Scary Math: Which Cross Series Is Most Predictive?

As we marvel at Lars van der Haar's ability to stay in the white jersey of UCI Cyclocross series leader, in the face of a track that represented his biggest obstacle on the World Cup circuit, and as I think back to a post I wrote early on predicting that he might survive this competition and even the Worlds race... I wonder, do any of these series do a good job of predicting a world champion?

For openers, you wouldn't necessarily expect a series to predict the winner of a one-day event, particularly in the fickle world of Cross, where one misplaced tire can send you careening into a barrier called "Palookaville." Then there are some of the practicalities -- like, that the series tend to load up on results early on or in the mid-season Holiday Madness, which doesn't tend to speak to February fitness.

Still, winning a series is indicative of quality, since the series winners tend to win individual rounds, not just finish second a lot. So then, if it should correlate at least a little, which one correlates the most? My impression of the series is that the SuperPrestige and BPost (formerlyGvA) series are harder than the UCI events. SuperPrestige is 100% Flanders, or at least Flemish Belgium, depending on how you want to use these terms, while BPost crosses the odd border, but I'm not sure you could call one harder than the other. The UCI should be the most prestigious, but in this Belgian-dominated sport, the reality is that it has a bit of trouble competing for the Belgians' attention, and is generally more open to non-Belgians as a technical matter (i.e. the race styles). Coming into this exercise, I expected the UCI World Cup to correlate best to the World Championships.

And I was half-right. The "data":

[n.b.: "Year" is when the series began; the world champion is named in the following calendar year, i.e. Nys is the 2013 champion at the end of 2012. Like the Super Bowl. Also, 2013* lists the current leader of the incomplete series.]

Dieter Runkel is an awesome name. Just had to be said.

Anyway, the tally is that the world champion was also the winner of...

• four SuperPrestige series titles
• three BPost/GvA series titles
• four UCI World Cup titles

That's over twenty years, or 19.5 anyway. The correlation is low enough to be regarded as coincidence, or at least that's what I'd guess a real statistician would say. FYI, the world champion in that time frame was coming off a win in the Belgian nats on three occasions, and the Dutch nats on two. Roughly the same rate of correlation in the case of the BKs -- surprisingly not that correlated. [The correlation with the Dutch champ is closer, but the fields are thinner up there.] You might think that the Nats races, coming two weeks before the worlds most times, would line up better with the worlds, but not really. Perhaps it's best to conclude that the Belgians are constantly locked in such intense competition that winning one series or race tends to predict that someone else will win the next big thing. In other words, an anti-correlation.

Or, in the case of Kevin Pauwels, a sign that your luck is about to run out.

Anyway, I suppose you could look at this and say that the UCI World Cup series is an OK predictor for the reasons mentioned above (course similarity), but then you could say that the SuperPrestige is an equally good predictor, for the other reasons mentioned above (excellence). And anyway, you're only one occurrence at the BPost for them all having the same predictive rate.

Here's a predictive element: home country. When the world championship race is in Belgium, a Belgian will win. Whatever you think about the country's ability to repel foreign invasions of the military type, they've consistently repelled the ones on knobby tires since 1986: four for four. Similarly, a race on Dutch soil is almost statistically certain to... preclude a Dutch winner. One for six overall. Not good.

I still say Little Lars is in excellent shape for this year, but it depends on how deep the mud gets. Can't wait.