/TAcharts

Apply popular TA tools and charts to candlestick data with NumPy.

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

TAcharts 0.0.30

By: Carl Farterson

Contributors: @rnarciso, @t3ch9

This repository provides technical tools to analyze OHLCV data, along with several TA chart functionalities. These functions are optimized for speed and utilize numpy vectorization over built-in pandas methods when possible.

Methods

Indicators With Chart Functionality

  • Bollinger(df=None, filename=None, interval=None, n=20, ndev=2): Bollinger Bands
  • Ichimoku(df=None, filename=None, interval=None): Ichimoku Cloud
  • Renko(df=None, filename=None, interval=None): Renko Chart

Indicators Without Chart Functionality

  • atr(high, low, close, n=2): average true range from candlestick data
  • cmf(df, n=2): Chaikin Money Flow of an OHLCV dataset
  • double_smooth(src, n_slow, n_fast): The smoothed value of two EMAs
  • ema(src, n=2): exponential moving average for a list of src across n periods
  • macd(src, slow=25, fast=13): moving average convergence/divergence of src
  • mmo(src, n=2): Murrey Math oscillator of src
  • roc(src, n=2): rate of change of src across n periods
  • rolling(src, n=2, fn=None, axis=1): rolling sum, max, min, mean, or median of src across n periods
  • rsi(src, n=2): relative strength index of src across n periods
  • sdev(src, n=2): standard deviation across n periods
  • sma(src, n=2): simple moving average of src across n periods
  • td_sequential(src, n=2): TD sequential of src across n periods
  • tsi(src, slow=25, fast=13): true strength indicator

utils

  • area_between(line1, line2): find the area between line1 and line2
  • crossover(x1, x2): find all instances of intersections between two lines
  • draw_candlesticks(ax, df): add candlestick visuals to a matplotlib chart
  • fill_values(averages, interval, target_len): Fill missing values with evenly spaced samples.
    • Example: You're using 15-min candlestick data to find the 1-hour moving average and want a value at every 15-min mark, and not every 1-hour mark.
  • group_candles(df, interval=4): combine candles so instead of needing a different dataset for each time interval, you can form time intervals using more precise data.
    • Example: you have 15-min candlestick data but want to test a strategy based on 1-hour candlestick data (interval=4).
  • intersection(a0, a1, b0, b1): find the intersection coordinates between vector A and vector B

How it works

Create your DataFrame

# NOTE: we are using 1-hour BTC OHLCV data from 2019.01.01 00:00:00 to 2019.12.31 23:00:00
from TAcharts.utils.ohlcv import OHLCV

df = OHLCV().btc

df.head()
  date open high low close volume
0 2019-01-01 00:00:00 3699.95 3713.93 3697.00 3703.56 660.279771
1 2019-01-01 01:00:00 3703.63 3726.64 3703.34 3713.83 823.625491
2 2019-01-01 02:00:00 3714.19 3731.19 3707.00 3716.70 887.101362
3 2019-01-01 03:00:00 3716.98 3732.00 3696.14 3699.95 955.879034
4 2019-01-01 04:00:00 3699.96 3717.11 3698.00 3713.07 534.113945

Bollinger Bands

from TAcharts.indicators.bollinger import Bollinger

b = Bollinger(df)
b.build(n=20, ndev=2)

b.plot()

bollinger

Ichimoku

from TAcharts.indicators.ichimoku import Ichimoku

i = Ichimoku(df)
i.build(20, 60, 120, 30)

i.plot()

ichimoku

Renko

from TAcharts.indicators.renko import Renko

r = Renko(df)
r.set_brick_size(auto=True, atr_interval=2)
r.build()

r.plot()

renko


wrappers

  • @args_to_dtype(dtype): Convert all function arguments to a specific data type

    from TAcharts.wrappers import args_to_dtype
    
    # Example: `src` is converted to a list
    @args_to_dtype(list)
    def rsi(src, n=2):
        pass
  • @pd_series_to_np_array: Convert function arguments from pd.Series to np.array using pd.Series.values. This wrapper is 10x quicker than using @args_to_dtype(np.array) when working with Pandas series.

    from TAcharts.wrappers import pd_series_to_np_array
    
    # Example: `high`, `low`, and `close` are all converted into `np.array` data types
    @pd_series_to_np_array
    def atr(high, low, close, n=14):
        pass