
||||
shares = -2000
|
||||
self.holding += shares
|
||||
return shares
|
||||
|
||||
orders["Shares"] = orders.apply(strat, axis=1)
|
||||
|
||||
def testPolicy(self, symbol="IBM",
|
||||
sd=dt.datetime(2009, 1, 1),
|
||||
ed=dt.datetime(2010, 1, 1),
|
||||
sv=10000):
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
self._add_indicators(df, symbol)
|
||||
data_x = df[self.indicators].to_numpy()
|
||||
data_y = pd.DataFrame(index=df.index, data=self.learner.query(data_x))
|
||||
|
||||
orders = pd.DataFrame(index=df.index)
|
||||
orders["Symbol"] = symbol
|
||||
orders["Order"] = ""
|
||||
orders["Shares"] = 0
|
||||
self.strat(data_y, orders)
|
||||
if self.testing:
|
||||
return orders
|
||||
else:
|
||||
return orders[["Shares"]]
|
||||
|
|
@ -0,0 +1,237 @@
|
|||
import pandas as pd
|
||||
import datetime as dt
|
||||
import sys
|
||||
|
||||
import util
|
||||
import indicators
|
||||
import crypto_eval.marketsim as marketsim
|
||||
import matplotlib.pyplot as plt
|
||||
from matplotlib.widgets import MultiCursor
|
||||
from BenchmarkStrategy import BenchmarkStrategy
|
||||
from ManualStrategy import ManualStrategy
|
||||
from StrategyLearner import StrategyLearner
|
||||
from QLearner import QLearner
|
||||
|
||||
|
||||
def plot_indicators(symbol, df):
|
||||
fig, ax = plt.subplots(4, sharex=True)
|
||||
|
||||
price_sma = indicators.price_sma(df, symbol, [8])
|
||||
bb = indicators.bollinger_band(df, symbol)
|
||||
rsi = indicators.rsi(df, symbol)
|
||||
macd = indicators.macd(df, symbol).copy()
|
||||
|
||||
df[[symbol]].plot(ax=ax[0])
|
||||
bb.plot(ax=ax[0])
|
||||
price_sma.plot(ax=ax[1])
|
||||
macd.plot(ax=ax[2])
|
||||
rsi.plot(ax=ax[3])
|
||||
for a in ax.flat:
|
||||
a.grid()
|
||||
m = MultiCursor(fig.canvas, ax, color='r', lw=0.5)
|
||||
plt.show()
|
||||
sys.exit(0)
|
||||
|
||||
|
||||
def visualize_correlations(symbol, df):
|
||||
indicators.price_sma(df, symbol, [8, 21])
|
||||
indicators.price_delta(df, symbol, 5)
|
||||
indicators.price_delta(df, symbol, 3)
|
||||
indicators.price_delta(df, symbol, 1)
|
||||
indicators.macd(df, symbol)
|
||||
indicators.rsi(df, symbol)
|
||||
|
||||
# df = df[df['rsi'] > 80]
|
||||
fig, ax = plt.subplots(3, 2) # sharex=True)
|
||||
df.plot.scatter(x="price_sma_8", y="pct_5", ax=ax[0, 0])
|
||||
df.plot.scatter(x="price_sma_8", y="pct_3", ax=ax[1, 0])
|
||||
df.plot.scatter(x="price_sma_8", y="pct_1", ax=ax[2, 0])
|
||||
# df.plot.scatter(x="rsi", y="pct_5", ax=ax[0, 1])
|
||||
# df.plot.scatter(x="rsi", y="pct_3", ax=ax[1, 1])
|
||||
# df.plot.scatter(x="rsi", y="pct_1", ax=ax[2, 1])
|
||||
df.plot.scatter(x="macd_diff", y="pct_5", ax=ax[0, 1])
|
||||
df.plot.scatter(x="macd_diff", y="pct_3", ax=ax[1, 1])
|
||||
df.plot.scatter(x="macd_diff", y="pct_1", ax=ax[2, 1])
|
||||
|
||||
for a in ax.flat:
|
||||
a.grid()
|
||||
plt.show()
|
||||
sys.exit(0)
|
||||
|
||||
|
||||
def compare_manual_strategies(symbol, sv, sd, ed):
|
||||
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
|
||||
bs = BenchmarkStrategy()
|
||||
orders = bs.testPolicy(symbol, sd, ed, sv)
|
||||
df["Benchmark"] = marketsim.compute_portvals(orders, sv)
|
||||
df["Orders Benchmark"] = orders["Shares"]
|
||||
|
||||
ms = ManualStrategy()
|
||||
orders = ms.testPolicy(symbol, sd, ed, sv, macd_strat=True)
|
||||
df["MACD Strat"] = marketsim.compute_portvals(orders, sv)
|
||||
df["Orders MACD"] = orders["Shares"]
|
||||
# df["Holding Manual"] = orders["Shares"].cumsum()
|
||||
|
||||
orders = ms.testPolicy(symbol, sd, ed, sv)
|
||||
df["Three Strat"] = marketsim.compute_portvals(orders, sv)
|
||||
df["Orders Three"] = orders["Shares"]
|
||||
|
||||
fig, ax = plt.subplots(3, sharex=True)
|
||||
df[[symbol]].plot(ax=ax[0])
|
||||
df[["Benchmark", "MACD Strat", "Three Strat"]].plot(ax=ax[1])
|
||||
df[["Orders Benchmark", "Orders MACD", "Orders Three"]].plot(ax=ax[2])
|
||||
|
||||
for a in ax:
|
||||
a.grid()
|
||||
MultiCursor(fig.canvas, ax, color='r', lw=0.5)
|
||||
|
||||
# plt.show()
|
||||
fig.set_size_inches(10, 8, forward=True)
|
||||
plt.savefig('figure_1.png', dpi=fig.dpi)
|
||||
|
||||
|
||||
def compare_all_strategies(symbol, sv, sd, ed):
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
normalize = indicators.normalize
|
||||
|
||||
bs = BenchmarkStrategy()
|
||||
orders = bs.testPolicy(symbol, sd, ed, sv)
|
||||
df["Benchmark"] = normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders Benchmark"] = orders["Shares"]
|
||||
|
||||
ms = ManualStrategy()
|
||||
orders = ms.testPolicy(symbol, sd, ed, sv)
|
||||
df["Manual"] = normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders Manual"] = orders["Shares"]
|
||||
|
||||
sl = StrategyLearner(testing=True)
|
||||
sl.addEvidence(symbol, sd, ed, sv)
|
||||
orders = sl.testPolicy(symbol, sd, ed, sv)
|
||||
df["Strategy"] = normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders Strategy"] = orders["Shares"]
