In reply to Damo:
This is a good point that is rarely raised. The media often come out with ridiculous stats like '1 in 10 who summit Everest die' etc which is completely inaccurate. If that was the case there would have been 60 deaths on Everest this past season. Summit/death ratio is only really useful for comparing the relative dangers of mountains, so you can see for example that K2 is statistically about 16 or 17 times as dangerous as Cho Oyu, or whatever. A summit/death rate of 25% for K2 does not mean you have a 1 in 4 chance of dying, it just shows you that it is 'pretty risky'.
As Damo mentioned, a much more accurate way of expressing the danger involved in attempting a given mountain is looking at how many people have attempted it vs number of deaths. (completely ignore summits) This shows that even attempting Annapurna with it's often quoted 40% death rate in fact only carries a 4% risk of dying. Everest is only 1.4%.(accurate as of 2006) Source:
http://www.himalayandatabase.com/downloads/HimalayaByNbrs.pdf p87
There is a slightly more up to date table of stats for all 8000ers here (still about a decade old though):
http://en.wikipedia.org/wiki/Eight-thousander
Importantly, it gives a break down of summit/death ratio for each mountain before 1990 and after 1990, so you can see a huge difference in some cases - notably Nanga Parbat with its pre 1990 77% (!) death rate, but post 1990 it is only 5.5%.
All summit/death ratios are skewed towards being higher than what the actual risk is today because climbing 8000ers has got considerably safer in recent years and they don't take into account all the people that attempt a mountain and neither summit nor die. If you have to look at summit/death ratios, then using the the post 1990 figure gives you a much more accurate picture of the dangers of a particular mountain if you were going to attempt it tomorrow.
If you take the above figure of 1.4% for overall risk of dying on Everest, then add the thousands of ascents and comparatively few deaths that have happened since 2006, it would likey come down significantly to closer to 1%. Then take into account that a huge proportion of the deaths happened pre-commercialisation and you would probably end up with an acutal risk of dying on Everest of even less than 1%, probably closer to 0.5%, if you attempt it now.