Crossresults.com -- Who's going to win Nationals?

Cross nationals are theoretically a great place to test the legitimacy of crossresults.com's points system -- it's the only time of the year when racers from every part of the country go head-to-head against each other. If the points system could correctly predict where a New Englander matches up against a Portlander -- and where they both fall against a ride from Kansas City -- then we'd really have something.

After running the points against almost every field out there, it was interesting to see what percentage of riders were unranked in the crossresults.com database -- almost 50% of non-elite riders. This is a good thing, really it is, because it means cross is even bigger than I realized. There were 80+ races added to crossresults.com in November, and it's quite possible there's 80 more we don't know about yet.

So, the first disclaimer is that we're only looking at about half the data.

Let's check out the women's race. Common sense should tell you that this is a two-woman race between Georgia Gould and Katie Compton. Here's your top 7:

Georgia Gould 248.532002836
Rachel Lloyd 266.668229669
Kelli Emmett 284.587457081
Sue Butler 290.630611044
Katie COMPTON 291.063175498
Laura Van Gilder 300.012391789
Maureen Bruno-Roy 303.89

How the heck is Katie ranked so low?? One word -- quantity. Katie's done (according to us) six races this year, and half of those were in Europe. When she's actually on this side of the pond, she's dominated Georgia, but when she's in Europe she's doing not quite as well, against competition without quite as low points. Meanwhile Gould has been killing the much more established (in our numbers' eyes, at least) women's scene stateside.

Yep, another flaw. We've only been tracking European races for a season -- things are still stabilizing over there. Remember that all elite women, regardless of intrinsic awesomeness, are worth 440 points the first time we see them. That's why winning a World Cup can be worth 287 points but winning a USGP is 246 -- we have a lot more data about the US scene, and the algorithm can tell that USGPs have all the best riders.

The other problem is that the European racers and North Americans don't cross-pollinate enough. If they raced each other regularly, the points would "match up" quickly, and soon Hanka or Daphny would be at the top of the charts by "stealing" the low points from the top US riders. But if they never race, the system just doesn't know! Seeing a rider like Compton go from being cream-of-the-crop here to cream-of-the-crop there (she's got two wins and a second in Europe, you know) doesn't give enough information about the relative speed of the fields.

I don't think anyone will argue that the racing is faster in Europe, but I'm not about to "hack" the points to reflect this.

If you look at the head-to-head matchups between the top 5, you can see how much this is Compton's race to lose - she's 4-1 against everyone else combined, with her lone "loss" being the World Cup she DNF'ed from. Georgia Gould hasn't lost to anyone in that group since USGP #1 -- she is as much a shoo-in for 2nd and Compton is for first.

With Lloyd leading Emmett by another 22 points, it's safe to say that the top 3 is most likely to look like this:

1)Compton
2)Gould
3)Lloyd

The men's race is, unsurprisingly, less cut-and-dried. Here's your top seven:

Todd Wells 129.666681648
Tim Johnson 130.063377424
Jeremy Powers 135.700068861
Andy Jacques-Maynes 150.424373869
Ryan Trebon 150.461717657
James Driscoll 152.700823423
Troy Wells 152.978070

The men's field is way tighter. But the top men have barely half the points of the top women -- so the points gaps, as a percentage, are actually similar. Look at the 15-point drop from Jeremy Powers in 3rd to Andy Jacques-Maynes in fourth -- that's over 10%.

And it matches up with common sense, too -- if Andy, Jamie or Troy Wells won Nationals it would be a huge upset. Those guys are 10% back of the leaders on points. You can effectively remove them from contention.

You probably noticed that I didn't put Trebon in that group, despite his points score of "only" 150. This number should be surprising to you, because Tree Farm is a favorite on just about every nationals prediction out there. Why does the crossresults.com points system hate him?

Unfortunately, there's a blip in Trebon's race history where he apparently decided to go back to Portland to "have fun." A good way to loosen up during an otherwise intense cross season -- but a terrible way to impress the points system. In Portland he had the following results:

SSCXWC SINGLE SPEED MEN 8 (277.3)
Cross Crusade #7 Singlespeed Kona 1 (228.72)
Cross Crusade #7 Category A Kona 1 (180.11)

The points system rightly detected that the SSWC race was full of clowns, and placing 8th against a bunch of clowns is baaaaad. Luckily Trebon gets to drop that one, but now he's used his drop -- so winning such B-quality races as Singlespeed and A's at a cross crusade goes on his points. That's right -- winning two cross crusade races, in the same day, is still bad for your points when you're as good as Trebon. It's basically like a A racer winning a C race -- the points algorithm scores you as "slightly faster" than second place, when in reality you were miles ahead.

Drop those two results and Tree Farm's points would be 130.14. Sound better?

With that issue out of the way, we're left at Wells, Johnson and Trebon in a dead heat with Powers slightly behind. Check them out head-to-head.

You can see why Powers is trailing -- against these three opponents he only has 3 win in 18 tries when at least one of them is present. He's never won when all 3 were there. While he's in the discussion -- he's definitely the outsider. Him winning Nats would be a minor upset.

Drop him from the comparison and you'll see how it's almost a dead heat... between Trebon and Johnson. Shocking, I know.

Trebon has won 8 in 17 tries. Johnson has won 7 in 15 tries. Wells is only 3 for 13 -- so now he's the outsider. The prediction is, then:

1) Trebon, by virtue of going 8-6 head to head against Johnson this year
2) Johnson
3) Wells
4) Powers

Isn't that some boring stuff? Did you really need this much math to determine that?

But wait, you ask, what about Jonathan Page? What's his deal?

Sadly he's suffering from the same Euro-competitive-imbalance Katie is. He's ranked a piddling 25th, with 206 points. There's no way JP will be 25th in this race, but with hardly going up against Trebon and Johnson this year, there's no way to know how close to the front 206 European points will put you.

Don't like it? Get the Euros to come here -- or North Americans to go THERE -- and it would all straighten out in about 8 races. Just like everyone else, I can't wait to see how Page does here.

This "article" should have revealed the biggest flaw in crossresults.com's points -- separate bodies of racers that don't meet up often enough to stay "connected." This affects European and North American elite riders -- but also New England and Californian masters riders. Within a population of riders who race regularly, it's often an amazingly accurate summary of an entire season, but across multiple populations it's only an approximation. Getting beat at KC by a guy with 40 more points than you shouldn't hurt too badly -- unless he's from your home region.

The off-season should see a few tweaks to the points algorithm, but there's no way to fix the "separate populations" issue without adding a lot bias, which would be even worse -- because then instead of defending my vague math here, I'd have to defend my personal judgement.