Real stats on engine failures?

Howdy. I was discussing with a friend of mine my concerns about flying
single-engine planes at night or in hard IFR, due to the possibility
of engine failure. My buddy is a CFI/CFII/ATP as well as an A&P, about
3500 hours, and been around airplanes for a long time, so I tend to
give credence to his experiences. He asked me how often I thought a
piston engine had an in-flight engine failure. I guestimated once
every 10,000 hours or so. He said that was *dramatically*
over-estimating the failure rate. He said that in his experience it is
at least 40,000 to 50,000 hours per in-flight engine failure. The
place where he works sometimes as a mechanic has plenty of planes come
in for overhauls and annuals, and he estimates that for every plane
that has had an engine failure before TBO, at least 20-30 make it to
TBO without any failure (which would extrapolate to a similar figure).
The flight school he teaches at has 7 Cessnas used for primary
training and rental that have flown at least 40,000 hours total in the
six years he has been there, and they have not experienced a single
engine failure.

I emailed Lycoming, and (unsurprisingly) they told me they did not
keep records about engine failure rates.

So I'd like to find out if anyone has done any objective analysis of
certificated, piston-engine failure rates in light airplanes. I have
seen all kinds of 'guesses', but little in the way of objective facts.
After analyzing NTSB accident data and comparing to annual GA
flight-hours, I'm starting to think my friend is on the right track,
but that is a relatively small sample, and has some methodologial
flaws. It's funny. I know 20,000 hour CFIs who have never had an
engine failure, and I also know 300 hour PP-ASELs who have had engine
failures.

Just for giggles, I asked 8 pilot friends/relatives if they had ever
had an engine failure. The only 'yes' was a relative who lost an
engine after takeoff on his first solo cross-country in 1958. And I
know one other pilot who had an engine failure, who I wasn't able to
talk to.

So what is it? If the engine-failure rate is one failure for every
50,000 flight hours, I'll feel much less reticent about night/IFR
single-engine flying than if it is one in 10,000 hours. Anybody have
any facts or hard data, or have any idea where I might be able to
track some down?

Thanks,

Cap