# MACD

## Overview

The MACD (Moving Average Convergence and Divergence) is another particular example of a strategy that is common and can be implemented. Similarly to the Golden Cross, the MACD takes in moving averages and calculates the difference between the two (convergence and divergence).

For those that have a keen eye, this strategy is extremely similar to the golden cross strategy. In fact if you used the 50 and the 200-day SMA as your configuration, you would produce the same algorithm just written differently!

Our condition for buying and selling will be when the MACD crosses above the zero (i.e. when two moving averages cross) and sell when they cross below 0.

## Implementing in Blankly

### Boilerplate Code

To get started, make sure you have already [set up](/getting-started/installation) your environment along with the necssary keys and settings.

We will be implementing a MACD strategyusing Blankly.Strategy that allows for a quick and easy way of building out our golden cross. We'll also be utilizing blankly.indicators to quickly implement the MACD calculations.

#### Create Strategy

from blankly import Strategy, StrategyState, Interface
from blankly import Alpaca
from blankly.indicators import macd

def init(symbol, state: StrategyState):
# run on a new price event to initialize variables
pass

def price_event(price, symbol, state: StrategyState):
# we'll come back to this soon
pass

alpaca = Alpaca()
s = Strategy(alpaca)
s.start()


### Initializing Variables and History

In order to speed things up, we should make one call to get the historical data that we need and append data as new prices come in. We also need to initialize some variabels for the price event.

We can actually easily do this on initialization and make sure the proper data is passed in to the proper price events:

SHORT_PERIOD = 12
LONG_PERIOD = 26
SIGNAL_PERIOD = 9

def init(symbol, state: StrategyState):
interface: Interface = state.interface
resolution: float = state.resolution
variables = state.variables
# initialize the historical data
variables['history'] = interface.history(symbol, 800,
resolution,
return_as='list')['close']
variables['short_period'] = SHORT_PERIOD
variables['long_period'] = LONG_PERIOD
variables['signal_period'] = SIGNAL_PERIOD
variables['has_bought'] = False


### Implementing the Price Event

Now that we have the code set up, let's take a deep dive into how to implement this price event.

We want to buy when the MACD crosses above the MACD Signal Line (which means the slope of MACD must be positive), thus, we're going to analyze the difference between two points of the MACD to detect positive slope and the crossing up. We also want to note that the MACD may skip over the MACD Signal Line (i.e. go from -0.00005 to 0.35), and in this case, we still want to buy when the MACD is 0.35.

We will use the 12 and 26 day EMA as our short and long periods and 9 as our signal period (see Investopedia)

Due to float precision errors with binary (try 2.1 * 0.2 in the terminal), we've carefully implemented a function called trunc that will allow you to easily calculate the correct amount you want to sell without worrying about floating point errors with the exchange. Simply import it from utils
def price_event(price, symbol, state: StrategyState):
interface: Interface = state.interface
# allow the resolution to be any resolution: 15m, 30m, 1d, etc.
variables = state.variables

variables['history'].append(price)
macd_res, macd_signal, macd_histogram = macd(variables['history'],
short_period=variables['short_period'],
long_period=variables['long_period'],
signal_period=variables['signal_period'])

slope_macd = (macd_res[-1] - macd_res[-5]) / 5  # get the slope of the last 5 MACD_points
prev_macd = macd_res[-2]
curr_macd = macd_res[-1]
curr_signal_macd = macd_signal[-1]

# We want to make sure this works even if curr_macd does not equal the signal MACD
is_cross_up = slope_macd > 0 and curr_macd >= curr_signal_macd > prev_macd

is_cross_down = slope_macd < 0 and curr_macd <= curr_signal_macd < prev_macd
if is_cross_up and not variables['has_bought']:
# buy with all available cash
variables['has_bought'] = True
elif is_cross_down and variables['has_bought']:
# sell all of the position
interface.market_order(symbol, 'sell', int(interface.account[symbol].available))
variables['has_bought'] = False

This example shows how global namespace can actually also be used in price_events, Blankly does not change namespaces so you can access anything in your general python file. This makes it easy to customize different strategies. You can also use
state.variables
as a way to integrate with local Strategy state

Now that we've gotten everything, let's bring it all together

from blankly import Strategy, StrategyState, Interface
from blankly import Alpaca
from blankly.utils import trunc
from blankly.indicators import macd

SHORT_PERIOD = 12
LONG_PERIOD = 26
SIGNAL_PERIOD = 9

def init(symbol, state: StrategyState):
interface: Interface = state.interface
resolution: float = state.resolution
variables = state.variables
# initialize the historical data
variables['history'] = interface.history(symbol, 800,
resolution,
return_as='list')['close']
variables['short_period'] = SHORT_PERIOD
variables['long_period'] = LONG_PERIOD
variables['signal_period'] = SIGNAL_PERIOD
variables['has_bought'] = False

def price_event(price, symbol, state: StrategyState):
interface: Interface = state.interface
# allow the resolution to be any resolution: 15m, 30m, 1d, etc.
variables = state.variables

variables['history'].append(price)
macd_res, macd_signal, macd_histogram = macd(variables['history'],
short_period=variables['short_period'],
long_period=variables['long_period'],
signal_period=variables['signal_period'])

slope_macd = (macd_res[-1] - macd_res[-5]) / 5  # get the slope of the last 5 MACD_points
prev_macd = macd_res[-2]
curr_macd = macd_res[-1]
curr_signal_macd = macd_signal[-1]

# We want to make sure this works even if curr_macd does not equal the signal MACD
is_cross_up = slope_macd > 0 and curr_macd >= curr_signal_macd > prev_macd

is_cross_down = slope_macd < 0 and curr_macd <= curr_signal_macd < prev_macd
if is_cross_up and not variables['has_bought']:
# buy with all available cash