Monday, June 23, 2014

Introducing the CFG Analysis 2014 Games Pick 'Em - NOW OPEN

I've toyed with the idea of putting together some sort of CrossFit Games pool or contest for the past couple of years, but a couple of things held me back.  One, I wasn't sure if there was enough interest and enough readership here to support it; and two, I really didn't know how a CrossFit Games pool should work.

Given the pretty decent volume of responses we got for the SWAG's during this years' Open, I think (hope?) there is enough interest to do something more formal for the Games this year. So we're going to try and see what happens.

But how will it work? Well, since the CrossFit Games are unlike most sports where you see pools or pick 'em contests (NCAA basketball, college football bowls, World Cup), the set-up of this contest will be decidedly different as well. It will also give me a chance to put some of the analysis I've been doing here to use. So here's the set-up:
  • Around the beginning of July, I will release my predictions for this year's CrossFit Games. These predictions, like last year, will contain the expected chance of each individual athlete winning, placing top 3 and placing top 10. These predictions are based almost entirely on the results from this year's Open and Regionals.
  • Using these predictions, you enter the contest by placing "wagers" on three male athletes and three female athletes. Each person entering the contest will get 20 points with which to make these wagers. Your result in the contest is based on how much your wagers end up paying off.
  • The payoff for each wager is based on my predictions. For instance, last year I gave Rich Froning a 59% chance of winning. If you wagered 5 points on Froning to win, you would receive 5/0.59 = 8.47 points for that pick. If you wagered 5 points on Jason Khalipa to win, you would receive 0 points since he did not win.
  • You must make exactly 6 wagers:
    • One male, one female to win
    • One male, one female to place top 3
    • One male, one female to place top 10
  • You may spread your 20 points among these 6 wagers in any way you choose. The wagers must be in whole point increments and you must wager at least 1 point on each pick.
As an example, let's pretend I was entering this contest last year. I decide to wager as follows:
  • 5 points Rich Froning to win (59%)
  • 3 points Jason Khalipa to place top 3 (34%)
  • 3 points Lucas Parker to place top 10 (27%)
  • 4 points Sam Briggs to win (32%)
  • 4 points Camille Leblanc-Bazinet to place top 3 (26%)
  • 1 points Alessandra Pichelli place top 10 (24%)
I would have made 8.47 on Froning, 8.82 on Khalipa, 0 on Parker, 12.50 on Briggs, 0 on Leblanc-Bazine and 4.17 on Pichelli. In total, I'd have 33.96 points (this would be 13.96 in profit). My guess is that would have been a pretty good effort.

There is no entry fee. The prize? Glory. Lots of glory. If this works out well this year, maybe I'll figure out a way to actually give something tangible away next year.

So be on the lookout for my predictions coming in the next couple weeks. After the predictions come out, post your wagers to the comments on THIS POST. I'll try to provide updates after each day of the Games regarding the current standings. Have fun everyone!

[UPDATE 7/1: Predictions are below and the contest is now open. These predictions are final, with the following exceptions:

  • If an athlete withdraws prior to the first event, all other athletes' chances will be adjusted upward to relect the loss of that athlete. I'll repost the predictions as soon as I can.
  • If an extra athlete gets a wild card before the Games, I will be forced to re-run all the predictions. I won't change the methodology, so you can expect the chances for most athletes to remain similar.
In the event that I do have to change the predictions for the reasons above, all wagers will be against the revised picks. You can change your own picks anytime up until the start of the first event, for any reason.

These predictions are rounded to the nearest 0.1%, but I will use the true value when calculating the payoffs. Any athletes that truly have odds of 0.0% will be given a payoff of 5,000-to-1.

OK, with that, below are the picks. For methodology details, please see this post.]

[UPDATE 7/8: HQ just announced they will be paying out prize money for the top 20 finishers, but anywhere you see the term "money" in this post, it refers to top 10.  I have re-posted these charts with the heading changed to say "Top 10", but I will not be re-doing the predictions to give odds for finishing top 20.]


Tuesday, June 10, 2014

2014 Regional Review

Every year, watching the Regional competition, seeing how incredibly difficult it is to qualify for the Games even for the most seasoned of athletes, it's hard to imagine that those who do qualify have any weaknesses. Yet in a few short weeks, we will once again see the programming elevate to yet another level, exposing any deficiencies and allowing the best of the best to shine. Even with a few big name athletes out of the mix, it's going to be another tremendous competition, possibly the best to date (the women's side in particular is WIDE open).

But before we move fully into Games mode, let's take a more thorough look back at this season's Regional competition.

First, we'll start with the programming. Although I thought the programming was solid this year, I still preferred last season's events a bit more. A big reason for that is the handstand walk event - I simply cannot get over the fact that a single attempt at walking on one's hands was worth the same amount of points as classic CrossFit metcon like event 3, event 4 or event 7. It just seems to me that if you're evaluating an athlete, those other events tell you a lot more than a handstand walk. Part of that is opinion, but it is also backed up by the numbers. Below is a chart showing the correlation between an atahlete's ranking in each Regional and Open event this season and the athlete's ranking across all other events. Higher numbers indicate that the best overall athletes are typically finishing high on the event, which generally means the event was a good test of fitness.


