When Not To Pool

Two groups of developers took the same coding test after training on different stacks.

Group	n	mean score	sd
JavaScript	12	74.0	22.0
Kotlin	60	71.0	6.0

The sample variances differ by a factor of 13, so you use Welch's test rather than the pooled one. A teammate writes the degrees-of-freedom helper:

def welch_df(s1, n1, s2, n2):
    a = s1**2 / n1
    b = s2**2 / n2
    return (a + b)**2 / (a**2 / n1 + b**2 / n2)

print(round(welch_df(22.0, 12, 6.0, 60), 2))    # 12.36

The denominator is wrong. Report the correct Welch-Satterthwaite degrees of freedom for this data.

Answer format: round to 2 decimal places.