|
||||
|
||||
fig, ax = plt.subplots(3, sharex=True)
|
||||
df[[symbol]].plot(ax=ax[0])
|
||||
df[["Benchmark", "Manual", "Strategy"]].plot(ax=ax[1])
|
||||
df[["Orders Benchmark", "Orders Manual", "Orders Strategy"]].plot(ax=ax[2])
|
||||
|
||||
for a in ax:
|
||||
a.grid()
|
||||
MultiCursor(fig.canvas, ax, color='r', lw=0.5)
|
||||
|
||||
# plt.show()
|
||||
fig.set_size_inches(10, 8, forward=True)
|
||||
plt.savefig('figure_2.png', dpi=fig.dpi)
|
||||
|
||||
|
||||
def compare_number_trades():
|
||||
symbol = "JPM"
|
||||
sv = 10000
|
||||
sd = dt.datetime(2008, 1, 1) # in-sample
|
||||
ed = dt.datetime(2009, 12, 31) # in-sample
|
||||
|
||||
df = util.get_data([symbol], pd.date_range(sd, ed))
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
|
||||
print(f"| commission | n_orders |")
|
||||
print(f"-------------------------")
|
||||
for commission in [0, 9.95, 20, 50, 100]:
|
||||
ql = QLearner(testing=True, commission=commission, impact=0.005)
|
||||
ql.addEvidence(symbol, sd, ed, sv)
|
||||
orders = ql.testPolicy(symbol, sd, ed, sv)
|
||||
n_orders = orders[orders["Shares"] != 0].shape[0]
|
||||
print(f"| {commission} | {n_orders} |")
|
||||
|
||||
def compare_q_learners():
|
||||
symbol = "JPM"
|
||||
sv = 10000
|
||||
sd = dt.datetime(2008, 1, 1) # in-sample
|
||||
ed = dt.datetime(2009, 12, 31) # in-sample
|
||||
sd_out = dt.datetime(2010, 1, 1) # out-sample
|
||||
ed_out = dt.datetime(2011, 12, 31) # out-sample
|
||||
|
||||
df = util.get_data([symbol], pd.date_range(sd_out, ed_out))
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
|
||||
bs = BenchmarkStrategy()
|
||||
orders = bs.testPolicy(symbol, sd_out, ed_out, sv)
|
||||
df["Benchmark"] = indicators.normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders Benchmark"] = orders["Shares"]
|
||||
|
||||
ql = QLearner(testing=True, verbose=False)
|
||||
ql.addEvidence(symbol, sd, ed, sv)
|
||||
orders = ql.testPolicy(symbol, sd_out, ed_out, sv)
|
||||
df["QL 5"] = indicators.normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders QL 5"] = orders["Shares"]
|
||||
|
||||
ql = QLearner(testing=True, verbose=False, n_bins=4)
|
||||
ql.addEvidence(symbol, sd, ed, sv)
|
||||
orders = ql.testPolicy(symbol, sd_out, ed_out, sv)
|
||||
df["QL 4"] = indicators.normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders QL 4"] = orders["Shares"]
|
||||
|
||||
fig, ax = plt.subplots(3, sharex=True)
|
||||
df[[symbol]].plot(ax=ax[0])
|
||||
df[["Benchmark", "QL 5", "QL 4"]].plot(ax=ax[1])
|
||||
df[["Orders Benchmark", "Orders QL 5", "Orders QL 4"]].plot(ax=ax[2])
|
||||
|
||||
for a in ax:
|
||||
a.grid()
|
||||
m = MultiCursor(fig.canvas, ax, color='r', lw=0.5)
|
||||
fig.set_size_inches(10, 8, forward=True)
|
||||
plt.savefig('figure_4.png', dpi=fig.dpi)
|
||||
sys.exit(0)
|
||||
|
||||
|
||||
def experiment1(create_report=False):
|
||||
symbol = "COINBASE_BTCUSD_1D"
|
||||
sv = 10000
|
||||
sd = dt.datetime(2020, 1, 1) # in-sample
|
||||
ed = dt.datetime(2020, 12, 31) # in-sample
|
||||
sd_out = dt.datetime(2020, 1, 1) # out-sample
|
||||
ed_out = dt.datetime(2020, 12, 31) # out-sample
|
||||
|
||||
df = util.get_data([symbol], pd.date_range(sd_out, ed_out), addSPY=True)
|
||||
|
||||
# if create_report:
|
||||
# compare_manual_strategies(symbol, sv, sd, ed)
|
||||
# compare_all_strategies(symbol, sv, sd, ed)
|
||||
# sys.exit(0)
|
||||
|
||||
# visualize_correlations(symbol, df)
|
||||
# plot_indicators(symbol, df)
|
||||
# compare_number_trades(symbol, sv, sd, ed)
|
||||
# compare_q_learners()
|
||||
# return
|
||||
|
||||
bs = BenchmarkStrategy(units=1)
|
||||
orders = bs.testPolicy(symbol, sd_out, ed_out, sv)
|
||||
pvs = marketsim.compute_portvals(orders, start_val=sv)
|
||||
df["Benchmark"] = indicators.normalize(pvs)
|
||||
df["Orders Benchmark"] = orders["Shares"]
|
||||
|
||||
ql = QLearner(testing=True, verbose=False, units=1)
|
||||
ql.addEvidence(symbol, sd, ed, sv)
|
||||
orders = ql.testPolicy(symbol, sd_out, ed_out, sv)
|
||||
df["QL"] = indicators.normalize(marketsim.compute_portvals(orders, sv))
|
||||
df["Orders QL"] = orders["Shares"]
|
||||
|
||||
fig, ax = plt.subplots(3, sharex=True)
|
||||
df[[symbol]].plot(ax=ax[0])
|
||||
df[["Benchmark", "QL"]].plot(ax=ax[1])
|
||||
df[["Orders Benchmark", "Orders QL"]].plot(ax=ax[2])
|
||||
|
||||
for a in ax:
|
||||
a.grid()
|
||||
m = MultiCursor(fig.canvas, ax, color='r', lw=0.5)
|
||||
plt.show()
|
||||
# fig.set_size_inches(10, 8, forward=True)
|
||||
# plt.savefig('figure_4.png', dpi=fig.dpi)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
experiment1()
|
|
@ -0,0 +1,8 @@
|
|||
import experiment1
|
||||
|
||||
def experiment2():
|
||||
experiment1.compare_number_trades()
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
experiment2()
|
|
@ -0,0 +1,339 @@
|
|||
"""MC3-P3: Strategy Learner - grading script.
