Merge remote-tracking branch 'origin/develop' into dev/travis-setup

examples/neural_network.md

@@ -167,18 +167,18 @@ And there you have it!  You can clearly see what the 'TestNN' predictions where
 
 Let's create our trading criteria, starting with some entry criteria.  We want to check if the stock's current price is 'below' the threshold_below value - this will indicate that the neural network is predicting that the stock price will go up by more than $5 in our prediction window.
 
-    enter_crit_long = criteria.Position(googl.close, 'below', threshold_below.value)
+    enter_crit_long = criteria.Below(googl.close, threshold_below.value)
 
 We also want to check if the current stock price is 'above' the threshold_above value, which would indicate that the neural network is predicting that the stock price will go down by more than $5 in our prediction window.
 
-    enter_crit_short = criteria.Position(symbol.close, 'above', threshold_above.value)
+    enter_crit_short = criteria.Above(symbol.close, threshold_above.value)
 
 As for our exit criteria, we only have one, and it would probably be wise if it matched our neural network training parameters from earlier.  We had trained the network to predict the price 1 day into the future.  Therefore, we should have an exit criteria that says, "When we've been in the market for 1 day(s), exit".
 
-    exit_crit_long = criteria.TimeSinceAction(googl, Long(), 1)
-    exit_crit_short = criteria.TimeSinceAction(googl, Short(), 1)
+    exit_crit_long = criteria.BarsSinceLong(googl, 1)
+    exit_crit_short = criteria.BarsSinceShort(googl, 1)
 
-Pretty straightforward.  The TimeSinceAction criteria takes a stock, an action (such as Long(), Short(), LongExit(), ShortExit()), and a number of days/hours/minutes/etc.  The criteria will return true if the action specified has happened for the stock in the past X days/hours/minutes/etc -  In this case, if we have Long'ed and/or Short'ed googl exactly 1 days ago.
+Pretty straightforward.  The BarsSinceLong/BarsSinceShort/BarsSinceLongExit/BarsSinceShortExit criteria takes a stock and a number of bars. The criteria will return true if the specific action has happened for the stock in the past X bars -  In this case, if we have Long'ed and/or Short'ed googl exactly 1 day ago.
 
 We could have just as easily used +/- $5 from our stock price instead of the neural network prediction price. Then we could have checked if the neural network price fell outside that range to trade accordingly.
 

examples/neural_network.py

@@ -43,12 +43,12 @@
 test_dataset.add_technical_indicator(threshold_below)
 # Criteria
 # Current price is below the threshold of our neural network's prediction price
-enter_crit_long = criteria.Position(symbol.close, 'below', threshold_below.value)
+enter_crit_long = criteria.Below(symbol.close, threshold_below.value)
 # Current price is above the threshold of our neural network's prediction price
-enter_crit_short = criteria.Position(symbol.close, 'above', threshold_above.value)
+enter_crit_short = criteria.Above(symbol.close, threshold_above.value)
 # Exit after 1 day - as per the network's parameters
-exit_crit_long = criteria.TimeSinceAction(symbol, Long(), 1)
-exit_crit_short = criteria.TimeSinceAction(symbol, Short(), 1)
+exit_crit_long = criteria.BarsSinceLong(symbol, 1)
+exit_crit_short = criteria.BarsSinceShort(symbol, 1)
 # Criteria Groups
 enter_crit_group1 = criteria_group.CriteriaGroup([enter_crit_long], Long(), symbol)
 enter_crit_group2 = criteria_group.CriteriaGroup([enter_crit_short], Short(), symbol)

examples/random_forest.md

@@ -81,7 +81,7 @@ Typically, this is the part where I would create the testing data set, but we've
 
 Our random forest technical indicator will spit out prediction prices, but we still need a threshold to determine whether or not we want to enter the trade.  Let's define a technical indicator that is $5 higher than the ensemble's prediction, and one that is $5 lower.
 
-_Note that we could have also simply used a Position criteria on the ensemble price and skipped the threshold step entirely._
+_Note that we could have also simply used a Above and Below criteria on the ensemble price and skipped the threshold step entirely._
 
     threshold_above = technical_indicator.Addition(random_forest.value, 5)
     test_dataset.add_technical_indicator(threshold_above)
@@ -133,18 +133,18 @@ Click [here](http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFr
 
 Let's create our trading criteria, starting with some entry criteria.  We want to check if the stock's current price is 'below' the threshold_below value - this will indicate that the random forest ensemble is predicting that the stock price will go up by more than $5 in our prediction window.
 
-    enter_crit_long = criteria.Position(symbol.close, 'below', threshold_below.value)
+    enter_crit_long = criteria.Below(symbol.close, threshold_below.value)
 
 We also want to check if the current stock price is 'above' the threshold_above value, which would indicate that the random forest ensemble is predicting that the stock price will go down by more than $5 in our prediction window.
 
-    enter_crit_short = criteria.Position(symbol.close, 'above', threshold_above.value)
+    enter_crit_short = criteria.Above(symbol.close, threshold_above.value)
 
 As for our exit criteria, we only have one, and it would probably be wise if it matched our ensemble fitting parameters from earlier.  We fit the model to predict the price 5 days into the future.  Therefore, we should have an exit criteria that says, "When we've been in the market for 5 days, exit".
 
