from collections import deque
import random
'''
Example below replicates
+75 MSFT 25.10
+50 MSFT 25.12
-100 MSFT 25.22
Realized P&L = 75 * (25.22 - 25.10) + 25 * (25.22 - 25.12) = $ 11.50
A Trade is split into a set of unit positions that are then dequeued on FIFO basis as part of Sell.
'''
number_of_sell_trades = 1000
max_sell_quentity = 5
min_sell_price = 23.00
max_sell_price = 27.00
class TradeManager():
def __init__(self):
# FIFO queue that we can use to enqueue unit buys and
# dequeue unit sells.
self.fifo = deque()
self.profit = []
def __repr__(self):
return 'position size: %d'%(len(self.fifo))
def execute_with_total_pnl(self, direction, quantity, price):
#print direction, quantity, price, 'position size', len(self.fifo)
if len(self.fifo) == 0:
return 0
if 'Sell' in (direction):
if len(self.fifo) >= quantity:
return sum([(price - fill.price) for fill in tm.execute(direction, quantity, price)])
else:
return 0
else:
return [tm.execute(direction, quantity, price)]
def execute(self, direction, quantity, price):
#print direction, quantity, price, 'position size', len(self.fifo)
if direction in ('Buy'):
for i, fill in Trade(direction, quantity, price):
self.fifo.appendleft(fill)
yield fill
elif direction in ('Sell'):
for i, fill in Trade(direction, quantity, price):
yield self.fifo.pop()
class Fill():
def __init__(self, price):
self.price = price
self.quantity = 1
class Trade():
def __init__(self, direction, quantity, price):
self.direction = direction
self.quantity = quantity
self.price = price
self.i = 0
def __iter__(self):
return self
def next(self):
if self.i < self.quantity:
i = self.i
self.i += 1
return i, Fill(self.price)
else:
raise StopIteration()
# create a TradeManager
tm = TradeManager()
# generate some buys
a = [i for i in tm.execute('Buy', 75, 25.10)]
a = [i for i in tm.execute('Buy', 50, 25.12)]
# generate sell
pnl = np.cumsum(tm.execute_with_total_pnl('Sell', 100, 25.22))
# how much did we make
print 'total pnl', pnl[-1:]
# try something more involved.
tm = TradeManager()
pnl_ending = []
# run n simulations
for step in range(0,50):
a = [i for i in tm.execute('Buy', 75000, 25)]
pnl = np.cumsum([tm.execute_with_total_pnl('Sell', quantity, random.uniform(min_sell_price, max_sell_price)) \
for quantity in [random.randint(0,max_sell_quentity) \
for i in range(0,number_of_sell_trades,1)]])
plot(pnl)
pnl_ending.append(pnl[-1:][0])
print 'step', step, 'pnl', pnl[-1:][0], 'avg. pnl', np.mean(pnl_ending), 'diff to mean', pnl[-1:][0]-np.mean(pnl_ending)
print 'avg, total pnl', np.mean(pnl_ending) #pnl[-1:][0]
show()
# bin the results
hist(pnl_ending, 25)
grid(True)
show()
# could lookat fitting and var.
Diff to Previous Revision
--- revision 1 2015-02-10 12:46:10
+++ revision 2 2015-02-10 13:34:48
@@ -15,14 +15,16 @@
'''
-number_of_sell_trades = 100
+number_of_sell_trades = 1000
max_sell_quentity = 5
min_sell_price = 23.00
-max_sell_price = 25.00
+max_sell_price = 27.00
-class TradeManager():
+class TradeManager():
def __init__(self):
+ # FIFO queue that we can use to enqueue unit buys and
+ # dequeue unit sells.
self.fifo = deque()
self.profit = []
@@ -30,48 +32,35 @@
return 'position size: %d'%(len(self.fifo))
def execute_with_total_pnl(self, direction, quantity, price):
- print direction, quantity, price, 'position size', len(self.fifo)
+ #print direction, quantity, price, 'position size', len(self.fifo)
if len(self.fifo) == 0:
return 0
if 'Sell' in (direction):
if len(self.fifo) >= quantity:
- return sum([(fill.price - price) for fill in tm.execute(direction, quantity, price)])
+ return sum([(price - fill.price) for fill in tm.execute(direction, quantity, price)])
else:
- return 0
- #raise MyShortSellException('No short selling allowed')
+ return 0
else:
- return [tm.execute(direction, quantity, price)]
-
- #def execute_with_pnl(self, direction, quantity, price):
- # if 'Sell' in (direction):
- # for fill in tm.execute(direction, quantity, price):
- # self.profit.append(fill.price - price)
- # yield (fill.price - price)
- # else:
- # yield tm.execute(direction, quantity, price)
+ return [tm.execute(direction, quantity, price)]
- def execute(self, direction, quantity, price, verbose=False):
- if verbose:
- print 'before', (self.fifo)
- if direction in ('Buy'):
- #print 'Buy'
+ def execute(self, direction, quantity, price):
+ #print direction, quantity, price, 'position size', len(self.fifo)
+ if direction in ('Buy'):
for i, fill in Trade(direction, quantity, price):
self.fifo.appendleft(fill)
yield fill
elif direction in ('Sell'):
for i, fill in Trade(direction, quantity, price):
- yield self.fifo.pop()
- if verbose:
- print 'after', len(self.fifo)
+ yield self.fifo.pop()
-class Fill():
+class Fill():
def __init__(self, price):
self.price = price
self.quantity = 1
-class Trade():
+class Trade():
def __init__(self, direction, quantity, price):
self.direction = direction
self.quantity = quantity
@@ -93,22 +82,35 @@
tm = TradeManager()
# generate some buys
-a = [i for i in tm.execute('Buy', 75, 25.12)]
-a = [i for i in tm.execute('Buy', 50, 25.22)]
+a = [i for i in tm.execute('Buy', 75, 25.10)]
+a = [i for i in tm.execute('Buy', 50, 25.12)]
-# generate some sells
-pnl = np.cumsum([tm.execute_with_total_pnl('Sell', quantity, random.uniform(min_sell_price, max_sell_price)) \
- for quantity in [random.randint(0,max_sell_quentity) \
- for i in range(0,number_of_sell_trades,1)]])
-plot(pnl)
+# generate sell
+pnl = np.cumsum(tm.execute_with_total_pnl('Sell', 100, 25.22))
+
+# how much did we make
print 'total pnl', pnl[-1:]
-
# try something more involved.
tm = TradeManager()
-a = [i for i in tm.execute('Buy', 75, 25.10)]
-pnl = np.cumsum([tm.execute_with_total_pnl('Sell', quantity, random.uniform(min_sell_price, max_sell_price)) \
+pnl_ending = []
+
+# run n simulations
+for step in range(0,50):
+ a = [i for i in tm.execute('Buy', 75000, 25)]
+ pnl = np.cumsum([tm.execute_with_total_pnl('Sell', quantity, random.uniform(min_sell_price, max_sell_price)) \
for quantity in [random.randint(0,max_sell_quentity) \
for i in range(0,number_of_sell_trades,1)]])
-plot(pnl)
-print 'total pnl', pnl[-1:]
+ plot(pnl)
+ pnl_ending.append(pnl[-1:][0])
+ print 'step', step, 'pnl', pnl[-1:][0], 'avg. pnl', np.mean(pnl_ending), 'diff to mean', pnl[-1:][0]-np.mean(pnl_ending)
+
+print 'avg, total pnl', np.mean(pnl_ending) #pnl[-1:][0]
+show()
+
+# bin the results
+hist(pnl_ending, 25)
+grid(True)
+show()
+
+# could lookat fitting and var.
