G. ADAM wrote:The expected average result for 1 batch of 3 words

is 1 word right, same as with replacement.

ABC 3 right

ACB 1

BAC 1

BCA 0

CAB 0

CBA 1

6 trials with each sequence showing up once only = 6 hits

You are using the expected result in a very strange way. The only way you can beat your expected average is to get 3/3. Don't you see how it is the wrong approach? If you get 1/3 100% of the tme over a large sample, you will have far exceeded the expected results. If you get 0/3 100% of the time over a large sample there will also be something very strange going on. Do you see what I mean?