Saturday, June 6, 2015

Pacing an Ultra


When I started running ultramarathons not so long ago I was surprised at how difficult it was to find practical information on choosing an appropriate pace. For races from the mile to the marathon, it is well established how performances compare across distances, and it is easy to find tables and online calculators that help you choose an appropriate pace for a marathon, say, based on your 5k time. But if you’re running a 100k, 100-miler or 24-hr race for the first time, choosing an appropriate goal pace is not so easy. Aphorisms are pretty easy to find, and many of them, like “start slow, then slow down” I’ve found to be surprisingly effective. But quantitative information on establishing a reasonable goal time (or distance) is hard to come by. The same is true for information on race pacing strategies.

So, here is an attempt to provide some quantitative guidance on how to pace yourself during an ultra. There are, of course, a number of factors other than distance that can have a large influence on your pace, particularly if it’s run on trails. Topography, trail conditions, heat/humidity, elevation and other factors can combine to slow you by up to a factor of two or more in some of the most difficult races, even if all goes well. In most races it is absolutely critical to take these factors into account. These are discussed below, but to start it’s useful to consider the baseline case of a flat road or track ultra (even those these races aren’t currently in vogue).

Setting a Goal Time

Figure 1. Race distance vs. finish time, mile to marathon.

    A widely used algorithm for predicting race times for distances between the mile and the marathon is based on Jack Daniels’ VDOT calculator, where VDOT values essentially correspond to an individual’s level of talent and fitness. Figure 1 shows calculated finish times across a range of distances for a few different VDOT values, along with world record times for men and women. Note that each axis on this figure is plotted with a logarithmic scale, on which an equal spacing between ticks indicates a doubling of the units. Plotting the data in this way is useful because the points form nice straight lines. The important take-away from Fig. 1 is that – woman or man, world-beater or back-of-the-packer – the average pace falls off at about the same rate as the distance increases. Of course there is individual variation – some people are better at marathons, others at the mile – but in general it’s a good rule of thumb that your average race pace will fall off by about 5% for every doubling of the distance. This is the basis for calculators that allow you to set a reasonable goal for your first marathon, based on your race times at shorter distances.

Figure 2. Running world records (women - blue; men - black)
So, can you simply extrapolate from Fig. 1, and assume that your pace will continue to fall off at the same rate beyond the marathon? Sadly, no. As shown in Fig. 2 to the right, an extrapolation of the world record lines for distances of 1500 m to a marathon significantly underestimates finish times at ultra distances. The longer the race, the greater the mismatch. (I know it doesn’t look like a huge misfit on this plot – but it really is!) Why the mismatch? One difference is that, generally speaking, the shorter distance races are competed at a higher level than ultras – there are simply more athletes competing at these distances, and more of the top athletes are able to devote full time to their running careers than at ultra distances. I’m sure this plays some role, but it can’t plausibly explain the mismatch. At 100 miles, the extrapolation underpredicts the actual world records by 3-4 hours – a woman would need to average 5:46/mi over 100 miles, and a man 5:11/mi! I’m quite confident this is not humanly possible, regardless of the training regime (or doping regime, for that matter – not even Lance Armstrong levels of PEDs can buy that kind of speed in a 100-miler!). Sure, if there were more money and more competition in ultras, the records would come down a bit, and the record time vs. race distance trends would smooth out. But I don’t think the changes are likely to be huge – and they certainly won’t be enough to bring the ultra records onto the middle-to-long distance world record curve.

         
Figure 3. World record pace vs. distance (blue - women; black - men)
  
To understand the origin of the mismatch, it’s useful to look at the data in a slightly different way  – in terms of the average pace for different distances (Fig. 3). This plot shows that ultras are in a different regime than middle-distance to long-distance races – just as middle- to long-distance races are in a different regime than sprints. In contrast to the relatively small change in pace between the mile and the marathon – slowing by 5% per doubling in distance – for ultramarathons the pace slows about 16.5% for every doubling in race distance.

