This article by Ray Charbonneau originally appeared in our Sept/Oct 2014 issue.

Conventional wisdom says that running even half-marathon splits is the best way to race a marathon. This comes from the fact that faster runners run more even splits than slower runners, as demonstrated by this scattergram that plots half-marathon splits against finish time for the 2013 Chicago Marathon. See Figure 1.

I used data from Chicago because its flat course profile reduces the impact of elevation change on splits. You’ll see a similar trend if you look at the data from Boston or any other marathon.

The thing is, while more even splits correlate with faster finish times, as we all (should) know, correlation does not imply causation.  See Figure 2.

That doesn’t mean that there is no causal relationship between splits and finish times, just because a correlation between the two exists doesn’t prove anything.

For that matter, half-marathon splits are an extremely blunt instrument to use when planning a race strategy or judging the results from a race. Two measurements from a race that covers 26.2 miles (and thousands of strides) don’t tell us much; they’re just what are readily available.

What happens if we take a closer look?

I think we can all agree that wildly different splits are bad. If someone runs a 1:30 first half and a 2:00 second half, they probably went out too fast or had an injury.

Excessively negative splits are just as bad. A runner who follows up a 2:00 first half with a 1:30 second half probably left a significant amount of time out on the course, perhaps in a portapotty.

So we start with the assumption that something closer to even splits is better. But within that narrower range, do we really know what’s best?

Ideally, to study this question we would take runners with a specific level of ability and training and have them run multiple marathons under the same conditions using different pacing strategies. Obviously, we can’t do that. Instead, let’s do a little gedankenexperiment with three runners, all of equal ability:

Runner A runs even splits, ticking off 7:45’s like clockwork. She judges her ability and the conditions perfectly, gives everything she has, and makes it to the finish without slowing down and with nothing left in the tank. Her time: 7:45 x 26.2 = 3:23:03.

Runner B decides to do the first 20 miles at a pace 15 seconds per mile slower than A. The energy she saves lets her pick up the pace and average 30 seconds per mile faster over the final 10K. Her time: (8:00 x 20) + (7:30 x 6.2) = 3:26:30.

Runner C goes out 15 seconds per mile faster than A and runs steadily for 20 miles before fatigue forces her to slow down. She doesn’t blow up entirely, but she drops 30 seconds per mile over the last 10K. Her time: (7:30 x 20) + (8:00 x 6.2) = 3:19:36.

Let’s state the obvious: Runner C is faster than both A and B.

It seems reasonable that someone who’s best possible even pace for a marathon is 7:45 could conceivably go faster using strategy C. Obviously, these numbers are just made up and they don’t prove a thing, but they certainly fit within a realistic range for a single runner’s capabilities.

Heck, let’s take someone a little faster, have her go out at our runners’ base pace (7:45) for 20 miles, then speed up by 15 seconds per mile over the last 10K. Her result: (7:45 x 20) + (7:30 x 6.2) = 3:21:30, still slower than runner C.

Think about effort instead of pace for a minute.

There’s only a Dri-FIT singlet’s thickness between the perfect race and one where you stumble across the finish on dead legs, assuming you make it that far. Your goal should be to put forth as much effort as you can over the entire 26.2 miles without flaming out.

If you aim at running an even pace the whole way, you’re always holding back a little, saving something to ensure you don’t slow down at the end. Doesn’t it seem likely that you can get more out of yourself by increasing your initial level of effort slightly, running a little more aggressively early on, and accepting a slight loss of speed as you inevitably wear down?

If you find that you’re in better shape than you expected, how would you prefer to be surprised, by holding your initial pace longer and finishing faster, or by having more left over at the end of the race so you look better in your finish line photos?

On the other hand, if you’re in worse shape than you expected, who really cares what path you take toward a slower finish than the one you’d hoped for?

In practice, the simplest way to implement this plan in a race is to figure out the even pace that would get you to your target finish time, stay a comfortable 5-20 seconds under that pace for the first 20 miles, then run your best 10K to the finish using whatever you have left, accepting that means that you’ll probably slow down some.

For those runners at the front of the pack, competing to win, tactics in reaction to other runners play a greater part in pacing (or so I’ve heard). But for the rest of us, I think we’ll do best by going out a little faster, putting seconds in the bank, and giving a little time back towards the end.

And while I know that correlation doesn’t imply causation, take a closer look at the splits for the fastest runners in the Chicago Marathon, using 5K splits instead of half-marathon splits this time. See figure 3.

Ray Charbonneau should spend more time training and less time writing if he intends to run faster than Runner C.  See more of Ray‘s writing about running at y42k.com.

To read more from the Sept/Oct 2014 issue, click here.