-    exit_crit_long = criteria.TimeSinceAction(symbol, Long(), 5)
-    exit_crit_short = criteria.TimeSinceAction(symbol, Short(), 5)
+    exit_crit_long = criteria.BarsSinceLong(symbol, 5)
+    exit_crit_short = criteria.BarsSinceShort(symbol, 5)
 
-Pretty straightforward.  The TimeSinceAction criteria takes a stock, an action (such as Long(), Short(), LongExit(), ShortExit()), and a number of days/hours/minutes/etc.  The criteria will return true if the action specified has happened for the stock in the past X days/hours/minutes/etc -  In this case, if we have Long'ed and/or Short'ed GOOGL exactly 5 days ago.
+Pretty straightforward.  The BarsSinceLong/BarsSinceShort/BarsSinceLongExit/BarsSinceShortExit criteria takes a stock and a number of bars.  The criteria will return true if the specific action has happened for the stock in the past X bars -  In this case, if we have Long'ed and/or Short'ed GOOGL exactly 5 days ago.
 
 We could have just as easily used +/- $5 from our stock price instead of the random forest prediction price. Then we could have checked if the random forest price fell outside that range to trade accordingly.
 

examples/random_forest.py

@@ -42,12 +42,12 @@
 test_dataset.add_technical_indicator(threshold_below)
 # Criteria
 # Current price is below the threshold of our random forest's prediction price
-enter_crit_long = criteria.Position(symbol.close, 'below', threshold_below.value)
+enter_crit_long = criteria.Below(symbol.close, threshold_below.value)
 # Current price is above the threshold of our random forest's prediction price
-enter_crit_short = criteria.Position(symbol.close, 'above', threshold_above.value)
+enter_crit_short = criteria.Above(symbol.close, threshold_above.value)
 # Exit after 5 days - as per the random forest's build parameters
-exit_crit_long = criteria.TimeSinceAction(symbol, Long(), 5)
-exit_crit_short = criteria.TimeSinceAction(symbol, Short(), 5)
+exit_crit_long = criteria.BarsSinceLong(symbol, 5)
+exit_crit_short = criteria.BarsSinceShort(symbol, 5)
 # Criteria Groups
 enter_crit_group1 = criteria_group.CriteriaGroup([enter_crit_long], Long(), symbol)
 enter_crit_group2 = criteria_group.CriteriaGroup([enter_crit_short], Short(), symbol)

nowtrade/configuration.py

@@ -14,8 +14,8 @@
 # Criteria module
 #LOGGING_Position_CONSOLE = logging.DEBUG
 #LOGGING_Position_FILE = logging.DEBUG
-#LOGGING_TimeSinceAction_CONSOLE = logging.DEBUG
-#LOGGING_TimeSinceAction_FILE = logging.DEBUG
+#LOGGING_BarsSinceLong_CONSOLE = logging.DEBUG
+#LOGGING_BarsSinceLong_FILE = logging.DEBUG
 
 # Criteria Group module
 #LOGGING_CriteriaGroup_CONSOLE = logging.DEBUG

nowtrade/criteria.py

@@ -1,6 +1,7 @@
 import numpy as np
 import pandas as pd
 from nowtrade import logger
+from nowtrade import action
 
 class Criteria(object):
     """
@@ -14,18 +15,18 @@ def __init__(self):
         # Some TIs don't require the entire data_frame history.
         self.num_bars_required = None # Requires all history by default
 
-class TimeSinceAction(Criteria):
+class BarsSinceAction(Criteria):
     def __init__(self, symbol, action, periods, condition=None):
         Criteria.__init__(self)
         self.symbol = str(symbol)
         self.action = action.raw()
         self.periods = periods
         self.condition = str(condition).upper()
         self.num_bars_required = self.periods + 1
-        self.label = 'TimeSinceAction_%s_%s_%s_%s' %(symbol, action, periods, condition)
+        self.label = 'BarsSinceAction_%s_%s_%s_%s' %(symbol, action, periods, condition)
         self.logger.info('Initialized - %s' %self)
     def __str__(self):
-        return 'TimeSinceAction(symbol=%s, action=%s, periods=%s, condition=%s)' %(self.symbol, self.action, self.periods, self.condition)
+        return 'BarsSinceAction(symbol=%s, action=%s, periods=%s, condition=%s)' %(self.symbol, self.action, self.periods, self.condition)
     def __repr__(self): return self.label
     def apply(self, data_frame):
         if self.condition == 'OVER':
@@ -40,6 +41,31 @@ def apply(self, data_frame):
                 return self.action == data_frame['ACTIONS_%s' %self.symbol][-1-self.periods]
             return False
 