Why do these different running regimes exist? I have no expertise in exercise physiology, and I suspect many factors are involved, but one clear difference among the regimes is the dominant metabolic process. Sprints are anaerobic. As the race distance increases there is a gradual transition to aerobic metabolism. For middle- and long-distance events, glycogen stored in the muscles is the predominant source of aerobic fuel, but stored glycogen alone can only fuel a race of approximately the marathon distance. Beyond the marathon, the body must rely increasingly on its fat stores as a fuel source, and on carbohydrates, fat and protein that are consumed during the race. These are metabolized slowly compared to muscle glycogen –if the fuel doesn’t burn as fast, the machine doesn’t run as fast.

So, how do you choose an appropriate goal time for an ultra race distance you’ve never run? It’s probably a reasonable assumption that your own (potential) performance varies with distance in a way that is not too dissimilar from the trend of world record performances. This seems to hold fairly well for middle- to long-distances (see Fig. 1), and there’s no reason to think it doesn’t hold for ultra-distances also. As a general rule of thumb, you could anticipate that the ratio of your performance to the world record performance is relatively constant across distances. For example, let’s say you’ve run a marathon in 3.5 hours and want to try a 50-miler. Your marathon time is 71% longer than the men’s world record time, and you could anticipate that it’s about 71% longer at 50 miles as well. Although world records are not tracked officially for the 50 mile distance, the trend shown in Fig. 3 gives a men’s world-record-equivalent pace for this distance of 3.65 min/km, which corresponds to a time of 4:52. So, you might consider a goal of around 8:20 for the 50-miler – IF it’s on a course similar to your marathon.

What About Trail Ultras?

Of course, it’s rather unlikely that the 50-miler is on flat terrain with good footing. Unlike marathons, most ultras these days (in contrast to 2-3 decades ago) are run on trails, many of them on technical trails with serious elevation gain (and loss). If your 50-miler is on this sort of course, it would be a major mistake to stick with your 8:20 plan – it’s going to take you a lot longer. But how much longer? Unfortunately, there is no simple way to account for all of the variables that come into play in a trail race. Your best bet is to rely on the performance of others who have run your target race, whose abilities you can infer from their performances in other races you are familiar with, or have run yourself.

Based on such inter-race comparisons, the folks at realendurance.com have calculated the relative finish times for a number of North American ultras, focusing primarily on 100 mile races. I am not sure exactly what their methods are, but in essence their calculations are based on a dataset for individuals who have run multiple ultras in the same year. The relative finish times for individuals in these races are averaged in some way, and normalized to the finish times for the Western States 100. The table on their site provides a useful baseline to estimate your time in a race you’ve never run, provided you’ve run at least one of the races on the list. However, if you haven’t run any of the races on the list, it’s necessary to have some other point of comparison.

Below (Table 1) I’ve made an attempt to re-calibrate the realendurance.com list of hundred-mile races to the road and track world record curve. My assumption is that the course records for the Western States 100 (14:46:44 for men, 16:47:19 for women) represent performances that are comparable to the world records for 100 miles on track. This seems reasonable, given the level of competition at WS100. The WS100 course record times are 28.8% and 21.7% longer than the corresponding track world record times for men and women, respectively, suggesting that the WS100 course is ~22-29% more difficult than a track. The higher value (based on the men’s records) seems somewhat more reasonable. On one end of the spectrum, it gives track-world-record equivalent times that are comparable to the actual track world record, for races that are run on flat non-technical courses. On the other end, it gives a track-world-record equivalent time for the Hard Rock 100 of 23:30, which is close to but actually a bit slower than the course record of 22:41 set by Kilian Jornet in 2014. This suggests that Jornet’s performance at HR100 may actually be superior to the track world record, or the WS100 world record held by Timothy Olson (which seems plausible enough). Similarly, on the women’s side, Ellie Greenwood’s course record at WS100 appears to be superior to Ann Trason’s earlier track 100 world record, since calibration to Greenwood’s WS100 record yields track-world-record equivalent times for the flatter races in the realendurance.com database that are ~9% faster than the actual world record.

