A package for obtaining quotation data from various sources and saving them to a database. Quotes can be quickly extracted and used for calculations and forecasts. It is possible to receive and process data in real time. There are a significant number of ready-to-use indicators. The integrity of the data stored in the database is carefully monitored.
One of the advantages of the live_trading_indicators library is the speed of work. Extracting 31 million quotes in one year on the 1s timeframe takes less than two seconds: performance test.
To calculate indicators, you can also use the Pandas Data Frame as a data source.
The current version allows you to receive exchange data from:
- Binance (spot, futures USD-M, futures COIN-M).
- Many different exchanges via CCXT (CryptoCurrency eXchange Trading Library)
The data can be obtained in numpy ndarray and Dataframe Pandas..
Package data from online sources is stored by default in the .lti folder of the user's home directory. A significant amount of data can be created in this folder, depending on the number of instruments and their timeframes. Only data received from online sources is saved.
- New indicator - Chandelier
- Fix some bugs
- New indicator - MFI
- Fix some bugs
- Change some default settings during a new installation
- The quotation database has been optimized (the conversion may take some time at the first launch)
- Fix some bugs (when ccxt is used for multiple exchanges at the same time)
- New indicator - Williams %R
- The quotation database has been optimized (the conversion may take some time at the first launch)
- New indicator - Ichimoku
- Migration of quote storage to sqlite3
- Added support for three compression algorithms: gzip, bz2 and lz4 (see)
- Add the depth parameter for ZigZag indicator
pip install live_trading_indicatorsAll the examples given here can be found in jupyter notebook examples.
import live_trading_indicators as lti
indicators = lti.Indicators('binance')
ohlcv = indicators.OHLCV('ethusdt', '4h', '2022-07-01', '2022-07-01')
print(ohlcv)<OHLCV data> symbol: ethusdt, timeframe: 4h
date: 2022-07-01T00:00 - 2022-07-01T20:00 (length: 6)
empty bars: count 0 (0.00 %), max consecutive 0
Values: time, open, high, low, close, volume
Now ohlcv contains quotes in numpy array (ohlcv.time, ohlcv.open, ohlcv.high, ohlcv.low, ohlcv.close, ohlcv.volume).
dataframe = ohlcv.pandas()
print(dataframe.head()) time open high low close volume
0 2022-07-01 00:00:00 1071.02 1117.00 1050.46 1054.52 430646.8720
1 2022-07-01 04:00:00 1054.52 1076.43 1045.41 1066.81 275557.9328
2 2022-07-01 08:00:00 1066.81 1086.44 1033.44 1050.22 252105.5665
3 2022-07-01 12:00:00 1050.21 1074.23 1043.00 1056.86 298465.0695
4 2022-07-01 16:00:00 1056.86 1083.10 1054.82 1067.91 158796.2248
import live_trading_indicators as lti
indicators = lti.Indicators('ccxt.bybit')
macd = indicators.MACD('ETHUSDT', '1h', '2022-07-01', '2022-07-30', period_short=15, period_long=26, period_signal=9)
print(macd[40:].pandas().head()) time macd signal hist
0 2022-07-02 16:00:00 -1.661969 -3.514499 1.852530
1 2022-07-02 17:00:00 -0.983912 -3.125461 2.141548
2 2022-07-02 18:00:00 -0.081701 -2.617233 2.535532
3 2022-07-02 19:00:00 0.464134 -2.064394 2.528529
4 2022-07-02 20:00:00 0.828222 -1.477419 2.305641
import pandas
import live_trading_indicators as lti
dataframe = pandas.read_csv('tests/data/ETHUSDT-1m-2022-08-15.zip', header=None)
dataframe.rename(columns={0: 'time', 1: 'open', 2: 'high', 3: 'low', 4: 'close', 5: 'volume', }, inplace=True)
indicators = lti.Indicators(dataframe)
macd = indicators.MACD(period_short=15, period_long=26, period_signal=9)
print(macd[40:].pandas().head()) time macd signal hist
0 2022-08-15 00:40:00 3.403958 2.320975 1.082984
1 2022-08-15 00:41:00 3.540428 2.643593 0.896835
2 2022-08-15 00:42:00 3.594786 2.930063 0.664722
3 2022-08-15 00:43:00 3.684476 3.170449 0.514027
4 2022-08-15 00:44:00 3.763257 3.354183 0.409074
Plotting uses matplotlib. These are optional features, so matplotlib must be installed separately. There are two methods for plotting: plot() and show(). plot() returns the drawn figure, show() returns None. For jupyter notepad, it is better to use show(), since plot() can draw a figure twice.
