Sampling and Experiments · Challenge problem

A randomized experiment puts $3$ subjects in treatment and $3$ in control. The treatment outcomes are $8, 9, 10$ and the control outcomes $4, 5, 6$. If the treatment had no effect, every split of these six outcomes into groups of $3$ would be equally likely. What is the probability that a random split gives a treatment-minus-control difference in means at least as large as the one observed?