+class BarsSinceLong(BarsSinceAction):
+    def __init__(self, symbol, periods, condition=None):
+        BarsSinceAction.__init__(self, symbol, action.Long(), periods, condition)
+        self.label = 'BarsSinceLong_%s_%s_%s' %(symbol, periods, condition)
+    def __str__(self):
+        return 'BarsSinceLong(symbol=%s, periods=%s, condition=%s)' %(self.symbol, self.periods, self.condition)
+class BarsSinceShort(BarsSinceAction):
+    def __init__(self, symbol, periods, condition=None):
+        BarsSinceAction.__init__(self, symbol, action.Short(), periods, condition)
+        self.label = 'BarsSinceShort_%s_%s_%s' %(symbol, periods, condition)
+    def __str__(self):
+        return 'BarsSinceShort(symbol=%s, periods=%s, condition=%s)' %(self.symbol, self.periods, self.condition)
+class BarsSinceLongExit(BarsSinceAction):
+    def __init__(self, symbol, periods, condition=None):
+        BarsSinceAction.__init__(self, symbol, action.LongExit(), periods, condition)
+        self.label = 'BarsSinceLongExit_%s_%s_%s' %(symbol, periods, condition)
+    def __str__(self):
+        return 'BarsSinceLongExit(symbol=%s, periods=%s, condition=%s)' %(self.symbol, self.periods, self.condition)
+class BarsSinceShortExit(BarsSinceAction):
+    def __init__(self, symbol, periods, condition=None):
+        BarsSinceAction.__init__(self, symbol, action.ShortExit(), periods, condition)
+        self.label = 'BarsSinceShortExit_%s_%s_%s' %(symbol, periods, condition)
+    def __str__(self):
+        return 'BarsSinceShortExit(symbol=%s, periods=%s, condition=%s)' %(self.symbol, self.periods, self.condition)
+
 class InMarket(Criteria):
     """
     Looks one period behind the current period because it expects the last
@@ -282,6 +308,66 @@ def apply(self, data_frame):
             else: results = data_frame[self.param1] == data_frame[self.param2]
         return results.iloc[-1]
 
+class Above(Criteria):
+    def __init__(self, param1, param2, lookback=1):
+        Criteria.__init__(self)
+        self.param1 = param1 # Technical indicator label
+        self.param2 = param2 # Technical indicator label or int or long or float
+        self.lookback = lookback
+        self.label = 'Above_%s_%s_%s' %(param1, param2, lookback)
+        self.num_bars_required = lookback
+        self.logger.info('Initialized - %s' %self)
+    def __str__(self):
+        return 'Above(param1=%s, param2=%s, lookback=%s)' %(self.param1, self.param2, self.lookback)
+    def __repr__(self): return self.label
+    def apply(self, data_frame):
+        if len(data_frame) < self.lookback: return False
+        # Second value is not a technical indicator, simply a number to compare
+        if isinstance(self.param2, (int, long, float)):
+            print 'comparing %s with %s' %(data_frame[self.param1][-self.lookback], self.param2)
+            return data_frame[self.param1][-self.lookback] > self.param2
+        else: return data_frame[self.param1][-self.lookback] > data_frame[self.param2][-self.lookback]
+
+class Below(Criteria):
+    def __init__(self, param1, param2, lookback=1):
+        Criteria.__init__(self)
+        self.param1 = param1 # Technical indicator label
+        self.param2 = param2 # Technical indicator label or int or long or float
+        self.lookback = lookback
+        self.label = 'Below_%s_%s_%s' %(param1, param2, lookback)
+        self.num_bars_required = lookback
+        self.logger.info('Initialized - %s' %self)
+    def __str__(self):
+        return 'Below(param1=%s, param2=%s, lookback=%s)' %(self.param1, self.param2, self.lookback)
+    def __repr__(self): return self.label
+    def apply(self, data_frame):
+        if len(data_frame) < self.lookback: return False
+        # Second value is not a technical indicator, simply a number to compare
+        if isinstance(self.param2, (int, long, float)):
+            return data_frame[self.param1][-self.lookback] < self.param2
+        else: return data_frame[self.param1][-self.lookback] < data_frame[self.param2][-self.lookback]
+
+class Equals(Criteria):
+    def __init__(self, param1, param2, lookback=1):
+        Criteria.__init__(self)
+        self.param1 = param1 # Technical indicator label
+        self.param2 = param2 # Technical indicator label or int or long or float
+        self.lookback = lookback
+        self.label = 'Equals_%s_%s_%s' %(param1, param2, lookback)
+        self.num_bars_required = lookback
+        self.logger.info('Initialized - %s' %self)
+    def __str__(self):
+        return 'Equals(param1=%s, param2=%s, lookback=%s)' %(self.param1, self.param2, self.lookback)
+    def __repr__(self): return self.label
+    def apply(self, data_frame):
+        if len(data_frame) < self.lookback: return False
+        # Second value is not a technical indicator, simply a number to compare
+        if isinstance(self.param2, (int, long, float)):
+            print 'comparing %s and %s' %(data_frame[self.param1][-self.lookback], self.param2)
+            return data_frame[self.param1][-self.lookback] == self.param2
+        else: return data_frame[self.param1][-self.lookback] == data_frame[self.param2][-self.lookback]
+Equal = Equals
+
 class InRange(Criteria):
     """
     Determines if a technical indicator is in between two values.