|
||||
|
||||
Usage:
|
||||
- Switch to a student feedback directory first (will write "points.txt" and "comments.txt" in pwd).
|
||||
- Run this script with both ml4t/ and student solution in PYTHONPATH, e.g.:
|
||||
PYTHONPATH=ml4t:MC1-P2/jdoe7 python ml4t/mc2_p1_grading/grade_marketsim.py
|
||||
|
||||
Copyright 2017, Georgia Tech Research Corporation
|
||||
Atlanta, Georgia 30332-0415
|
||||
All Rights Reserved
|
||||
|
||||
Template code for CS 4646/7646
|
||||
|
||||
Georgia Tech asserts copyright ownership of this template and all derivative
|
||||
works, including solutions to the projects assigned in this course. Students
|
||||
and other users of this template code are advised not to share it with others
|
||||
or to make it available on publicly viewable websites including repositories
|
||||
such as github and gitlab. This copyright statement should not be removed
|
||||
or edited.
|
||||
|
||||
We do grant permission to share solutions privately with non-students such
|
||||
as potential employers. However, sharing with other current or future
|
||||
students of CS 7646 is prohibited and subject to being investigated as a
|
||||
GT honor code violation.
|
||||
|
||||
-----do not edit anything above this line---
|
||||
|
||||
Student Name: Tucker Balch (replace with your name)
|
||||
GT User ID: tb34 (replace with your User ID)
|
||||
GT ID: 900897987 (replace with your GT ID)
|
||||
"""
|
||||
|
||||
import pytest
|
||||
from grading.grading import grader, GradeResult, run_with_timeout, IncorrectOutput
|
||||
|
||||
import os
|
||||
import sys
|
||||
import traceback as tb
|
||||
|
||||
import datetime as dt
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
from collections import namedtuple
|
||||
|
||||
import time
|
||||
import util
|
||||
import random
|
||||
|
||||
# Test cases
|
||||
StrategyTestCase = namedtuple('Strategy', ['description','insample_args','outsample_args','benchmark_type','benchmark','impact','train_time','test_time','max_time','seed'])
|
||||
strategy_test_cases = [
|
||||
StrategyTestCase(
|
||||
description="ML4T-220",
|
||||
insample_args=dict(symbol="ML4T-220",sd=dt.datetime(2008,1,1),ed=dt.datetime(2009,12,31),sv=100000),
|
||||
outsample_args=dict(symbol="ML4T-220",sd=dt.datetime(2010,1,1),ed=dt.datetime(2011,12,31),sv=100000),
|
||||
benchmark_type='clean',
|
||||
benchmark=1.0, #benchmark updated Apr 24 2017
|
||||
impact=0.0,
|
||||
train_time=25,
|
||||
test_time=5,
|
||||
max_time=60,
|
||||
seed=1481090000
|
||||
),
|
||||
StrategyTestCase(
|
||||
description="AAPL",
|
||||
insample_args=dict(symbol="AAPL",sd=dt.datetime(2008,1,1),ed=dt.datetime(2009,12,31),sv=100000),
|
||||
outsample_args=dict(symbol="AAPL",sd=dt.datetime(2010,1,1),ed=dt.datetime(2011,12,31),sv=100000),
|
||||
benchmark_type='stock',
|
||||
benchmark=0.1581999999999999, #benchmark computed Nov 22 2017
|
||||
impact=0.0,
|
||||
train_time=25,
|
||||
test_time=5,
|
||||
max_time=60,
|
||||
seed=1481090000
|
||||
),
|
||||
StrategyTestCase(
|
||||
description="SINE_FAST_NOISE",
|
||||
insample_args=dict(symbol="SINE_FAST_NOISE",sd=dt.datetime(2008,1,1),ed=dt.datetime(2009,12,31),sv=100000),
|
||||
outsample_args=dict(symbol="SINE_FAST_NOISE",sd=dt.datetime(2010,1,1),ed=dt.datetime(2011,12,31),sv=100000),
|
||||
benchmark_type='noisy',
|
||||
benchmark=2.0, #benchmark updated Apr 24 2017
|
||||
impact=0.0,
|
||||
train_time=25,
|
||||
test_time=5,
|
||||
max_time=60,
|
||||
seed=1481090000
|
||||
),
|
||||
StrategyTestCase(
|
||||
description="UNH - In sample",
|
||||
insample_args=dict(symbol="UNH",sd=dt.datetime(2008,1,1),ed=dt.datetime(2009,12,31),sv=100000),
|
||||
outsample_args=dict(symbol="UNH",sd=dt.datetime(2010,1,1),ed=dt.datetime(2011,12,31),sv=100000),
|
||||
benchmark_type='stock',
|
||||
benchmark= -0.25239999999999996, #benchmark computed Nov 22 2017
|
||||
impact=0.0,
|
||||
train_time=25,
|
||||
test_time=5,
|
||||
max_time=60,
|
||||
seed=1481090000
|
||||
),
|
||||
]
|
||||
|
||||
max_points = 60.0
|
||||
html_pre_block = True # surround comments with HTML <pre> tag (for T-Square comments field)
|
||||
|
||||
MAX_HOLDINGS = 1000
|
||||
|
||||
# Test functon(s)
|
||||
@pytest.mark.parametrize("description, insample_args, outsample_args, benchmark_type, benchmark, impact, train_time, test_time, max_time, seed", strategy_test_cases)
|
||||
def test_strategy(description, insample_args, outsample_args, benchmark_type, benchmark, impact, train_time, test_time, max_time, seed, grader):
|
||||
"""Test StrategyLearner.