indicators = lti.Indicators('binance', '2022-07-01', '2022-07-15')
bb = indicators.BollingerBands('btcusdt', '4h', '2022-07-05', '2022-07-15', period=14)
bb.show()
You can find other examples of charts here.
To get real-time data, you don't have to specify an end date.
import datetime as dt
import live_trading_indicators as lti
utcnow = dt.datetime.utcnow()
print(f'Now is {utcnow} UTC')
indicators = lti.Indicators('binance', utcnow - dt.timedelta(minutes=3))
ohlcv = indicators.OHLCV('btcusdt', '1m')
print(ohlcv.pandas())Now is 2022-11-04 09:32:31.528230 UTC
time open high low close volume
0 2022-11-04 09:29:00 20594.39 20595.60 20591.06 20592.38 177.35380
1 2022-11-04 09:30:00 20592.38 20600.98 20591.75 20600.30 178.40869
2 2022-11-04 09:31:00 20600.98 20623.93 20600.30 20621.45 431.11917
To get data containing an incomplete bar, you must specify with_incomplete_bar=True when creating Indicators.
utcnow = dt.datetime.utcnow()
print(f'Now is {utcnow} UTC')
indicators = lti.Indicators('binance', utcnow - dt.timedelta(minutes=3), with_incomplete_bar=True)
ohlcv = indicators.OHLCV('btcusdt', '1m')
print(ohlcv.pandas())Now is 2022-11-04 09:37:07.372986 UTC
time open high low close volume
0 2022-11-04 09:34:00 20614.55 20618.50 20610.76 20615.97 263.96754
1 2022-11-04 09:35:00 20615.61 20624.00 20610.29 20616.53 258.53777
2 2022-11-04 09:36:00 20615.69 20617.75 20609.74 20611.46 199.43313
3 2022-11-04 09:37:00 20611.11 20611.89 20608.17 20609.02 15.15800
live-trading-indicators supports the following timeframes: 1s, 1m, 3m, 5m, 10m, 15m, 30m, 1h, 2h, 4h, 6h, 8h, 12h, 1d. The specific supported timeframes for the source depend on the source.
live-trading-indicators check the integrity of quotes when they are loaded. The fraction of lost quotes should not exceed max_empty_bars_fraction. The number of lost quotes in a row should not exceed max_empty_bars_consecutive. The values of max_empty_bars_fraction and max_empty_bars_consecutive are set to 0 by default. That is, if there is at least one lost quote, LTIExceptionTooManyEmptyBars will be raised:
live_trading_indicators.exceptions.LTIExceptionTooManyEmptyBars: Too many empty bars: fraction 0.014076769406392695, consecutive 79200. Source binance, symbol ethusdt, timeframe 1s, date 2021-01-01T00:00:00.000 - 2021-12-31T23:59:59.000.
The values of max_empty_bars_fraction and max_empty_bars_consecutive can be set as follows:
import live_trading_indicators as lti
lti.config(max_empty_bars_fraction=0.1, max_empty_bars_consecutive=10)
If you don't need integrity control at all, do:
import live_trading_indicators as lti
lti.config(max_empty_bars_fraction=-1, max_empty_bars_consecutive=-1)
The presence of the first and last bars in the date range is also checked. For more details, see Settings.
By default, log messages are output to the console, and you will see similar messages:
2022-11-04 12:32:31,528 Download using api symbol btcusdt timeframe 1m from 2022-11-04T00:00:00.000...
