I was interested (rather than offended) to read about Noah Hoffman’s pacing strategies in Sunday’s classic WC race. I have a limited supply of data on split times (what I do have is thanks to Jan over at worldofxc.com, though) so the following data is definitely incomplete.
Hoffman seemed determined to not start too fast, something that apparently he does quite often. Since pacing interval start races is so much different than mass start or pursuits, we’ll only look at his splits for interval start races. I only have split times from a total of eight WC-level interval start races for him (of varying lengths). Here’s a simple graph showing them all together, with Sunday’s race highlighted with the black dashed line:
This is a very crude representation of split times, where I’ve simply plotted how fast Hoffman skied each timed section compared to the field. So, for example, the y-axis means he had the 20th fastest, 40th fastest, etc. split time on that section.
In order to compare races of different length I’ve converted the x-axis from raw kilometers to a percentage of the total race distance. (This may be dubious, since pacing strategies will be markedly different in a 15k versus a 30k. However, these data consist of two 10k’s, five 15k’s and only one 30k, so I think we’re on fairly safe ground.)
Certainly Noah’s first split was slower than his subsequent splits on Sunday. And it was the 3rd slowest initial split of the eight I have. But it seems to me like he proceeded to ski the rest of the race fairly consistently, rather than gradually accelerating. At least, until he faded a bit on the last section.
This is in contrast to several of the other lines here that begin with fairly quick initial splits, but by mid-race he’s clocking only ~60th fastest time on each section or so. So whatever he did seemed to work.
I should say that I’m fairly cautious about my ability to analyze split times. I’m sure the coaches are keeping more detailed data on this sort of thing than I have access to. But it’s interesting, nonetheless.
Continuing with my series of posts looking at some split time data (kindly provided by Jan at WorldOfXC.com), we turn to identifying some folks who tend to ski better (as measured by their % behind the leaders) towards the end of a race than the beginning.
I actually fit a model on this one (gasp!) rather than just exploring the data and making a graph. So after a little bit of fun with some linear mixed effects models, here are the 12 skiers with the strongest tendency to ski closer to the leaders as the race progresses:
Once again, you can click through for a slightly larger version, and the y axis is a highly adjusted measure based on percent back relative to that skier in that race. You can go back and check some of my previous posts using this data for a more detailed explanation of what I’m measuring on the y axis. The basic story is that it’s a relative measure for each skier. So the actual numbers don’t signify anything across athletes, just the overall trends.
You’ll notice that several folks here appear to be dramatically further behind the leaders early in mass start and pursuit races. This is in part a function of how these races play out. Early on, there will often be a very large pack of skiers. If someone, say Alex Harvey, is just sitting comfortably mid pack, at 1-2k into the race he might be as much as 5-10 seconds behind the “leaders” just because the pack is so large and moving so slowly. That can easily translate into a fairly large % back given the small amount of time that has passed.
Still, this model included a variable that accounted for the global differences in race type (interval vs mass vs pursuit) so these skiers are still exhibiting a stronger trend than other skiers, even accounting for the overall differences in race type.
Teichmann’s tendency to pick up the pace during a race is fairly well known (my benchmark here being that even I know about that one!). But maybe some of the other’s are a bit of a surprise to some folks…
Last time I looked at some split timing data for Americans in World Cup races. Let’s do the same with our neighbors to the north:
Visualizing split times is trickier than it seems. You have data from an array of different situations that arguably aren’t terribly comparable. We all know that mass start and pursuit races play out very differently than interval start races. It may also be difficult to compare split times from a Tour de Ski prologue to a 50k. Add to all this the fact that each race will record their splits at different points in the race, even for races of the same length, and you’ve got quite a jumble of information. (Once again, I’m using split time data kindly supplied by Jan at WorldOfXC.com.)
Still, we can do some stuff with this: Read more
Pacing is a frequent topic of conversation in skiing, or really any endurance sport. Typically, the refrain is ‘don’t start to fast‘. In fact, I feel like I hardly ever hear people recommending that one should start a race harder. It must happen occasionally (and I’ll share a story about this later), but I suspect endurance athletes have a general bias towards worrying about going too hard early in a race rather than too easy.
If someone tells me that I shouldn’t start a race ‘too fast’, my first question is ‘Too fast compared to what?’ There are only two things skiers could compare themselves to: (1) the maximum speed that you, personally, could sustain during an entire race, and (2) the speed your competitors are travelling at.