Table 1. Relative difficulty of various North American 100-mile events, with estimated finish times that are equivalent performances to the track 100 mi world records. Based on inter-race comparisons by realendurance.com
Event
Time factor relative to WS
Time factor relative to track 100 mi WR (M)
Estd. time (hr) equivalent to track 100 mi WR (M)
Time factor relative to track 100  mi WR (F)
Estd. time (hr) equivalent to track 100 mi WR (F)
Hard Rock
1.59
2.05
23.48
1.94
26.69
HURT
1.28
1.65
18.91
1.56
21.49
Susitna
1.28
1.65
18.91
1.56
21.49
Plain
1.27
1.64
18.76
1.55
21.32
Coyote Two Moon
1.21
1.56
17.87
1.47
20.31
Grand Mesa
1.20
1.55
17.72
1.46
20.15
Wasatch
1.19
1.53
17.58
1.45
19.98
Mt. Rushmore
1.15
1.48
16.99
1.40
19.31
Massanutten
1.13
1.46
16.69
1.38
18.97
Bear
1.12
1.44
16.54
1.36
18.80
Bighorn
1.12
1.44
16.54
1.36
18.80
Superior Sawtooth
1.11
1.43
16.39
1.35
18.63
Grindstone
1.11
1.43
16.39
1.35
18.63
Pine to Palm
1.10
1.42
16.25
1.34
18.47
Angeles Crest
1.09
1.40
16.10
1.33
18.30
Tahoe Rim
1.09
1.40
16.10
1.33
18.30
Moab
1.07
1.38
15.80
1.30
17.96
Grand Teton
1.07
1.38
15.80
1.30
17.96
Eagle
1.05
1.35
15.51
1.28
17.63
Ozark Trail
1.04
1.34
15.36
1.27
17.46
Cascade Crest
1.04
1.34
15.36
1.27
17.46
Leadville
1.04
1.34
15.36
1.27
17.46
Cactus Rose
1.02
1.31
15.07
1.24
17.12
McNaughton Park
1.01
1.30
14.92
1.23
16.96
Western States
1.00
1.29
14.77
1.22
16.79
Virgil Crest
1.00
1.29
14.77
1.22
16.79
Oil Creek
0.99
1.28
14.62
1.20
16.62
Pinhoti
0.98
1.26
14.47
1.19
16.45
PCTR Headlands
0.98
1.26
14.47
1.19
16.45
Zumbo
0.97
1.25
14.33
1.18
16.28
San Diego
0.97
1.25
14.33
1.18
16.28
Pueblo Nuevo
0.96
1.24
14.18
1.17
16.12
Rio del Lago
0.94
1.21
13.88
1.14
15.78
Mohican
0.92
1.18
13.59
1.12
15.45
Haliburton Forest
0.92
1.18
13.59
1.12
15.45
Ancient Oaks
0.91
1.17
13.44
1.11
15.28
Javelina Jundred
0.91
1.17
13.44
1.11
15.28
Burning River
0.91
1.17
13.44
1.11
15.28
Old Dominion
0.91
1.17
13.44
1.11
15.28
Pony Express
0.90
1.16
13.29
1.10
15.11
Arkansas Traveller
0.89
1.15
13.15
1.08
14.94
Kettle Moraine
0.89
1.15
13.15
1.08
14.94
Sulphur Springs
0.89
1.15
13.15
1.08
14.94
Vermont
0.87
1.12
12.85
1.06
14.61
Squamish Stormy
0.87
1.12
12.85
1.06
14.61
Hundred in the Hood
0.87
1.12
12.85
1.06
14.61
Beast of Burden Winter
0.87
1.12
12.85
1.06
14.61
Run Woodstock
0.87
1.12
12.85
1.06
14.61
Mother Road
0.84
1.08
12.41
1.02
14.10
Rocky Raccoon
0.83
1.07
12.26
1.01
13.93
Lean Horse
0.83
1.07
12.26
1.01
13.93
Keys
0.82
1.06
12.11
1.00
13.77
Heartland
0.82
1.06
12.11
1.00
13.77
Umstead
0.82
1.06
12.11
1.00
13.77
Iron Horse
0.82
1.06
12.11
1.00
13.77
Boulder
0.82
1.06
12.11
1.00
13.77
Old Dominion Memorial Day
0.82
1.06
12.11
1.00
13.77
Ontario Northland
0.75
0.97
11.08
0.91
12.59
Dan Rossi Memorial
0.75
0.97
11.08
0.91
12.59
Olander Park
0.75
0.97
11.08
0.91
12.59