|
||||
|
||||
Requires test description, insample args (dict), outsample args (dict), benchmark_type (str), benchmark (float)
|
||||
max time (seconds), points for this test case (int), random seed (long), and a grader fixture.
|
||||
"""
|
||||
points_earned = 0.0 # initialize points for this test case
|
||||
try:
|
||||
incorrect = True
|
||||
if not 'StrategyLearner' in globals():
|
||||
import importlib
|
||||
m = importlib.import_module('StrategyLearner')
|
||||
globals()['StrategyLearner'] = m
|
||||
outsample_cr_to_beat = None
|
||||
if benchmark_type == 'clean':
|
||||
outsample_cr_to_beat = benchmark
|
||||
def timeoutwrapper_strategylearner():
|
||||
#Set fixed seed for repetability
|
||||
np.random.seed(seed)
|
||||
random.seed(seed)
|
||||
learner = StrategyLearner.StrategyLearner(verbose=False,impact=impact)
|
||||
tmp = time.time()
|
||||
learner.addEvidence(**insample_args)
|
||||
train_t = time.time()-tmp
|
||||
tmp = time.time()
|
||||
insample_trades_1 = learner.testPolicy(**insample_args)
|
||||
test_t = time.time()-tmp
|
||||
insample_trades_2 = learner.testPolicy(**insample_args)
|
||||
tmp = time.time()
|
||||
outsample_trades = learner.testPolicy(**outsample_args)
|
||||
out_test_t = time.time()-tmp
|
||||
return insample_trades_1, insample_trades_2, outsample_trades, train_t, test_t, out_test_t
|
||||
msgs = []
|
||||
in_trades_1, in_trades_2, out_trades, train_t, test_t, out_test_t = run_with_timeout(timeoutwrapper_strategylearner,max_time,(),{})
|
||||
incorrect = False
|
||||
if len(in_trades_1.shape)!=2 or in_trades_1.shape[1]!=1:
|
||||
incorrect=True
|
||||
msgs.append(" First insample trades DF has invalid shape: {}".format(in_trades_1.shape))
|
||||
elif len(in_trades_2.shape)!=2 or in_trades_2.shape[1]!=1:
|
||||
incorrect=True
|
||||
msgs.append(" Second insample trades DF has invalid shape: {}".format(in_trades_2.shape))
|
||||
elif len(out_trades.shape)!=2 or out_trades.shape[1]!=1:
|
||||
incorrect=True
|
||||
msgs.append(" Out-of-sample trades DF has invalid shape: {}".format(out_trades.shape))
|
||||
else:
|
||||
tmp_csum=0.0
|
||||
for date,trade in in_trades_1.iterrows():
|
||||
tmp_csum+= trade.iloc[0]
|
||||
if (trade.iloc[0]!=0) and\
|
||||
(trade.abs().iloc[0]!=MAX_HOLDINGS) and\
|
||||
(trade.abs().iloc[0]!=2*MAX_HOLDINGS):
|
||||
incorrect=True
|
||||
msgs.append(" illegal trade in first insample DF. abs(trade) not one of ({},{},{}).\n Date {}, Trade {}".format(0,MAX_HOLDINGS,2*MAX_HOLDINGS,date,trade))
|
||||
break
|
||||
elif abs(tmp_csum)>MAX_HOLDINGS:
|
||||
incorrect=True
|
||||
msgs.append(" holdings more than {} long or short in first insample DF. Date {}, Trade {}".format(MAX_HOLDINGS,date,trade))
|
||||
break
|
||||
tmp_csum=0.0
|
||||
for date,trade in in_trades_2.iterrows():
|
||||
tmp_csum+= trade.iloc[0]
|
||||
if (trade.iloc[0]!=0) and\
|
||||
(trade.abs().iloc[0]!=MAX_HOLDINGS) and\
|
||||
(trade.abs().iloc[0]!=2*MAX_HOLDINGS):
|
||||
incorrect=True
|
||||
msgs.append(" illegal trade in second insample DF. abs(trade) not one of ({},{},{}).\n Date {}, Trade {}".format(0,MAX_HOLDINGS,2*MAX_HOLDINGS,date,trade))
|
||||
break
|
||||
elif abs(tmp_csum)>MAX_HOLDINGS:
|
||||
incorrect=True
|
||||
msgs.append(" holdings more than {} long or short in second insample DF. Date {}, Trade {}".format(MAX_HOLDINGS,date,trade))
|
||||
break
|
||||
tmp_csum=0.0
|
||||
for date,trade in out_trades.iterrows():
|
||||
tmp_csum+= trade.iloc[0]
|
||||
if (trade.iloc[0]!=0) and\
|
||||
(trade.abs().iloc[0]!=MAX_HOLDINGS) and\
|
||||
(trade.abs().iloc[0]!=2*MAX_HOLDINGS):
|
||||
incorrect=True
|
||||
msgs.append(" illegal trade in out-of-sample DF. abs(trade) not one of ({},{},{}).\n Date {}, Trade {}".format(0,MAX_HOLDINGS,2*MAX_HOLDINGS,date,trade))
|
||||
break
|
||||
elif abs(tmp_csum)>MAX_HOLDINGS:
|
||||
incorrect=True
|
||||
msgs.append(" holdings more than {} long or short in out-of-sample DF. Date {}, Trade {}".format(MAX_HOLDINGS,date,trade))
|
||||
break
|
||||
# if (((in_trades_1.abs()!=0) & (in_trades_1.abs()!=MAX_HOLDINGS) & (in_trades_1.abs()!=2*MAX_HOLDINGS)).any().any() or\
|
||||
# ((in_trades_2.abs()!=0) & (in_trades_2.abs()!=MAX_HOLDINGS) & (in_trades_2.abs()!=2*MAX_HOLDINGS)).any().any() or\
|
||||
# ((out_trades.abs()!=0) & (out_trades.abs()!=MAX_HOLDINGS) & (out_trades.abs()!=2*MAX_HOLDINGS)).any().any()):
|
||||
# incorrect = True
|
||||
# msgs.append(" illegal trade. abs(trades) not one of ({},{},{})".format(0,MAX_HOLDINGS,2*MAX_HOLDINGS))
|
||||
# if ((in_trades_1.cumsum().abs()>MAX_HOLDINGS).any()[0]) or ((in_trades_2.cumsum().abs()>MAX_HOLDINGS).any()[0]) or ((out_trades.cumsum().abs()>MAX_HOLDINGS).any()[0]):
|
||||
# incorrect = True
|
||||
# msgs.append(" holdings more than {} long or short".format(MAX_HOLDINGS))
|
||||
if not(incorrect):
|
||||
if train_t>train_time:
|
||||
incorrect=True
|
||||
msgs.append(" addEvidence() took {} seconds, max allowed {}".format(train_t,train_time))
