Walk-forward folds

mediumMLacceptance —~25 min

Implement walk_forward(n, folds, min_train) — expanding-window validation splits for n observations. The first min_train observations are always training; the remaining n - min_train are cut into folds contiguous test blocks of equal size, with the remainder given to the LAST block. Each fold trains on every observation before its test block. Return one row per fold: [train_start, train_end, test_start, test_end], ends inclusive.

One accuracy number is a guess; a distribution across folds is evidence. The ai quants walk-forward harness reports exactly this.

example 1

input

n=100, folds=4, min_train=20

output

[[0, 19, 20, 39], [0, 39, 40, 59], [0, 59, 60, 79], [0, 79, 80, 99]]

example 2

input

n=100, folds=3, min_train=20

output

[[0, 19, 20, 45], [0, 45, 46, 71], [0, 71, 72, 99]]

note

80 test rows into 3 folds leaves 2 over — they belong to the last fold.

constraints

folds ≥ 1
n - min_train ≥ folds
min_train ≥ 1

teacher · 2 hints · the first is free

↗ read the theory

the chapter this problem is cut from

▶ open in the workbench

run the same math on a real tape

●solution.py· utf-8 · spaces:4

def walk_forward(n: int, folds: int, min_train: int) -> list:
    """[train_start, train_end, test_start, test_end] per expanding fold."""
    size = (n - min_train) // folds
    out = []
    for f in range(folds):
        test_start = min_train + f * size
        # TODO: the last fold should run to n - 1 — the remainder rows are being dropped
        test_end = test_start + size - 1
        out.append([0, test_start - 1, test_start, test_end])
    return out

ln 1, col 1tab indent · esc leave editor

python 3.13 · pyodide

+ 6 hidden on submit

input · keyword arguments to walk_forward()

n=100, folds=4, min_train=20

expected

[[0, 19, 20, 39], [0, 39, 40, 59], [0, 59, 60, 79], [0, 79, 80, 99]]

the judge calls walk_forward() with exactly these inputs · outputs are compared to 6 decimal places

● standing bypython 3.13 in your browser · pyodide 0.28.3 · judged on the server to 6 dp⌘↵ run · ⌘⇧↵ submit

paper only · educational · not advice