Hi Jack,

perhaps a MANOVA will do what you are looking for (https://en.m.wikipedia.org/wiki/Multivariate_analysis_of_variance).

Are your three ‘performance variables‘ outcome measures (i.e. dependent variables)?

From what I understood you want to examine the differences between bowling spells on three separate response variables. Assuming the assumptions are satisfied, you may run T tests on each of these response variables. Otherwise, a Wilcoxon rank sum test will suffice. R makes it simple to run both tests.

Hello:

I suggest that you use the tests and procedures available freely at VassarStats:

http://www.vassarstats.net/?msclkid=f7a047aac6eb11ec9726d00c4a942e90

Ideally, you should have >30 measurements per variable to have a robust design allowing sound comparisons.

If you have lower numbers, use non-parametic procedures such as the Wilcoxon sugn test available here:

http://www.vassarstats.net/?msclkid=f7a047aac6eb11ec9726d00c4a942e90