To disable these messages, run the following code and restart python.
import live_trading_indicators as lti
lti.config(print_log=False)When getting indicator values from online source, the first two parameters should be symbol and timeframe. Further, the period can optionally be specified. Then the parameters of the indicator are specified by name. When getting indicator values offline from Pandas DataFrame parameters symbol and timeframe are not specified.
indicators = lti.Indicators('binance', '2022-07-01', '2022-08-30')
sma = indicators.SMA('ethusdt', '1h', period=9)
macd = indicators.MACD('ethusdt', '1h', '2022-07-01', '2022-07-30', period_short=15, period_long=26, period_signal=9)dataframe = pandas.readcsv('ETHUSDT-1m-2022-08-15.zip')
indicators = lti.Indicators(dataframe)
macd = indicators.MACD(period_short=15, period_long=26, period_signal=9)
sma = indicators.SMA('2022-08-15T03:00', '2022-08-15T06:00', period=9)The list of supported indicators and their parameters can be obtained by calling lti.help(). Parameters symbol, timeframe, time_start, time_end are omitted for brevity.
import live_trading_indicators as lti
print(lti.help())- ADL(ma_period=None, ma_type='sma') - Accumulation/distribution line.
- ADX(period=14, smooth=14, ma_type='mma') - Average directional movement index.
- ATR(smooth=14, ma_type='mma') - Average true range.
- Aroon(period=14) - Aroon oscillator.
- Awesome(period_fast=5, period_slow=34, ma_type_fast='smw', ma_type_slow='sma', normalized=False) - Awesome oscillator.
- BollingerBands(period=20, deviation=2, ma_type='sma', value='close') - Bollinger bands.
- CCI(period=) - Commodity channel index.
- Chandelier(period=22, multiplier=3, use_close=False) - Chandelier Exit.
- EMA(period=, value='close') - Exponential moving average.
- Ichimoku(period_short=9, period_mid=26, period_long=52, offset_senkou=26, offset_chikou=26) - Ichimoku indicator.
- Keltner(period=10, multiplier=1, period_atr=10, ma_type='ema', ma_type_atr='mma') - Keltner channel.
- MA(period=, value='close', ma_type='sma') - Moving average of different types: 'sma', 'ema', 'mma', 'ema0', 'mma0'
- MACD(period_short=, period_long=, period_signal=, ma_type='ema', ma_type_signal='sma', value='close') - Moving Average Convergence/Divergence.
- MFI(period=14) - Money flow index.
- OBV() - On Balance Volume.
- OHLCV() - Quotes: open, high, low, close, volume.
- OHLCVM(timeframe_low='1m', bars_on_bins=6) - Quotes and the price of the maximum volume: open, high, low, close, volume, mv_price.
- ParabolicSAR(start=0.02, maximum=0.2, increment=0.02) - Parabolic SAR.
- ROC(period=14, ma_period=14, ma_type='sma', value='close') - Rate of Change.
- RSI(period=, ma_type='mma', value='close') - Relative strength index.
- SMA(period=, value='close') - Simple moving average.
- Stochastic(period=, period_d=, smooth=3, ma_type='sma') - Stochastic oscillator.
- Supertrend(period=10, multipler=3, ma_type='mma') - Supertrend indicator.
- TEMA(period=, value='close') - Triple exponential moving average.
- TRIX(period=, value='close') - TRIX oscillator.
- VWAP() - Volume-weighted average price.
- VWMA(period=, value='close') - Volume Weighted Moving Average.
- VolumeClusters(timeframe_low='1m', bars_on_bins=6) - OHLCVM and volume clusters is determined by the lower timeframe.
- WilliamsR(period=14) - Williams %R oscillator.
- ZigZag(delta=0.02, depth=1, type='high_low', end_points=False) - Zig-zag indicator (pivots).
There are three strategies for specifying a time period:
Indicator values can be obtained for any period within the interval specified for Indicators. When exiting the specified interval, an exception will be raised LTIExceptionOutOfThePeriod.
indicators = lti.Indicators('binance', 20220901, 20220930) # the base period
ohlcv = indicators.OHLCV('um/ethusdt', '1h') # the period is not specified, the base period is used
sma22 = indicators.SMA('um/ethusdt', '1h', 20220905, 20220915, period=22) # the period is specified
sma15 = indicators.SMA('um/ethusdt', '1h', 20220905, 20221015, period=15) # ERROR, going beyond the boundaries of the base periodIn this variant, when getting indicator data, the period should always be specified. When the interval is extended, data may be updated, this may slow down the work.
indicators = lti.Indicators('binance') # period not specified
ohlcv = indicators.OHLCV('um/ethusdt', '1h', 20220801, 20220815) # the period must be specified
ma22 = indicators.SMA('um/ethusdt', '1h', 'close', 22, 20220905, 20220915) # the period must be specifiedIn this variant, when creating Indicators, only the start date is specified. The data is always received up to the current moment. When creating Indicators, you can specify with_incomplete_bar=True, then the data of the last, incomplete bar will be received. See the example above.
