-
-
Save ultragtx/6831eb04dfe9e6ff50d0f334bdcb847d to your computer and use it in GitHub Desktop.
| # calculating RSI (gives the same values as TradingView) | |
| # https://stackoverflow.com/questions/20526414/relative-strength-index-in-python-pandas | |
| def RSI(series, period=14): | |
| delta = series.diff().dropna() | |
| ups = delta * 0 | |
| downs = ups.copy() | |
| ups[delta > 0] = delta[delta > 0] | |
| downs[delta < 0] = -delta[delta < 0] | |
| ups[ups.index[period-1]] = np.mean( ups[:period] ) #first value is sum of avg gains | |
| ups = ups.drop(ups.index[:(period-1)]) | |
| downs[downs.index[period-1]] = np.mean( downs[:period] ) #first value is sum of avg losses | |
| downs = downs.drop(downs.index[:(period-1)]) | |
| rs = ups.ewm(com=period-1,min_periods=0,adjust=False,ignore_na=False).mean() / \ | |
| downs.ewm(com=period-1,min_periods=0,adjust=False,ignore_na=False).mean() | |
| return 100 - 100 / (1 + rs) | |
| # calculating Stoch RSI (gives the same values as TradingView) | |
| # https://www.tradingview.com/wiki/Stochastic_RSI_(STOCH_RSI) | |
| def StochRSI(series, period=14, smoothK=3, smoothD=3): | |
| # Calculate RSI | |
| delta = series.diff().dropna() | |
| ups = delta * 0 | |
| downs = ups.copy() | |
| ups[delta > 0] = delta[delta > 0] | |
| downs[delta < 0] = -delta[delta < 0] | |
| ups[ups.index[period-1]] = np.mean( ups[:period] ) #first value is sum of avg gains | |
| ups = ups.drop(ups.index[:(period-1)]) | |
| downs[downs.index[period-1]] = np.mean( downs[:period] ) #first value is sum of avg losses | |
| downs = downs.drop(downs.index[:(period-1)]) | |
| rs = ups.ewm(com=period-1,min_periods=0,adjust=False,ignore_na=False).mean() / \ | |
| downs.ewm(com=period-1,min_periods=0,adjust=False,ignore_na=False).mean() | |
| rsi = 100 - 100 / (1 + rs) | |
| # Calculate StochRSI | |
| stochrsi = (rsi - rsi.rolling(period).min()) / (rsi.rolling(period).max() - rsi.rolling(period).min()) | |
| stochrsi_K = stochrsi.rolling(smoothK).mean() | |
| stochrsi_D = stochrsi_K.rolling(smoothD).mean() | |
| return stochrsi, stochrsi_K, stochrsi_D | |
| # calculating Stoch RSI | |
| # -- Same as the above function but uses EMA, not SMA | |
| def StochRSI_EMA(series, period=14, smoothK=3, smoothD=3): | |
| # Calculate RSI | |
| delta = series.diff().dropna() | |
| ups = delta * 0 | |
| downs = ups.copy() | |
| ups[delta > 0] = delta[delta > 0] | |
| downs[delta < 0] = -delta[delta < 0] | |
| ups[ups.index[period-1]] = np.mean( ups[:period] ) #first value is sum of avg gains | |
| ups = ups.drop(ups.index[:(period-1)]) | |
| downs[downs.index[period-1]] = np.mean( downs[:period] ) #first value is sum of avg losses | |
| downs = downs.drop(downs.index[:(period-1)]) | |
| rs = ups.ewm(com=period-1,min_periods=0,adjust=False,ignore_na=False).mean() / \ | |
| downs.ewm(com=period-1,min_periods=0,adjust=False,ignore_na=False).mean() | |
| rsi = 100 - 100 / (1 + rs) | |
| # Calculate StochRSI | |
| stochrsi = (rsi - rsi.rolling(period).min()) / (rsi.rolling(period).max() - rsi.rolling(period).min()) | |
| stochrsi_K = stochrsi.ewm(span=smoothK).mean() | |
| stochrsi_D = stochrsi_K.ewm(span=smoothD).mean() | |
| return stochrsi, stochrsi_K, stochrsi_D | |
@justmeonthegit, you say only numpy but as I understand it dropna() is a pandas function. I can only get it to work by doing this
delta = pd.Series(numpy.diff(series)).dropna() otherwise the line of code you originally had does not work. delta = series.diff().dropna()
check the screenshot of how vastly different the numbers are on the binance 1m chart and the values coming from this function with 1m closes plugged in. Not only that the numbers jump wildly from 80 to 50 to 10 back to 100 within a couple of closes. Maybe I am using it incorrectly, any insight would be appreciated.
If you have trouble to the first line of code ( delta = series.diff().dropna()) you are passing a wrong type of parameter to the function as it describes here.
The difference in calculation can be from a lot of things. Just check if all settings on tradingview are the same, and align with your code.
This is how I am calling the function
'r, t_1, t_2 = StochRSI(numpy.array(chistory[sym]))'
Where chistory contains the closing prices from the 1m chart for whatever ticker symbol is being iterated.
the I will put the .diff().dropna() line back in and run a breakpoint to see if I can get some more info.
Further everything I have read about dropna() says it is a pandas function, pandas is built on numpy but the functions are specific to pandas
This works for that line of code
'r, t_1, t_2 = StochRSI(pd.DataFrame(numpy.array(chistory[sym])))' also works as a pd.Series()
The numbers jump all over the shop. Check this out.
Here are the settings for the chart I am comparing it to.
Did you remove from input: all rows with timestamps who do not have any olhvc value ?
Yes, I am giving only the closing values for the last 120 periods (minutes in this case) I then append the current value every time the function is called.
This is a screenshot of the data that is being fed to the stochastic rsi function.
What should we pass in the series? The OHLC data or just open data?
You need to pass closing series (OHLC 'Close')
I was getting "NaN" as the output when passing in a list of numbers into the StochRSI function until I added the 'min_periods=0' argument to the 'rolling' functions at the bottom. Did anyone else have this problem?
after careful review of ku and binance api candle data i noticed trading view close values , for btc, are 100 to 200 off from both exchange API candle closes vs what TV charts say for the same UTC candles close price....manipulation to give edge against us and make the indicators relying on close values wrong or make the charts give basic traders the wrong idea? ...
You are awesome , this the only code that i had found similar to Trading view.
Thanks
I looked into my code. Only numpy is needed. (import numpy as np) You have to pass a series and not a pd.DataFrame. A series is only 1 column from a df.