Chart Patterns

The price turns up from 100 three times (days 4, 12 and 20). Grouping the swing lows to the nearest point gives one level with three touches: support at 100. A close 1% below it, under 99, would count as a break; day 24 closes at 106, so it holds.

• Levels are zones, not lines: how close counts as a touch is a parameter, and changing it changes the levels you find.
• Every test also uses up the buyers waiting there, so heavily tested levels often break.
• Round numbers and old highs work partly because many traders watch them, a self-fulfilling effect that can stop at any time.

def swings(s, n=3):                     # a swing high (low) is the extreme of the 2n+1 bars around it
    w = s.rolling(2 * n + 1, center=True)
    return s[s == w.max()], s[s == w.min()]
c = df['close']
highs, lows = swings(c)
touches = lows.round(0).value_counts()               # swing lows grouped to the nearest point
level = touches.idxmax()
support = level if touches.max() >= 2 else None
broken = c.iloc[-1] < level * 0.99                    # a close 1% under it

Assumes import numpy as np, import pandas as pd, and a DataFrame df with columns open, high, low and close for the data above.