The problem with statistics is that the real world is different to your perfect statistical model. However, mathammer should still give some steer as to how to actually play the game. While you can never guarantee to kill a unit or blow something up, you can get a very good idea of which actions should (in that perfect world) give you the more favourable outcomes. Never forget the best way to get the result you want is always to roll more dice. The more shots you take, the more hits you get and the more wounds you cause. The more wounds you cause, the more saving throws your opponant has to make and the more chance they will fail. I'll talk in depth with more specific examples later on but for now I'm going to mention one of those occasions where the mathammer failed. Everyone who has ever played the game will have had a similar experience I guarantee but I'm going to discuss the time that a Scout Sergeant with a Power Fist took out Bjorn The Fell Handed.
So, he has two attacks and needs 4+ to hit. Then he needs a 6 to pen. Then Bjorn needs to fail his 5+ inv. save. Then the sergeant needs a 5+ to wreck or destroy and then another 5+ after the venerable re-roll.
P(Kill Bjorn) = Number of attacks x P(Hit) x P(Pen) x P(Fail) x P(Damage) x P(damage re-roll)
= 2 x 1/2 x 1/6 x 2/3 x 1/3 x 1/3
Very unlikely but it still happened. Always with any mathammer you might do, remember two things, statistics is nothing without context and, in the words of Arthur Conan Doyle, "Once you remove the impossible, whatever remains, no matter how improbable, must be the truth" The Scout killing Bjorn was very improbable but it was definitely the truth.