As I noted last week, the events that really stood out on the positive side were Regional Even 3, Regional Event 7 and Open Event 4. All three had correlations near 80% for both men and women. I've admitted already that my doubts about Regional Event 7 were unfounded, but I'll state it again: this was a pretty great event. On the flip side, you can see that the handstand walk (Regional Event 2) had one of the lowest correlations for both men and women, although the run/rope climb workout (Regional Event 5) had the lowest correlation of any event. My theory here is that since most other events generally favored smaller athletes, many top athletes on the other workouts struggled a bit here since the height was a clear advantage.

Which brings me to a second topic on the programming: did the Regional programming really favor smaller athletes, as many (including me) expected it would? This came up on the Update Show last week, and predictably it was dismissed as being a myth. The evidence? Well, since Tommy Hackenbruck, Jason Khalipa and Elizabeth Akinwale all qualified, the programming must not have favored smaller aithletes, according to Pat Sherwood. Needless to say, as a stats guy (and just a fan of the sport in general), hearing this topic brushed off so casually was irritating for a number of reasons:

  • Anyone who knows anything about data knows that you can't prove a point like this simply by cherry picking a few oddball cases. Can we assume height is not advantageous in basketball because Nate Robinson is in the NBA? Of course not.
  • Tommy Hackenbruck is not even that tall. The man is 6'1", so let's keep things in perspective here. And Elizabeth Akinwale is only 5'7"!
  • As much as HQ talks about data points and all the insights they can gain from the numbers, we so often get responses like this instead of actual analysis. HQ has the numbers at their disposal to provide better evidence to support their point, yet they choose not to. I love this sport, but we lose credibility when the organization in charge so rarely admits any fault.
None of this is to say that Sherwood was necessarily wrong. His opinions were simply not supported by facts. That's what is irritating.

However, the numbers are available to see if the HQ company line is in fact correct. I spent a couple of hours compiling the heights and weights of male Games qualifiers from the past three years (I'd like to get the female numbers at some point when I have some more time). Here are the raw averages for the field in those years:
  • 2012 - 69.9 inches, 198 pounds
  • 2013 - 69.9 inches, 197 pounds
  • 2014 - 69.4 inches, 192 pounds
Clearly there is a drop in size in the past year, and if we throw out the bottom and top two values from each year (including 6-foot-5 Aja Barto in 2012 and 2013), the trend actually becomes a bit more apparent.

  • 2012 - 69.9 inches, 199 pounds
  • 2013 - 69.7 inches, 196 pounds
  • 2014 - 69.4 inches, 192 pounds
Now, is this a drastic drop? No. But the fact that the average size of the Games athletes is a half-inch and 5 pounds lighter than a year ago is not meaningless. Perhaps you could argue that the programming in prior years was too favorable towards larger athletes (I wouldn't argue that, but you could). But to just ignore the numbers altogether is a bit disingenous.

Something else that came out of this investigation (which I hope to expand on at a later date) was that I decided to look at how the distribution of heights for Games qualifiers compared to the general population. My theory was actually that since the programming is generally so balanced, we'd see that the Games athletes have a distribution of heights that are pretty typical. Below are the comparisons (Games athletes distribution based on the qualifiers from the past three years combined, general public data from the Census bureau).


Although the shape of the distribution isn't terribly different, we do see that Games athletes are more concentrated around the average (approximately 5-10) than the general public. In fact, this has been a trend in recent years. The standard deviation in height among Games qualifiers has dropped from 2.4 inches in 2012 to 2.1 inches in 2014, and the standard deviation in weight has dropped from 15.9 pounds in 2012 to 13.2 pounds in 2014. I expect this will continue in the future as the field grows and it becomes harder and harder to overcome any natural disadvantages.

This brings me to my final topic for today: who were the most impressive athletes at this year's Regionals? Obviously, Rich Froning remains the dominant athlete on the men's side, staying well clear of the field despite (reportedly) battling a nasty cold all weekend. And for sure, Scott Panchik again showed that he's a podium contender by hanging right with Froning (and also finishing well ahead of the third-place athlete in the Cross-Regional Comparison). On the women's side, clearly Camille showed up big-time, as she typically has in past Regionals.

But considering what we saw above regarding the concentration of "average size" athletes, it does make performances by athletes like Hackenbruck, Nate Schrader (6'1") and Becca Voigt (5'9") even more impressive. In fact, I'd venture to say these athletes will do even better at the Games, since the programming could be more favorable to their size. Hackenbruck and Schrader were both in the low 20s in the Cross-Regional Comparison (the best of any athletes over 5-11), and I could easily see both in the top 10 at the Games.

That's it for today. In the next few weeks, we'll get into adjusting those Cross-Regional Comparisons, and of course, assessing who actually could win the CrossFit Games. Until then, enjoy your training, and I'll see you back here soon.