I'm trying to rewrite a script to Python, but I can't figure out how pivothigh() and pivotlow() function work, and I can't find source code, I know how to calculate Pivot Points, but what leftbars and rightbars means in this two function? Please help.
Leftbars and rightbars are number of bars that the pivot
functions looks when is searching for a pivot.
For example:
pivothigh(10,10)
will search for high price that was not exceeded during 10 bars to the left (past data) and 10 bars to the right (future data). Note that the function won't be able to determine the pivot if there is less than 10 bars to the right.
I too had a need to better understand how the pivothigh()
and pivotlow()
functions work internally so I made the effort to code a Pine Script version (using version 5 of Pine Script) for myself and tested it side-by-side with the ta.pivotlow()
and ta.pivothigh()
functions and it seems to work well. Maybe this will help you as well.
my_pivothigh(float _series = high, int _leftBars, int _rightBars) =>
float _pivotHigh = na
int _pivotRange = ( _leftBars + _rightBars )
float _leftEdgeValue = nz(_series[_pivotRange], na)
if not na(_series) and _leftBars > 0 and _rightBars > 0 and not na(_leftEdgeValue)
float _possiblePivotHigh = _series[_rightBars]
float[] _arrayOfSeriesValues = array.new_float(0)
for _barIndex = _pivotRange to 0
array.push(_arrayOfSeriesValues, _series[_barIndex])
//end for
int _pivotHighRightBars = array.size(_arrayOfSeriesValues) - array.lastindexof(_arrayOfSeriesValues, array.max(_arrayOfSeriesValues)) - 1
_pivotHigh := ( _pivotHighRightBars == _rightBars ) ? _possiblePivotHigh : na
//end if
_pivotHigh
my_pivotlow(float _series = low, int _leftBars, int _rightBars) =>
float _pivotLow = na
int _pivotRange = ( _leftBars + _rightBars )
float _leftEdgeValue = nz(_series[_pivotRange], na)
if not na(_series) and _leftBars > 0 and _rightBars > 0 and not na(_leftEdgeValue)
float _possiblePivotLow = _series[_rightBars]
float[] _arrayOfSeriesValues = array.new_float(0)
for _barIndex = _pivotRange to 0
array.push(_arrayOfSeriesValues, _series[_barIndex])
//end for
int _pivotLowRightBars = array.size(_arrayOfSeriesValues) - array.lastindexof(_arrayOfSeriesValues, array.min(_arrayOfSeriesValues)) - 1
_pivotLow := ( _pivotLowRightBars == _rightBars ) ? _possiblePivotLow : na
//end if
_pivotLow
I know this is an old post but I've made a very simple python implementation that anyone can build on, it does the same thing as Pine Scripts ta.pivot functions.
code:
def pivots_high(data, LBR, LBL):
pivots = []
for i in range(len(data)-LBR):
pivots.append(0)
pivot = True
if i > LBL:
for j in range(LBL + 1):
if data[i - j] > data[I]: # do if data[i - j] < data[i] for pivot low
pivot = False
for j in range(LBR + 1):
if data[i + j] > data[I]: # do if data[i + j] < data[i] for pivot low
pivot = False
if pivot is True:
pivots[len(pivots)-1] = data[i]
for p in range(LBR):
pivots.append(0) # This is so the pivots length matches your data length
return pivots # The Pivots will be any value that is not 0 and it will be where the lowest/highest value is
the lookback variables simply means that if you take one price point and look n(Looback left) candles left and n(lookback right) candles right, and still its the lowest/highest then that's the pivot
UPDATE
I ran into some problems with this code with different combinations of numbers on large datasets so I had to completely change it to always match up with tradingview.
New Code:
def checkhl(data_back, data_forward, hl):
if hl == 'high' or hl == 'High':
ref = data_back[len(data_back)-1]
for i in range(len(data_back)-1):
if ref < data_back[i]:
return 0
for i in range(len(data_forward)):
if ref <= data_forward[i]:
return 0
return 1
if hl == 'low' or hl == 'Low':
ref = data_back[len(data_back)-1]
for i in range(len(data_back)-1):
if ref > data_back[i]:
return 0
for i in range(len(data_forward)):
if ref >= data_forward[i]:
return 0
return 1
def pivot(osc, LBL, LBR, highlow)
left = []
right = []
for i in range(len(osc)):
pivots.append(0.0)
if i < LBL + 1:
left.append(osc[i])
if i > LBL:
right.append(osc[i])
if i > LBL + LBR:
left.append(right[0])
left.pop(0)
right.pop(0)
if checkhl(left, right, highlow):
pivots[i - LBR] = osc[i - LBR]
return pivots
then just do:
pivots_low = pivot(data, lbl, lbr, 'low')
pivots_high = pivot(data, lbl, lbr, 'high')
output is pivots in their actual location with 0.0s otherwise.