|
||||
else:
|
||||
points_earned += 1.0
|
||||
if test_t > test_time:
|
||||
incorrect = True
|
||||
msgs.append(" testPolicy() took {} seconds, max allowed {}".format(test_t,test_time))
|
||||
else:
|
||||
points_earned += 2.0
|
||||
if not((in_trades_1 == in_trades_2).all()[0]):
|
||||
incorrect = True
|
||||
mismatches = in_trades_1.join(in_trades_2,how='outer',lsuffix='1',rsuffix='2')
|
||||
mismatches = mismatches[mismatches.iloc[:,0]!=mismatches.iloc[:,1]]
|
||||
msgs.append(" consecutive calls to testPolicy() with same input did not produce same output:")
|
||||
msgs.append(" Mismatched trades:\n {}".format(mismatches))
|
||||
else:
|
||||
points_earned += 2.0
|
||||
student_insample_cr = evalPolicy2(insample_args['symbol'],in_trades_1,insample_args['sv'],insample_args['sd'],insample_args['ed'],market_impact=impact,commission_cost=0.0)
|
||||
student_outsample_cr = evalPolicy2(outsample_args['symbol'],out_trades, outsample_args['sv'],outsample_args['sd'],outsample_args['ed'],market_impact=impact,commission_cost=0.0)
|
||||
if student_insample_cr <= benchmark:
|
||||
incorrect = True
|
||||
msgs.append(" in-sample return ({}) did not beat benchmark ({})".format(student_insample_cr,benchmark))
|
||||
else:
|
||||
points_earned += 5.0
|
||||
if outsample_cr_to_beat is None:
|
||||
if out_test_t > test_time:
|
||||
incorrect = True
|
||||
msgs.append(" out-sample took {} seconds, max of {}".format(out_test_t,test_time))
|
||||
else:
|
||||
points_earned += 5.0
|
||||
else:
|
||||
if student_outsample_cr < outsample_cr_to_beat:
|
||||
incorrect = True
|
||||
msgs.append(" out-sample return ({}) did not beat benchmark ({})".format(student_outsample_cr,outsample_cr_to_beat))
|
||||
else:
|
||||
points_earned += 5.0
|
||||
if incorrect:
|
||||
inputs_str = " insample_args: {}\n" \
|
||||
" outsample_args: {}\n" \
|
||||
" benchmark_type: {}\n" \
|
||||
" benchmark: {}\n" \
|
||||
" train_time: {}\n" \
|
||||
" test_time: {}\n" \
|
||||
" max_time: {}\n" \
|
||||
" seed: {}\n".format(insample_args, outsample_args, benchmark_type, benchmark, train_time, test_time, max_time,seed)
|
||||
raise IncorrectOutput("Test failed on one or more output criteria.\n Inputs:\n{}\n Failures:\n{}".format(inputs_str, "\n".join(msgs)))
|
||||
except Exception as e:
|
||||
# Test result: failed
|
||||
msg = "Test case description: {}\n".format(description)
|
||||
|
||||
# Generate a filtered stacktrace, only showing erroneous lines in student file(s)
|
||||
tb_list = tb.extract_tb(sys.exc_info()[2])
|
||||
for i in range(len(tb_list)):
|
||||
row = tb_list[i]
|
||||
tb_list[i] = (os.path.basename(row[0]), row[1], row[2], row[3]) # show only filename instead of long absolute path
|
||||
# tb_list = [row for row in tb_list if row[0] in ['QLearner.py','StrategyLearner.py']]
|
||||
if tb_list:
|
||||
msg += "Traceback:\n"
|
||||
msg += ''.join(tb.format_list(tb_list)) # contains newlines
|
||||
elif 'grading_traceback' in dir(e):
|
||||
msg += "Traceback:\n"
|
||||
msg += ''.join(tb.format_list(e.grading_traceback))
|
||||
msg += "{}: {}".format(e.__class__.__name__, str(e))
|
||||
|
||||
# Report failure result to grader, with stacktrace
|
||||
grader.add_result(GradeResult(outcome='failed', points=points_earned, msg=msg))
|
||||
raise
|
||||
else:
|
||||
# Test result: passed (no exceptions)
|
||||
grader.add_result(GradeResult(outcome='passed', points=points_earned, msg=None))
|
||||
|
||||
def compute_benchmark(sd,ed,sv,symbol,market_impact,commission_cost,max_holdings):
|
||||
date_idx = util.get_data([symbol,],pd.date_range(sd,ed)).index
|
||||
orders = pd.DataFrame(index=date_idx)
|
||||
orders['orders'] = 0; orders['orders'][0] = max_holdings; orders['orders'][-1] = -max_holdings
|
||||
return evalPolicy2(symbol,orders,sv,sd,ed,market_impact,commission_cost)
|
||||
|
||||
def evalPolicy(student_trades,sym_prices,startval):
|
||||
ending_cash = startval - student_trades.mul(sym_prices,axis=0).sum()
|
||||
ending_stocks = student_trades.sum()*sym_prices.iloc[-1]
|
||||
return float((ending_cash+ending_stocks)/startval)-1.0
|
||||
|
||||
def evalPolicy2(symbol, student_trades, startval, sd, ed, market_impact,commission_cost):
|
||||
orders_df = pd.DataFrame(columns=['Shares','Order','Symbol'])
|
||||
for row_idx in student_trades.index:
|
||||
nshares = student_trades.loc[row_idx][0]
|
||||
if nshares == 0:
|
||||
continue
|
||||
order = 'sell' if nshares < 0 else 'buy'
|
||||
new_row = pd.DataFrame([[abs(nshares),order,symbol],],columns=['Shares','Order','Symbol'],index=[row_idx,])
|
||||
orders_df = orders_df.append(new_row)
|
||||
portvals = compute_portvals(orders_df, sd, ed, startval,market_impact,commission_cost)
|
||||
return float(portvals[-1]/portvals[0])-1
|
||||
|
||||
def compute_portvals(orders_df, start_date, end_date, startval, market_impact=0.0, commission_cost=0.0):
|
||||
"""Simulate the market for the given date range and orders file."""