- For the spot market, they completely coincide with the code on binance (btcusdt, ethusdt, etc.)
- For the futures market USD-M, codes are prefixed with um/ (um/btcusdt, um/ethusdt, etc.)
- For the futures market COIN-M, codes are prefixed with cm/ (cm/btcusd_perp, cm/ethusd_perp, etc.)
Using CCXT, you can download data from a large number of exchanges, currently there are more than 100. The available symbols, their names and timeframes depend on the specific source. More information can be found in the CCXT documentation. The use of ccxt is optional, so it must be installed separately. It can be done like this:
pip install ccxt
Then you can use all available ccxt exchanges by specifying them through a dot. To download, for example, from binance via ccxt, you need to specify ccxt.binance. To download from okx, we use ccxt.okx, Bybit - ccxt.bybit, etc.
indicators = lti.Indicators('ccxt.okx')
ohlcv = indicators.OHLCV('BTC/USDT', '1h', 20220701, 20220702)live-trading-indicators has not been tested with all quotation sources supported by ccxt. If you find a problem with some data source, open the problem here.
Sometimes the ccxt source may need additional parameters passed through params. In this case, these parameters are passed via exchange_params when creating Indicators:
indicators = lti.Indicators('ccxt.okx', exchange_params={'limit': 300})live-trading-indicators supports the following types of moving averages:
- 'sma' - simple move average
- 'ema' - classical exponential moving average with alpha = 2 / (n + 1), initialized by SMA (as in binance EMA)
- 'ema0' - classical exponential moving average with alpha = 2 / (n + 1), initialized by the first value
- 'mma' - Modified moving average with alpha = 1 / n, initialized by SMA (as in some binance indicators)
- 'mma0' - Modified moving average с alpha = 1 / n, initialized by the first value
The settings can be obtained as dict using config():
import live_trading_indicators as lti
print(lti.config())Result:
{'cache_folder': '/home/user/.lti/data/timeframe_data', 'sources_folder': '/home/user/.lti/data/sources', 'log_folder': '/home/hal/.lti/logs', 'endpoints_required': True, 'max_empty_bars_fraction': 0.0, 'max_empty_bars_consecutive': 0, 'restore_empty_bars': True, 'print_log': True, 'log_level': 'INFO', 'request_timeout': 10, 'request_trys': 3}
config() is also used to change the settings:
import live_trading_indicators as lti
lti.config(request_timeout=15)When creating Indicators, you can specify the settings that will be used instead of the saved ones:
indicators = lti.Indicators(test_source, time_begin, time_end, timeout=15, request_trys=5)Directory for storing quotation data.
Directory of log files.
Control of the presence of the first and last bar in the selected date range. In the absence of the first or last bar, LTIExceptionQuotationDataNotFound is raised. Default: False.
The maximum fraction of lost bars, if exceeded, an error will occur. Default: 1 (100% empty bars are allowed).
The maximum number of lost bars in a row, if exceeded, LTIExceptionTooManyEmptyBars will be raised. Default: -1 (any number of empty bars in a row is allowed).
If True, it restores the lost bars (open=close=close of the previous one, volume=0). The control of the number of lost bars (max_empty_bars_fraction, max_empty_bars_consecutive) is performed BEFORE recovery. Default: True.
If True, outputs log messages to standard output. Default: True.
Log registration level. Default: INFO.
Timeout of requests to download quotes, seconds. Default: 30.
The number of attempts to download quotes. Default: 3.
Path to the sqlite3 database for storing quotes.
The algorithm for compressing quotes when saving to the database. Can be gzip, bz2, lz4 and auto:
- bz2 - good compression, slow
- gzip - medium compression, medium speed
- lz4 - low compression, high speed
- auto - automatic selection
Default: auto