Guardrail Degradation

Problem

Compute guardrail metric degradation in percent, counting only harm: max(0, (control - treatment) / control) * 100.

Return 0.0 when control is 0 or the treatment matches or beats the control.

Hint

Guardrails only care about regressions, so improvements clamp to 0.

case 1

in: [0.05,0.06]

expected: 0

case 2

in: [0.06,0.05]

expected: 16.66666666666666

case 3

in: [0.2,0.15]

expected: 25.000000000000007

case 4

in: [0,0.1]

expected: 0

Starter codePython

def guardrail_degradation(control, treatment):
    # Your code here
    pass

4 test cases · run your code against them in the editor