|
||||
symbols = []
|
||||
orders = []
|
||||
orders_df = orders_df.sort_index()
|
||||
for date, order in orders_df.iterrows():
|
||||
shares = order['Shares']
|
||||
action = order['Order']
|
||||
symbol = order['Symbol']
|
||||
if action.lower() == 'sell':
|
||||
shares *= -1
|
||||
order = (date, symbol, shares)
|
||||
orders.append(order)
|
||||
symbols.append(symbol)
|
||||
symbols = list(set(symbols))
|
||||
dates = pd.date_range(start_date, end_date)
|
||||
prices_all = util.get_data(symbols, dates)
|
||||
prices = prices_all[symbols]
|
||||
prices = prices.fillna(method='ffill').fillna(method='bfill')
|
||||
prices['_CASH'] = 1.0
|
||||
trades = pd.DataFrame(index=prices.index, columns=symbols)
|
||||
trades = trades.fillna(0)
|
||||
cash = pd.Series(index=prices.index)
|
||||
cash = cash.fillna(0)
|
||||
cash.iloc[0] = startval
|
||||
for date, symbol, shares in orders:
|
||||
price = prices[symbol][date]
|
||||
val = shares * price
|
||||
# transaction cost model
|
||||
val += commission_cost + (pd.np.abs(shares)*price*market_impact)
|
||||
positions = prices.loc[date] * trades.sum()
|
||||
totalcash = cash.sum()
|
||||
if (date < prices.index.min()) or (date > prices.index.max()):
|
||||
continue
|
||||
trades[symbol][date] += shares
|
||||
cash[date] -= val
|
||||
trades['_CASH'] = cash
|
||||
holdings = trades.cumsum()
|
||||
df_portvals = (prices * holdings).sum(axis=1)
|
||||
return df_portvals
|
||||
|
||||
if __name__ == "__main__":
|
||||
pytest.main(["-s", __file__])
|
|
@ -0,0 +1,140 @@
|
|||
import pandas as pd
|
||||
import datetime as dt
|
||||
import matplotlib.pyplot as plt
|
||||
from util import get_data
|
||||
|
||||
|
||||
def author():
|
||||
return "felixm"
|
||||
|
||||
|
||||
def normalize(timeseries):
|
||||
return timeseries / timeseries.iloc[0]
|
||||
|
||||
|
||||
def bollinger_band(df, symbol, period=20, m=2):
|
||||
boll_sma = df[symbol].rolling(period).mean()
|
||||
std = df[symbol].rolling(period).std()
|
||||
boll_up = boll_sma + m * std
|
||||
boll_lo = boll_sma - m * std
|
||||
key_sma, key_up, key_lo = "boll_sma", "boll_up", "boll_lo"
|
||||
df[key_sma] = boll_sma
|
||||
df[key_up] = boll_up
|
||||
df[key_lo] = boll_lo
|
||||
return df[[key_sma, key_up, key_lo]]
|
||||
|
||||
|
||||
def sma(df, symbol, period):
|
||||
"""Adds SMA for one or multiple periods to df and returns SMAs"""
|
||||
if type(period) is int:
|
||||
period = [period]
|
||||
keys = []
|
||||
for p in period:
|
||||
key = f"sma_{p}"
|
||||
df[key] = df[symbol].rolling(p).mean()
|
||||
keys.append(key)
|
||||
return df[keys]
|
||||
|
||||
|
||||
def ema(df, symbol, period):
|
||||
"""Adds EMA for one or multiple periods to df and returns EMAs"""
|
||||
if type(period) is int:
|
||||
period = [period]
|
||||
keys = []
|
||||
for p in period:
|
||||
key = f"ema_{p}"
|
||||
df[key] = df[symbol].ewm(span=p).mean()
|
||||
keys.append(key)
|
||||
return df[keys]
|
||||
|
||||
|
||||
def price_sma(df, symbol, period):
|
||||
"""Calculates SMA and adds new column price divided by SMA to the df."""
|
||||
if type(period) is int:
|
||||
period = [period]
|
||||
keys = []
|
||||
for p in period:
|
||||
key = f"price_sma_{p}"
|
||||
sma = df[symbol].rolling(p).mean()
|
||||
df[key] = df[symbol] / sma
|
||||
keys.append(key)
|
||||
return df[keys]
|
||||
|
||||
|
||||
def rsi(df, symbol, period=14):
|
||||
"""Calculates relative strength index over given period."""
|
||||
|
||||
def rsi(x):
|
||||
pct = x.pct_change()
|
||||
avg_gain = pct[pct > 0].mean()
|
||||
avg_loss = pct[pct <= 0].abs().mean()
|
||||
rsi = 100 - (100 /
|
||||
(1 + ((avg_gain / period) /
|
||||
(avg_loss / period))))
|
||||
return rsi
|
||||
|
||||
key = "rsi"
|
||||
# Add one to get 'period' price changes (first change is nan).
|
||||
period += 1
|
||||
df[key] = df[symbol].rolling(period).apply(rsi)
|
||||
return df[[key]]
|
||||
|
||||
|
||||
def macd(df, symbol):
|
||||
macd = df[symbol].ewm(span=12).mean() - df[symbol].ewm(span=26).mean()
|
||||
k1 = "macd"
|
||||
k2 = "macd_signal"
|
||||
k3 = "macd_diff"
|
||||
df[k1] = macd
|
||||
df[k2] = macd.rolling(9).mean()
|
||||
df[k3] = df[k1] - df[k2]
|
||||
return df[[k1, k2, k3]]
|
||||
|
||||
|
||||
def price_delta(df, symbol, period=1):
|
||||
"""Calculate percentage change for period."""