Pacing the Race

            Once you have identified a target race, and have set a realistic goal time (or distance, for a timed race), what strategy gives you the best chance to reach your goal? Clearly you need to put in the training, and there is a lot you may need to figure out about how to deal with various issues that appear during the race, especially if it’s further than you’ve run before. To maximize your chances, you also need to have an appropriate pacing plan. In my experience, one of the most common mistakes – in races of all distances, but particularly in longer ultras – is to go out too fast. In a hundred-miler or 24 hour race, the pace you should start at may well be significantly slower than the easiest pace you’ve run in training. It’s all too easy to go out too hard, costing yourself precious time when you later have to slow down more than you should. Not only that, but a too-fast start could well prevent you from finishing the race altogether. Given how easy the early pace feels, there is a tendency in many runners to “bank miles” early in the race – planning ahead for the (almost) inevitable slow-down that occurs later, when your legs are fried and your energy is waning. It’s tempting to do this, and in my observation most runners (90% or more?) follow (whether by design or not) the miles-banking strategy. It’s not a good idea, in my opinion. There is nothing inherently wrong with positive splits in an ultra, but the key to a good race is to ensure that your splits are not too positive.

            It’s a truism in endurance athletics that an even effort maximizes performance. Before looking at this in slightly more detail, I would suggest that it’s not a bad strategy to aim for even splits in an ultra, whatever the distance. If your goal time is a bit ambitious, there’s a decent chance you’ll have positive splits anyway – just less positive than they would have been otherwise. In this case, striving for an even pace is likely to produce a better result. And if you nail your goal time, or even come in under it with negative splits, it’s not the end of the world. Maybe you could have gone a little faster with a more aggressive start, but the lesson learned would at least be an enjoyable one (it’s much more fun to run negative splits than positive!).

            Let’s say you’re dialed in on a clear time/distance goal and want to optimize your chances of achieving it. What’s the best pace strategy? In a 5k, you may aim for slightly negative splits. For example, Galen Rupp’s recent US 5k record had mile splits of 4:14, 4:12 and 4:04. In the marathon, even pacing to slightly positive splits (~1-2% between the first half and second half) seems to be the best strategy. Dennis Kimetto’s world record race at Berlin in 2014 had almost perfectly even splits between the first and second half.

What about ultras? Pam Smith set the 100-mile women’s track record in 2013, running even splits (50 miles in 7:06, 100 miles in 14:11). In the same race, Zach Bitter set the US 100 mile record (and world 12-hr record) with a 4% positive split between the first and second halves. So, even to slightly positive splits appear to work quite well for times up to around half a day. For longer times, the splits tend to become a bit more positive. Sabrina Little won silver with 152 miles at the IAU 24 hr world championships in 2013; her pace in the last 52 miles was 10% slower than in the first 100. I was able to find hour-by-hour splits for Yiannis Kouros’ world record (303.5 km/189.5 mi) track 24-hr race in 1997, which are shown in Fig. 4. Kouros started at a pace that was 90% of his average pace and held it steady for about 5 hrs before it gradually dropped off. His average pace in the second half of the race was 14% slower than in the first. For longer races, the splits tend to become even more positive. In Traci Falbo’s 2014 US record 48 hr track performance (and world record for indoor), her pace in the second half was 26% slower than in the first.

So, my suggestion for pacing ultras that take on the order of a day or less would be to shoot for something between even splits and a positive split of no more than 15% between the first and second half. This of course applies to races in which the first and second halves are roughly equal in difficulty, which isn’t always the case. If it's not, your level of effort may be a better guide than your pace. It can be valuable to calibrate your level of effort by doing a timed event on a loop course, or a race on flat terrain. Some races report aid station split times for all runners, which can be very valuable in judging the relative difficulty of various sections. Finally, many races, even highly technical ones, have at least some flat, smooth sections on which you can check your pace.

One final note – while your effort level should be relatively even over the course of the race, hour to hour, that doesn’t mean it needs to be absolutely even on a shorter time scale. It often makes sense to pursue a run/walk strategy.

1 comment:

  1. In a nutshell:

    (1) Your initial pace shouldn't be faster than 90% of your projected average pace. For example, if your projected average pace is 10:00/mi, you shouldn't start faster than 9:00/mi (assuming constant course/conditions, of course).

    (2) Projecting your pace from shorter distance events (assuming similar course/conditions).

    50k --> 50 mi: Multiply 50k pace x1.11
    50k --> 100k; 50 mi --> 100 mi: x1.165
    50k --> 100 mi: x1.293

    ReplyDelete