You could construct something similar with a one-liner with pandas:
pivots = high_column.shift(-len_right, fill_value=0).rolling(len_left).max()
For 'High' pivots pd.Series 'high_column' and:
pivots = low_column.shift(-len_right, fill_value=0).rolling(len_left).min()
for the 'Lows'. It avoids using loops and is a fast vectorized function.
This is my implementation of ta.pivotHigh for javascript, hope this helps.
const pivotHigh = (series, period) => {
let ph = 0;
let phIndex = 0;
// left + right bars + 1 pivot bar
for ( let i = period + period + 1, len = series.length; i--; ) {
const cur = series[len - i];
// [!] > -1 logic. can also checks: NaN
if ( cur > -1 ) {} else {
break;
}
if ( cur > ph ) {
ph = cur;
phIndex = len - i;
}
}
// found?
return phIndex === period
? ph
: 0;
};
usage:
const series = [0,1,2,3,4,5,4,3,2,1,0];
const period = 5;
const ph = pivotHigh(series, period);
Here is my humble approach
import numpy as np
from talib import MAX, MIN
def PIVOTHIGH(high: np.ndarray, left:int, right: int):
pivots = np.roll(MAX(high, left + 1 + right), -right)
pivots[pivots != high] = np.NaN
return pivots
def PIVOTLOW(low: np.ndarray, left:int, right: int):
pivots = np.roll(MIN(low, left + 1 + right), -right)
pivots[pivots != low] = np.NaN
return pivots
Use them like this:
a = np.array([1, 1, 2., 2., 2., 3., 4., 5., 6., 7., 8., 9., 8., 7., 6., 7., 8., 9., 10, 9., 8., 9., 10])
PIVOTHIGH(a, 2, 2)
#output
array([nan, nan, 2., nan, nan, nan, nan, nan, nan, nan, nan, 9., nan,
nan, nan, nan, nan, nan, 10., nan, nan, nan, nan])
PIVOTLOW(a, 2, 2)
#output
array([nan, nan, nan, nan, 2., nan, nan, nan, nan, nan, nan, nan, nan,
nan, 6., nan, nan, nan, nan, nan, 8., nan, nan])
I've tried to create a simple version of it in Pine Script, which does not uses pivothigh/pivotlow - instead does the candlestick comparison.
https://www.tradingview.com/script/BYHsrYPG-Broken-Fractal-Someone-s-broken-dream-is-your-profit
I was also able to convert this into Ruby code (Python code should be as easy).
if (candles[i-1][:h] > candles[i-2][:h]) and (candles[i-1][:h] > candles[i][:h])
puts "DownFractal"
end
if (candles[i-1][:l] < candles[i-2][:l]) and (candles[i-1][:l] < candles[i][:l])
puts "UpFractal"
end
Here's my Python implementation. Works exactly as ta.pivothigh
and ta.pivotlow
.
Pivot high:
def get_pivot_high(ohlcvs: List[OHLCV], left_bars: int, right_bars: int, key_name: str = 'high_price') -> Optional[float]:
if len(ohlcvs) < left_bars + right_bars:
return None
highest_value = max(ohlcv.get(key_name) for ohlcv in ohlcvs[-(left_bars + right_bars + 1):])
return highest_value if highest_value == ohlcvs[-right_bars].get(key_name) else None
Pivot low:
def get_pivot_low(ohlcvs: List[OHLCV], left_bars: int, right_bars: int, key_name: str = 'low_price') -> Optional[float]:
if len(ohlcvs) < left_bars + right_bars:
return None
lowest_value = min(ohlcv.get(key_name) for ohlcv in ohlcvs[-(left_bars + right_bars + 1):])
return lowest_value if lowest_value == ohlcvs[-right_bars].get(key_name) else None
Oneliner same as the tradingview Pivots HL:
LEN = 50 #Lookback and Lookforward
OHLC['PivotHigh'] = OHLC['high'] == OHLC['high'].rolling(2 * LEN + 1, center=True).max()
OHLC['PivotLow'] = OHLC['low'] == OHLC['low'].rolling(2 * LEN + 1, center=True).min()
(Left and right will be the same length)
This Python function was designed to precisely match the logic of the pivotlow and pivothigh functions in Pine Script version 4. It has been successfully tested on large datasets. Previous answers in this thread did not provide me with an exact match to the Pine Script logic.