|
||||
k = f"pct_{period}"
|
||||
df[k] = df[symbol].pct_change(periods=period)
|
||||
df[k] = df[k].shift(-period)
|
||||
return df[k]
|
||||
|
||||
|
||||
def test_indicators():
|
||||
symbol = "JPM"
|
||||
|
||||
sd = dt.datetime(2008, 1, 1)
|
||||
ed = dt.datetime(2009, 12, 31)
|
||||
df = get_data([symbol], pd.date_range(sd, ed))
|
||||
df.drop(columns=["SPY"], inplace=True)
|
||||
df_orig = df.copy()
|
||||
# df = normalize(df)
|
||||
|
||||
sma(df, symbol, 21)
|
||||
ema(df, symbol, 21)
|
||||
df.plot(title="21 SMA and EMA")
|
||||
plt.savefig('figure_1.png')
|
||||
|
||||
df = df_orig.copy()
|
||||
sma(df, symbol, 8)
|
||||
price_sma(df, symbol, 8)
|
||||
df.plot(title="SMA and price / SMA", subplots=True)
|
||||
plt.savefig('figure_2.png')
|
||||
|
||||
df = df_orig.copy()
|
||||
bollinger_band(df, symbol)
|
||||
df.plot(title="Bollinger Band")
|
||||
plt.savefig('figure_3.png')
|
||||
|
||||
df = df_orig.copy()
|
||||
rsi(df, symbol)
|
||||
fig, axes = plt.subplots(nrows=2, sharex=True)
|
||||
df[symbol].plot(ax=axes[0], title="JPM price action")
|
||||
df["JPM-rsi(14)"].plot(ax=axes[1], title="RSI")
|
||||
plt.savefig('figure_4.png')
|
||||
|
||||
df = df_orig.copy()
|
||||
macd(df, symbol)
|
||||
fig, axes = plt.subplots(nrows=2, sharex=True)
|
||||
df[symbol].plot(ax=axes[0], title="JPM price action")
|
||||
df[["JPM-macd", "JPM-macd-signal"]].plot(ax=axes[1])
|
||||
plt.savefig('figure_5.png')
|
|
@ -0,0 +1,179 @@
|
|||
"""MC2-P1: Market simulator.
|
||||
|
||||
Copyright 2018, Georgia Institute of Technology (Georgia Tech)
|
||||
Atlanta, Georgia 30332
|
||||
All Rights Reserved
|
||||
|
||||
Template code for CS 4646/7646
|
||||
|
||||
Georgia Tech asserts copyright ownership of this template and all derivative
|
||||
works, including solutions to the projects assigned in this course. Students
|
||||
and other users of this template code are advised not to share it with others
|
||||
or to make it available on publicly viewable websites including repositories
|
||||
such as github and gitlab. This copyright statement should not be removed
|
||||
or edited.
|
||||
|
||||
We do grant permission to share solutions privately with non-students such
|
||||
as potential employers. However, sharing with other current or future
|
||||
students of CS 7646 is prohibited and subject to being investigated as a
|
||||
GT honor code violation.
|
||||
|
||||
-----do not edit anything above this line---
|
||||
|
||||
Student Name: Tucker Balch (replace with your name)
|
||||
GT User ID: felixm (replace with your User ID)
|
||||
GT ID: 1337 (replace with your GT ID)
|
||||
"""
|
||||
|
||||
import pandas as pd
|
||||
from util import get_data, plot_data
|
||||
from optimize_something.optimization import calculate_stats
|
||||
|
||||
|
||||
def read_orders(orders_file):
|
||||
"""
|
||||
Parser orders into the form:
|
||||
|
||||
Date datetime64[ns]
|
||||
Symbol object
|
||||
Order object
|
||||
Shares int32
|
||||
|
||||
This is how the order book looks like:
|
||||
|
||||
Date,Symbol,Order,Shares
|
||||
2011-01-10,AAPL,BUY,1500
|
||||
2011-01-10,AAPL,SELL,1500
|
||||
"""
|
||||
orders = pd.read_csv(orders_file,
|
||||
index_col=['Date'],
|
||||
dtype='|str, str, str, i4',
|
||||
parse_dates=['Date'])
|
||||
orders.sort_values(by="Date", inplace=True)
|
||||
return orders
|
||||
|
||||
|
||||
def get_order_book_info(orders):
|
||||
"""Return start_date, end_date, and symbols (as a list)."""
|
||||
start_date = orders.index[0]
|
||||
end_date = orders.index[-1]
|
||||
symbols = sorted(list((set(orders.Symbol.tolist()))))
|
||||
return start_date, end_date, symbols
|
||||
|
||||
|
||||
def get_portfolio_value(holding, prices):
|
||||
"""Calculate the current portofolio value."""
|
||||
value = 0
|
||||
for ticker, shares in holding.items():
|
||||
if ticker == 'cash':
|
||||
value += shares
|
||||
else:
|
||||
value += shares * prices[ticker]
|
||||
return value
|
||||
|
||||
|
||||
def handle_order(date, order, holding, prices, commission, impact):
|
||||
"""Process the order."""
|
||||
symbol, order, shares = order
|
||||
if shares == 0 and order == "":
|
||||
return # empty order
|
||||
if pd.isnull(shares):
|
||||
return # shares is nan
|
||||
|
||||
# Allow indicating buying and selling via shares. If shares is positive we
|
||||
# buy and if it is negative we sell.
|
||||
if shares > 0 and order == "":
|
||||
order = "BUY"
|
||||
elif shares < 0 and order == "":
|
||||
order = "SELL"
|
||||
shares = abs(shares)
|
||||
|
||||
adj_closing_price = prices[symbol]
|
||||
cost = shares * adj_closing_price
|
||||
# Charge commission and deduct impact penalty
|
||||
holding['cash'] -= (commission + impact * adj_closing_price * shares)
|
||||
if order.upper() == "BUY":
|
||||
# print(f"Buy {shares:6} of {symbol:4} on {date}")
|
||||
holding['cash'] -= cost
|
||||
holding[symbol] += shares
|
||||
elif order.upper() == "SELL":
|
||||
# print(f"Sell {shares:6} of {symbol:4} on {date}")
|
||||
holding['cash'] += cost
|
||||
holding[symbol] -= shares
|
||||
else:
|
||||
raise Exception("Unexpected order type.")
