Looking at this from another angle if statistics are to be believed then why do a lot of pilots fall outside of said statistics?
The usual error of trying to draw conclusions from statistics using only mean values. One needs higher order moments as well in order to draw any kind of meaningful conclusions from the data itself (at the very least some variance measurement). If one were to do that, I guess one would find that the variation among pilots is so high that the mean values are quite pointless.
The other big problem is that the sample sizes are too large. People try to create large sample sizes by grouping together all sorts of pilots undertaking all sorts of operations in some belief that large sample sizes make the statistics better. What one ends up with is a fairly accurate value of something nobody is interested in measuring. If I want to know, say, my probabilities of having an engine failure, I want a sample consisting of pilots similar to me doing similar things in similar planes. The proper sample size is one individual, and the result will not be available until my career is over!
The third rather obvious point is that engine failures are not random events. An engine fails for a reason. If one has a perfectly good, healthy engine that is handled properly, the chances of it failing is close to zero. If there is a fatigue crack in a rod with a remaining life of some 3000 cycles, the chances of the engine failing in the next five minutes is about 100%. Since one does not know if one belongs to the former or the latter cathegory, one fundamentally has no idea of the probability of having an engine failure during that flight. Statistics is useful for all sorts of things, but not for making predictions of individual events.