Just a suggestion - extend rolling to support a rolling window with a step size, such as R's rollapply(by=X).

Code Sample

Pandas - inefficient solution (apply function to every window, then slice to get every second result)

import pandas
ts = pandas.Series(range(0, 40, 2))
ts.rolling(5).apply(max).dropna()[::2]

Suggestion:

ts = pandas.Series(range(0, 40, 2))
ts.rolling(window=5, step=2).apply(max).dropna()

Inspired by R (see rollapply docs):

require(zoo)
TS <- zoo(seq(0, 40, 2))
rollapply(TS, 5, FUN=max, by=2)

8 12 16 20 24 28 32 36 40