Quantile outputs → signal

mediumMLacceptance —~20 min

Implement quantile_signal(p10, p50, p90) — turn a quantile-regression forecast into a position for each asset: 1 (long) when the 10th percentile of the forecast return is above zero, -1 (short) when the 90th percentile is below zero, else 0 (flat). The three inputs are lists of equal length.

A median forecast alone says nothing about conviction; the interval does. This is the exact rule the workbench's quantile model uses to commit. Return the list of signals.

example 1

input

p10=[0.01, -0.02, 0.005], p50=[0.03, 0.0, 0.02], p90=[0.05, 0.02, 0.04]

output

[1, 0, 1]

example 2

input

p10=[0.01, -0.06, -0.02], p50=[0.03, -0.03, 0.0], p90=[0.05, -0.01, 0.02]

output

[1, -1, 0]

constraints

1 ≤ len(p10) == len(p50) == len(p90) ≤ 1_000
p10 ≤ p50 ≤ p90 elementwise

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 quantile_signal(p10: list, p50: list, p90: list) -> list:
    """1 long, -1 short, 0 flat — one signal per asset."""
    out = []
    for lo, mid, hi in zip(p10, p50, p90):
        if lo > 0:
            out.append(1)
        # TODO: the short side — when the whole upper tail (p90) is below zero
        else:
            out.append(0)
    return out

ln 1, col 1tab indent · esc leave editor

python 3.13 · pyodide

+ 6 hidden on submit

input · keyword arguments to quantile_signal()

p10=[0.01, -0.02, 0.005], p50=[0.03, 0.0, 0.02], p90=[0.05, 0.02, 0.04]

expected

[1, 0, 1]

the judge calls quantile_signal() 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