how to find df cells that are above/below pivot points, but only once

Question

I've got pandas dataframe with candle highs and lows. I've made new column, which shows pivot points (if the value in cell is lowest\highest between two above it an two below it, it has value 1 for lowest or 2 for highest , else 0) example:

date	high	low	pivot	takes pivot
01.01.2020	207	204	0
02.01.2020	208	205	0
03.01.2020	209	205	2
04.01.2020	207	203	0
05.01.2020	206	202	0
06.01.2020	205	200	0
07.01.2020	204	199	1
08.01.2020	206	201	0
09.01.2020	207	202	0
10.01.2020	208	205	0
11.01.2020	210	207	0
12.01.2020	212	209	0
13.01.2020	214	210	2
14.01.2020	207	204	0
15.01.2020	203	202	0
16.01.2020	201	198	0
17.01.2020	199	196	0

i need to find exact candles that made new high or low(according to pivotpoints,not all candles), something like this: for example in this table takes pivot=1 for 16.01.2020 and takes pivot=2 for 11.01.2020 only(takes pivot = 1 for lows and 2 for highs)

I've tried to make a loop, but cant really understand how exactly it should look

Answer 1

Code

rolling and use np.select

import numpy as np
cond1 = df['high'].rolling(5, center=True).max().eq(df['high'])
cond2 = df['low'].rolling(5, center=True).min().eq(df['low'])
df['pivot'] = np.select([cond1, cond2], [2, 1], 0)

df

    date    high    low pivot
0   01.01.2020  207 204 0
1   02.01.2020  208 205 0
2   03.01.2020  209 205 2
3   04.01.2020  207 203 0
4   05.01.2020  206 202 0
5   06.01.2020  205 200 0
6   07.01.2020  204 199 1
7   08.01.2020  206 201 0
8   09.01.2020  207 202 0
9   10.01.2020  208 205 0
10  11.01.2020  210 207 0
11  12.01.2020  212 209 0
12  13.01.2020  214 210 2
13  14.01.2020  207 204 0
14  15.01.2020  203 202 0
15  16.01.2020  201 198 0
16  17.01.2020  199 196 0

---

Example Code

import pandas as pd
data1 = {'date': ['01.01.2020', '02.01.2020', '03.01.2020', '04.01.2020', '05.01.2020', '06.01.2020', '07.01.2020', '08.01.2020', '09.01.2020', '10.01.2020', '11.01.2020', '12.01.2020', '13.01.2020', '14.01.2020', '15.01.2020', '16.01.2020', '17.01.2020'], 'high': [207, 208, 209, 207, 206, 205, 204, 206, 207, 208, 210, 212, 214, 207, 203, 201, 199], 'low': [204, 205, 205, 203, 202, 200, 199, 201, 202, 205, 207, 209, 210, 204, 202, 198, 196]}
df = pd.DataFrame(data1)