# This function is adapted to work with a pandas DataFrame.
# The function takes as input a DataFrame with OHLCV data, the index
# of the current candle - 1, and the window size for checking, prd.
# The function returns the pivot value if found, otherwise None.
# It is assumed that the DataFrame has columns "high" and "low"
# which contain the highs and lows of the candle, respectively.
def find_pivot_highs(df, index, prd):
# Create a window of size (prd * 2 + 1) around the current index
# Extract values from the "high" column of DataFrame df
window = df["high"].iloc[index - prd * 2 : index + 1].values
# Find the maximum value in the last prd elements of the window
high_max = max(window[-prd:])
# Find the maximum value in the entire window
max_value = max(window)
# Check if the current value is the maximum in the window
# and if it's greater than the maximum value in the last prd elements
if max_value == window[prd] and window[prd] > high_max:
return window[prd]
return None
def find_pivot_lows(df, index, prd):
# Create a window of size (prd * 2 + 1) around the current index
# Extract values from the "low" column of DataFrame df
window = df["low"].iloc[index - prd * 2 : index + 1].values
# Find the minimum value in the last prd elements of the window
low_min = min(window[-prd:])
# Find the minimum value in the entire window
min_value = min(window)
# Check if the current value is the minimum in the window
# and if it's less than the minimum value in the last prd elements
if min_value == window[prd] and window[prd] < low_min:
return window[prd]
return None
I had found this thread after I have searched for this kind of implementation. Here is my own implementation for those that have been utilizing the Binance API. (Written in java)
From my own testings, it has the same results as the pine script.
private boolean checkHighOrLow(Candlestick candlestick , int lengthForCheck, int currentCandleIndex, boolean checkForHigh) {
double currentCandleStickClosePrice = Double.parseDouble(candlestick.getClose());
for (int i = 0; i < lengthForCheck; i++) {
double afterCandleStick = Double.parseDouble(candlestickList.get(currentCandleIndex + i + 1).getClose());
double beforeCandleStick = Double.parseDouble(candlestickList.get(currentCandleIndex - i - 1).getClose());
if(checkForHigh) {
if (afterCandleStick > currentCandleStickClosePrice)
return false;
if (beforeCandleStick > currentCandleStickClosePrice)
return false;
}else{
if(afterCandleStick < currentCandleStickClosePrice)
return false;
if(beforeCandleStick < currentCandleStickClosePrice)
return false;
}
}
return true;
}
public void findHighsAndLows(){
int lengthForCheck = 1;
int numOfCandles = candlestickList.size();
for(int i = lengthForCheck; i < numOfCandles - lengthForCheck; i ++)
{
Candlestick currentCandle = candlestickList.get(i);
if(checkHighOrLow(currentCandle,numOfCandles,lengthForCheck,i,true))
highs.add(currentCandle);
if(checkHighOrLow(currentCandle,numOfCandles,lengthForCheck,i,false))
lows.add(currentCandle);
}
}
Logic still applies. Enjoy
Result:
FOUND LOW | Wed Aug 25 04:20:00 IDT 2021
FOUND HIGH | Wed Aug 25 05:05:00 IDT 2021
FOUND LOW | Wed Aug 25 05:20:00 IDT 2021
FOUND HIGH | Wed Aug 25 05:30:00 IDT 2021
FOUND LOW | Wed Aug 25 05:35:00 IDT 2021
FOUND HIGH | Wed Aug 25 05:45:00 IDT 2021
FOUND LOW | Wed Aug 25 06:15:00 IDT 2021
FOUND HIGH | Wed Aug 25 06:25:00 IDT 2021
FOUND LOW | Wed Aug 25 06:35:00 IDT 2021
FOUND HIGH | Wed Aug 25 06:40:00 IDT 2021
FOUND LOW | Wed Aug 25 06:55:00 IDT 2021
FOUND HIGH | Wed Aug 25 07:05:00 IDT 2021
FOUND LOW | Wed Aug 25 07:25:00 IDT 2021
FOUND HIGH | Wed Aug 25 07:45:00 IDT 2021
FOUND LOW | Wed Aug 25 07:50:00 IDT 2021
FOUND HIGH | Wed Aug 25 08:20:00 IDT 2021
FOUND LOW | Wed Aug 25 08:25:00 IDT 2021
FOUND HIGH | Wed Aug 25 08:35:00 IDT 2021
FOUND LOW | Wed Aug 25 08:45:00 IDT 2021
FOUND HIGH | Wed Aug 25 08:50:00 IDT 2021
FOUND LOW | Wed Aug 25 09:15:00 IDT 2021
FOUND HIGH | Wed Aug 25 09:30:00 IDT 2021
FOUND LOW | Wed Aug 25 09:35:00 IDT 2021
FOUND HIGH | Wed Aug 25 09:40:00 IDT 2021
FOUND LOW | Wed Aug 25 09:55:00 IDT 2021
FOUND HIGH | Wed Aug 25 10:00:00 IDT 2021
FOUND LOW | Wed Aug 25 10:05:00 IDT 2021
FOUND HIGH | Wed Aug 25 10:15:00 IDT 2021
FOUND LOW | Wed Aug 25 10:45:00 IDT 2021
FOUND HIGH | Wed Aug 25 10:50:00 IDT 2021
FOUND LOW | Wed Aug 25 11:15:00 IDT 2021
FOUND HIGH | Wed Aug 25 11:20:00 IDT 2021
FOUND LOW | Wed Aug 25 11:35:00 IDT 2021
FOUND HIGH | Wed Aug 25 11:45:00 IDT 2021
FOUND LOW | Wed Aug 25 11:55:00 IDT 2021
FOUND HIGH | Wed Aug 25 12:15:00 IDT 2021
© 2022 - 2024 — McMap. All rights reserved.