Indicators

Overview

Indicators are a central part of developing strong quantitative models. As we know, price data is noisy, unpredictable, and constantly changing. Using indicators that have been backed by years of research have shown to improve performance and ultimately have become the backbone of many modern quantitative models. Whether it's starting with the most basic golden cross, or using deep learning with inputs including RSI, TEMA, and more, indicators are extremely useful to know about and to learn.

Blankly has implemented a multitude of indicators for your own use and development. Specifically, we take the tulipy library and wrap it with Blankly data and Strategies for backtest.

As always, if you have any indicators that you'd like to be added, or an implementation for one, we'd love a PR!

Moving Averages

Moving averages are just as they sound: they are moving ("rolling window") averages ("aggregating data points together"). The point of moving averages is to greatly improve the quality of data points. As we mentioned before, price data is extremely noisy, with constant ups and downs, moving averages "smooth" out the data so it's very clear to see a specific trend. In Blankly, we offer a multitude of various types of moving averages. We describe them below.

sma(data, period=50, use_series=False)

Arguments

Returns

The simple moving average (SMA) is the simplest type of moving average, it takes a time series (f.e. price data) and divides it up into windows defined by the period. Then it averages all of those points to produce one point. Thus, if you have 250 total data points, you will be returned $250/5=5$ data points. A common use-case for the SMA is the golden cross or the MACD strategies where we look fro the crossing of various simple moving averages (for example the 50 day SMA and the 100 day SMA) to see trend changes. The longer the period, the "harder" it is to change (the less it follows local market changes).

ema(data, period=50, use_series=False)

Arguments

Returns

Exponential moving average (EMA) is calculated a little differently than the SMA, the key difference between the SMA and the EMA is that the EMA places greater weight and value to more recent data points in the dataset. However, the premise for looking at crossovers and divergences with moving averages stays the same.

For more information, check out investopedia

vwma(data, volume_data, period=50, use_series=False)

Arguments

Returns

This is a moving average that makes more weight for the days that have higher volume than others. Check out Investopedia

wma(data, period=50, use_series=False)

Arguments

Returns

A weighted moving average is very similar to the EMA, however their weighting function for recent points is different. Namely WMAs are linear weighting and EMA is exponential weighting. Check out this great article by Investopedia

zlema(data, period=50, use_series=False)

Arguments

Returns

The zero-lag exponential moving average reduces aspects of "lag", check out these docs

hma(data, period=50, use_series=False)

Arguments

Returns

Similar to the ZLEMA, the Hull Moving Average (HMA) modifies the Weighted Moving Average, check out these docs

trima(data, period=50, use_series=False)

Arguments

Returns

The Triangle Moving Average (TRIMA) puts more weight on the middle items in the dataset and less weight on the new and old data within a single period. Info here

kaufman_adaptive_ma(data, period=50, use_series=False)

Arguments

Returns

The KAMA adjusts its weighting relative to the current market conditions of noise. Check out this

macd(data, short_period=12, long_period=26, signal_period=9)

Arguments

Returns

The MACD (Moving Average Convergence and Divergence) is a way of looking at two different moving averages and identifying key characteristics of price movements relative to the crossing and diverging of the two moving averages. It utilizes the exponential moving average, subtracting the long_period from the short_period and using that to compare against the signal_period. Check out this in Investopedia.

Oscillators

Oscillators vary in moving averages in that instead of trying to smooth out the data, they attempt to capture the noise and volatility in the price data. Just as the name states, these indicators attempt to capture how much prices oscillates.

rsi(data, period = 14, round_rsi = False, use_series=False)

Arguments

Returns

The Relative Strength Index is perhaps one of the most commonly used oscillators. The RSI attempts to capture the momentum of recent price changes (how strong is the current uptrend and or how week is the downtrend). The RSI is calculated as shown here by Investopedia

aroon_oscillator(high_data, low_data, period=14, use_series=False)

Arguments

Returns

The aroon oscillator utilizes the aroon indicator to determine the current strength of a given trend. A value greater than zero signifies an uptrend and a value less than 0 signifies a downtrend. For more info, see here.

Example Use

from blankly.strategy import Strategy
from blankly import Alpaca, Interface
from blankly.metrics import aroon_oscillator

def price_event(price, symbol, state):
  interface: Interface = state.interface
  history = interface.history(symbol, 500, resolution=state.resolution)
  h = history['high']
  l = history['low']
  oscillation = aroon_oscillator(h, l)
  ...
a = Alpaca()
s = Strategy(a)
s.add_price_event(price_event, resolution='30m')

chande_momentum_oscillator(data, period=14, use_series=False)

Arguments

Returns

The chande momentum oscillator takes the sum of all recent gains, the sum of all recent losses and then divides that by the sum of all recent price movements (sum of all gains + losses). The formula is described by Investopedia

absolute_price_oscillator(data, short_period=12, long_period=26, use_series=False)

Arguments

Returns

The absolute price oscillator helps detect the difference between two EMAs and expresses the results as an absolute value. More information can be seen here

percentage_price_oscilator(data, short_period=12, long_period=26, use_series=False)

Arguments

Returns

The same as the absolute price, but returns values as a percentage difference.

stochastic_oscillator(high_data, low_data, close_data, pct_k_period=14, pct_k_slowing_period=3, pct_d_period=3, use_series=False)

Arguments

Returns

The stochastic oscillator compares the closing data of a security to a range of varying prices over a specific time period. More information can be found on Investopedia

stochastic_rsi(data, period=14, smooth_pct_k=3, smooth_pct_d=3, use_series=False)

Arguments

Returns

The stochastic RSI combines the RSI with the stochastic oscillators. More info can be found here.

Other Indicators

bbands(data, period=14, stddev=2)

Arguments

Returns

Bollinger Bands are a very popular technical indicator to measure the strength of an uptrend or downtrend. It is also a way to help determine trend reversals. View Investopedia's Explanation.

wad(high_data, low_data, close_data, use_series=False)

Arguments

Returns

The Williams accumulation/distribution helps determine a trend direction. View this for more info.

wilders(data, period=50, use_series=False)

Arguments

Returns

willr(high_data, low_data, close_data, period=50, use_series=False)

Arguments

Returns

The Williams %R is a measure similar to RSI and the stochastic oscillator where it measures overbought and oversold territory. More information is on Investopedia

true_range(high_data, low_data, close_data, use_series=False)

Arguments

Returns

The true range takes the greatest of high - close and low - close for every data point

average_true_range(high_data, low_data, close_data, period=50, use_series=False)

Arguments

Returns

The average true range takes the greatest of high - close and low - close for every data point and then takes a moving average of the points. More information can be found here.

Statistics

We offer a variety of general statistics as part of our indicators. All of our statistics are implemented so that they work on a period of data (i.e. they're rolling and return an Array-Like).

stddev_period(data, period=14, use_series=False)

Arguments

Returns

var_period(data, period=14, use_series=False)

Arguments

Returns

stderr_period(data, period=14, use_series=False)

Arguments

Returns

min_period(data, period, use_series=False)

Arguments

Returns

max_period(data, period, use_series=False)

Arguments

Returns