|
||||
|
||||
|
||||
def compute_portvals(orders_file, start_val=1000000, commission=9.95, impact=0.005):
|
||||
if isinstance(orders_file, pd.DataFrame):
|
||||
orders = orders_file
|
||||
else:
|
||||
orders = read_orders(orders_file)
|
||||
|
||||
start_date, end_date, symbols = get_order_book_info(orders)
|
||||
|
||||
# Tickers in the orderbook over the date_range in the order book.
|
||||
prices = get_data(symbols, pd.date_range(start_date, end_date))
|
||||
prices['Portval'] = pd.Series(0.0, index=prices.index)
|
||||
|
||||
# A dictionary to keep track of the assets we are holding.
|
||||
holding = {s: 0 for s in symbols}
|
||||
holding['cash'] = start_val
|
||||
|
||||
# Iterate over all trading days that are in the (inclusive) range of the
|
||||
# order book dates. This implicitly ignores orders placed on non-trading
|
||||
# days.
|
||||
for date, values in prices.iterrows():
|
||||
# Process orders for that day.
|
||||
for date, order in orders.loc[date:date].iterrows():
|
||||
handle_order(date, order, holding, values, commission, impact)
|
||||
# Compute portfolio value at the end of day.
|
||||
values['Portval'] = get_portfolio_value(holding, values)
|
||||
|
||||
return prices[['Portval']]
|
||||
|
||||
|
||||
def test_code():
|
||||
of = "./orders/orders-02.csv"
|
||||
sv = 1000000
|
||||
|
||||
portvals = compute_portvals(orders_file=of, start_val=sv)
|
||||
|
||||
if isinstance(portvals, pd.DataFrame):
|
||||
portvals = portvals[portvals.columns[0]] # just get the first column
|
||||
else:
|
||||
raise Exception("warning, code did not return a DataFrame")
|
||||
|
||||
start_date = portvals.index[0]
|
||||
end_date = portvals.index[-1]
|
||||
cum_ret, avg_daily_ret, \
|
||||
std_daily_ret, sharpe_ratio = calculate_stats(portvals.to_frame(), [1])
|
||||
|
||||
spy = get_data(['SPY'], pd.date_range(start_date, end_date))
|
||||
cum_ret_SPY, avg_daily_ret_SPY, \
|
||||
std_daily_ret_SPY, sharpe_ratio_SPY = calculate_stats(spy, [1])
|
||||
|
||||
# Compare portfolio against $SPY
|
||||
print(f"Date Range: {start_date} to {end_date}")
|
||||
print()
|
||||
print(f"Sharpe Ratio of Fund: {sharpe_ratio}")
|
||||
print(f"Sharpe Ratio of SPY : {sharpe_ratio_SPY}")
|
||||
print()
|
||||
print(f"Cumulative Return of Fund: {cum_ret}")
|
||||
print(f"Cumulative Return of SPY : {cum_ret_SPY}")
|
||||
print()
|
||||
print(f"Standard Deviation of Fund: {std_daily_ret}")
|
||||
print(f"Standard Deviation of SPY : {std_daily_ret_SPY}")
|
||||
print()
|
||||
print(f"Average Daily Return of Fund: {avg_daily_ret}")
|
||||
print(f"Average Daily Return of SPY : {avg_daily_ret_SPY}")
|
||||
print()
|
||||
print(f"Final Portfolio Value: {portvals[-1]}")
|
||||
|
||||
|
||||
def author():
|
||||
return 'felixm'
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
test_code()
|
|
@ -0,0 +1,8 @@
|
|||
from experiment1 import experiment1
|
||||
from experiment2 import experiment2
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
experiment1(create_report=True)
|
||||
experiment2()
|
||||
|
32
util.py
32
util.py
|
@ -14,22 +14,42 @@ def symbol_to_path(symbol, base_dir=None):
|
|||
base_dir = os.environ.get("MARKET_DATA_DIR", '../data/')
|
||||
return os.path.join(base_dir, "{}.csv".format(str(symbol)))
|
||||
|
||||
def get_data(symbols, dates, addSPY=True, colname = 'Adj Close'):
|
||||
def get_data(symbols, dates, addSPY=True, colname='Adj Close', datecol='Date'):
|
||||
"""Read stock data (adjusted close) for given symbols from CSV files."""
|
||||
df = pd.DataFrame(index=dates)
|
||||
if addSPY and 'SPY' not in symbols: # add SPY for reference, if absent
|
||||
symbols = ['SPY'] + list(symbols) # handles the case where symbols is np array of 'object'
|
||||
# handles the case where symbols is np array of 'object'
|
||||
symbols = ['SPY'] + list(symbols)
|
||||
|
||||
for symbol in symbols:
|
||||
df_temp = pd.read_csv(symbol_to_path(symbol), index_col='Date',
|
||||
parse_dates=True, usecols=['Date', colname], na_values=['nan'])
|
||||
if 'BTC' in symbol or 'ETH' in symbol:
|
||||
colname = 'close'
|
||||
datecol = 'time'
|
||||
elif symbol == 'SPY':
|
||||
colname = 'close'
|
||||
datecol = 'time'
|
||||
else:
|
||||
colname = 'Adj Close'
|
||||
datecol = 'Date'
|
||||
|
||||
df_temp = pd.read_csv(symbol_to_path(symbol),
|
||||
index_col=datecol,
|
||||
parse_dates=True, usecols=[datecol, colname],
|
||||
na_values=['nan'])
|
||||
df_temp = df_temp.rename(columns={colname: symbol})
|
||||
|
||||
if datecol == 'time':
|
||||
df_temp['date'] = pd.to_datetime(df_temp.index, unit='s')
|
||||
df_temp['date'] = pd.DatetimeIndex(df_temp['date']).normalize()
|
||||
df_temp.set_index('date', drop=True, inplace=True)
|
||||
|
||||
df = df.join(df_temp)
|
||||
if symbol == 'SPY': # drop dates SPY did not trade
|
||||
df = df.dropna(subset=["SPY"])
|
||||
|
||||
pass
|
||||
# df = df.dropna(subset=["SPY"])
|
||||
return df
|
||||
|
||||
|
||||
def plot_data(df, title="Stock prices", xlabel="Date", ylabel="Price"):
|
||||
import matplotlib.pyplot as plt
|
||||
"""Plot stock prices with a custom title and meaningful axis labels."""
|
||||
|
|
Loading…
Reference in New Issue