Split times from races give you direct information about (2), which in turn gives us only indirect information about (1). Think for a moment about how runners can learn to tell how fast they’re going based on how they feel. Runners can run on a fixed course (e.g. track, fixed road or trail course) and glance at their watch every lap or every mile and get instant feedback on the connection between how they feel and how fast they’re running. Even with the aid of GPS devices, skiers don’t have a concrete, objective measure of speed to compare themselves to that’s independent of weather, snow conditions, wax, etc. They can’t look at their watch, see that they skied 1km in 2:47 and have that mean something to them. Obviously, skiers do develop a sense for pacing, and probably a pretty good one, too. It’s just harder and they have to do it mostly be feel.
Let’s think carefully about how we’d show with data that someone started ‘too hard’ in an interval start race. The first thing we’d look for is their split times slowing down during the later stages of the race, right? But how do we know for sure that this race tactic would be slower compared to the alternate universe in which the skier started at a slower pace? It’s possible that the time they lost late in the race is balanced out, or even outweighed, by the time they gained early in the race. Ultimately, it’s tough to know for sure without a time machine that let’s us go back and repeat the race using different tactics. Runners (particularly track runners) can experiment with these sorts of tactics with a fairly high level of precision, but skiers have to use whatever innate sense they develop over time.
Obviously, I’m not saying that it’s impossible to tell when you’ve started too hard. I’ve bonked in races myself, you know. My point is that while the extreme cases of poor pacing in either direction are easy to spot, there’s probably a large grey area that’s difficult to navigate with any certainty, at least in skiing.
Let’s get back to the old adage, ‘Don’t start too hard’. My point with this post is that this advice only makes sense compared to yourself (1) not your competitors (2). As I said above, how fast you’re going compared to everyone else is surely correlated with how fast you can go relative to own potential, but if you start too hard in a race against Petter Northug and bonk, it isn’t Northug that caused your body to shut down, it’s your own limitations.
In fact, we can easily flip this around and say that compared to everyone else, you can never start too fast. What do I mean by that? Well, one of the first things that has jumped out at me while digging through the split times from the past WC season is how closely correlated early splits are with final performance, particularly in interval start races: Read more
Thanks to some help from Jan at WorldOfXC.com, I’ve been slowly gathering the split time data for World Cup races from this season. Analyzing them is tricky, though, for a variety of reasons.
First, the data quality is poor. There are numerous instances where the live timing data is obviously wrong in a way that I can’t fix by hand, and so a certain number of split times need to be omitted completely. Second, the live timing data itself doesn’t say exactly how far into the race each timing instance is, so I have to infer that from the times and the length of the race. It works pretty well, but I’m sure it’s only accurate to +/- 500 meters or so. Lastly, handicap start races have to be dealt with separately, since the live timing tracks the time from when the first racer starts, not from when each racer starts.
Still, I’m beginning to find some interesting stuff. A little later I’ll show you some graphs on individual skiers that can shed some light on tactics and performance that I think are kind of interesting. For now, though, the obvious: Read more
One of the interesting things we can look at with sprint races is differences between the various heats themselves, rather than individual skiers. Interestingly, reconstructing this information, namely who was in what quarterfinal or semifinal, using just the split times took a little bit of thought. Not a monumental challenge, but it was a bit trickier than my average graph.
Let’s start with the quarterfinals:
This graph tracks the members of each quarterfinal through the entire sequence of heats. I’m still tinkering a little with the y-xis. Rather than plotting everything relative to the fastest time of the day, I’m plotting them relative to the median time from the whole day. The stats geek in me finds this more appropriate, but it doesn’t really change much. The relative differences between each skier will be the same, but the interpretation is slightly different: 0% is the median skier, not the fastest.
Federico Pellegrino (the big winner in qualification and loser in the final) certainly did get quite lucky with a slow, slow, slow quarterfinal (red). Quarterfinal 5 was probably the next easiest, at least in terms of competitiveness. (Keep in mind that these graphs don’t tell us anything about crashes or other “extraneous events”, that can sometimes account for the speed of a heat.)
On the other hand, the women didn’t see such a stark difference between the quarterfinals:
Once again, these are the raw times, so this doesn’t reflect Kowalczyk’s relegation to 6th place in the final. So that pair of dueling orange lines are Kowalczyk and Randall.
I think the most interesting quarter here is #4 (purple) where only Majdic made it all the way to the final thanks to lucky loser status. So far this season it seems like Majdic has consistently been present in the sprints, but hasn’t really put anything impressive together in a final.
How about the semifinals: Read more