I've been thinking about the Afton Trail Race a lot lately. This year's winner, Mike Borst, ran an extremely fast first half and died in the second, still just three minutes off the course record (he claims a pit stop cost him the record, but the uneven pacing is more likely to be the cause). Not many people run anything close to even pace in that race, though it's generally accepted that even pacing is the most efficient way to race. The course starts before the heat of the day and heats up rapidly, so it's easy to run fast for the first hour or two - but is it better to run fast when you can do it easily and then slow, or to run evenly? I've gone back and forth on this, but cumulative heat stress effects tell me that even pacing is still best.
Not that I'm exactly an expert at even pacing... my official best mile had an insanely fast first 400 and a wicked kick, with a plodding 800 in the middle. My second-best marathon had a first half in 1:14:29, with a finish of 2:43:24.
There was a study that showed that 5K's might be best run with positive splits, but they're an unusual case (and not coincidentally my best distance). Top runners can run anaerobically, with increasing lactate levels, for about as long as they can hold their breath - about 3 minutes (not many average runners can do much more than a minute). 3 minutes is a sizable portion of a 13 minute 5K and slowing that last 3 minutes is reasonable. For a 1500m runner, 3 minutes is almost the whole race, so it doesn't alter first and second halves. At 10K, the effect is lessened and the problem of heat dissipation enters, so 10K runners vary from positive to even to negative splits, depending on the runner's strengths and the day. 100m sprinters spend the first part of the race getting to speed, so they almost always run negative splits. Everyone else: even splits.
I may just have to see if I can get in shape so I can try running even pace there next year. 50 weeks to go...