Actually, that was his original math claim. Refer to the quote.
*weary sigh*
Sorry if what I posted was confusing.
What I meant is that you can estimate that the most probable outcome will be a multiplication of the outcome of one probability roll.
If there is a 1 in 4 chance that you will kill a mage with one harpy attack then the most likely outcome of four attacks is that you will get 1 kill and the most likely outcome of 6 attacks is 1.5 kills.
Technically its not correct: no probability is ever guaranteed and indeed you are as likely to roll 1,1,1,1,1 on 5 dice as you are to roll any other sequence of numbers: that's the basis for most national lotteries by the way: any one sequence of numbers is as probable (or improbable as the case may be) to occur as any other sequence.
BUT
Commonsense says that if you roll a dice 6 times and there is a 1 in 3 chance that you will get the score you need (e.g. rolling a D6 six times, needing to get a 5+) then its fair to say that, on average, you should get the score you need twice. It's of course fully possible that you will get the score you need more than twice and equally possible that you will get the score you need less than twice. Its not guaranteed, but for the purposes of a fairly casual side hobby its probably a fair enough way to measure possible likelihoods of success. [Edit: I might add, looking back at my original post, that I did say "on average" at that stage too. No mention of 'guaranteed' at all]