You can always compare the two groups. However you should be fully transparent on the criteria chosen.

Assume sets A, B, and C have void intersection. Define D as the union of A, B, and C

You may want to normalise the values for members of A or B by the average value on D, etc

Now 10 and 40 are not large numbers... You get clouds of points and possible maximum likelihood fittings rather than massive (and significant) statistics..

1) You can use ANOVA and conduct a mean group comparison test, where group classification is determined by investment strategy (0 = "Other", 1 = Dogs of the Dow).

2) You can use regression analysis and include investment strategy (0 = "Other", 1 = Dogs of the Dow) as a binary independent variable.

50 observations is not a large sample, which may reduce your ability to reject null hypotheses.

Since your problem is a bit rough and not perfectly conditioned for classical statistics, here is something you may try. It's bottom up and fully data driven, but it seems robust:

each company is described by data (financial and other KPIs and parameter). With these data, you form for company i vector v(i).

You have two sets to compare, A and B. For each set say C (which can be either A or B), you study the autocorrelation matrix M formed of m(i, j) = which is the scalar product of parameters (data) for companies i an j, this is statistically an intercorrelation.

Matrix M is a Gram-Schmidt matrix, hence it diagonalises with eigenvalues lambda(k) with k=1,...,Card(C)

with Card(C) denoting the number of companies in set C.

Compare sets A and B by looking at their eigenvalues : each of them, or the min, the max, the average, the median, as you see fit from the experimental values you have.

This will give you a rough but deep structural comparison on how these set of companies perform.

Does it work for your actual data?

If not try some normalisation of the parameters (ratio of individual parameter to an industry average, or set-wide average, etc).

You can also introduce weights...

Looking forward to hear from you