Breaking Down the Open

09 Mar 2020

The 2020 Crossfit Games open was held this past October. With hundreds of thousands of people participating in the competition from around the world, we took at look at the scores and did some data analysis of the results.

We have said it before, and we will say it again, and probably will keep saying it - here at Wodly, we LOVE data! We believe that data is crucial to improving and tracking our progress over time. Being the nerds that we are, we decided to turn our data loving brains to the most recent Crossfit Games Open results to see if we could find anything interesting.

For the uninitiated, the open is a yearly world wide online competition lasting for 5 weeks with one workout per week. This is the first chance of the season to qualify for the Crossfit Games. Athletes compete in front of a judge and/or on video and submit their scores online over a 4 day period each week. The workouts are a secret each week, however, multiple attempts can be made before the 4 day period is over. There are many divisions across age groups including scaled and RX workout standards.

With the leader board finalized in January, we now have the true final results of the open. These results are publicly available on the Crossfit Games website for anyone to look at. Using the magic of modern technology we pulled an anonymized set of all RX results for both men and women in the senior category (18-35).

Before we continue, some quick notes to the reader which might help understanding what is going on in the below graphs. A very common way we measure performance in workouts in by Power. Power, in physics, is basically work done divided by the time taken to do the work. For the purposes of this analysis we have broken down Power as number of reps divided by time taken. For the graphs below, we multiplied this by 100 just to scale the graphs to make it easier to understand. On the below graphs, the number on the bottom is power and the number on the left is the number of athletes who got that power score for the workout. The lower the power, the further down the leader board you were, the higher the power, the better you did. For any of the workouts below, divide the number of reps you got by the time you got in seconds, multiply the number by 100 and that is where you would fall on the graph.

Finally, the graphs were largely the same for both mens and womens divisions in terms of shape and what they can tell us. However, the graphs below are based on the data from the mens division.

20.1

20.1 was a classic relatively low skill couplet of snatches and bar facing burpees. The workout was 10 rounds for time of 8 ground to overhead (snatch or clean and jerk 43/29kg) and 10 bar facing burpees. This workout had a 15 minute time cap.

With some knowledge of this type of workout, it is fairly easy to see that this workout doesn't really have a movement that people wouldn't be able to do. It is relatively low skill. Thinking about this from a data point of view, there is a concept called a normal distribution. Most people are familiar with this, it's essentially a bell shape. Amazingly, with enough data points, a huge amount of metrics measured in human populations, when plotted, tend to naturally form this curve.

Below is the weighted distribution of results from 20.1.

20.1 Score Distribution

As we can see, this is almost a normal distribution; this should make sense. What it's telling us is, there are two extremes, the most fit people and the least fit people, these are at either end of the scales and there are not very many of these people. Then the majority of the population are somewhere around the middle or median of the scores.

This makes a lot of sense when you think about the programming, most people should have been able to do this entire workout without hitting a wall where they just couldn't do something. So the only real test here is work capacity and how fast you can move. In a large section of of the population, it makes sense that we should see the very fit on one end, the not so fit on the other, and everyone else clustered around the middle.

From a programming stand point, this was a highly accessible workout to the entire community and fits with a pattern in recent years of the open starting with a work capacity test that everyone can do.

20.2

This workout had a larger skill component than 20.1, but it was still relatively accessible to most athletes with a year or more experience. The workout was - as many rounds as possible in 20 minutes of 4 dumbbell thrusters (22.5/15kg), 6 toes to bar and 24 double unders. The reps each round are small, so there shouldn't have been much that would break down. The largest factor would be fatigue catching up with the athlete. For the average athlete, they would be able to hold on and push through a decent handful of rounds. However, the elite athletes were pushing and constantly moving for the whole time pushing for more rounds. Nothing should have broken down, but the athletes with the largest work capacity should have managed to hold on.

With the above in mind, we can assume work capacity is the largest factor in an athletes score on this workout. Below is the distribution of the scores for 20.2.

20.2 Score Distribution

This graph looks a little different, but, it's still more or less a normal distribution. So, what are all the peaks about? These peaks seem to indicate clusters of people who got the same number of fully completed rounds. For example, with something like toes to bar, average athletes who can do them, can often run into a point where they completely break down and the athlete can do no more. These peaks would seem to indicate this number of reps is relatively common for groups of athletes.

Similar to 20.1, the data would seem to indicate that this workout, while more technically challenging, was more or less accessible to a large amount of athletes. However, from the right side of the graph, you can see the very top performers in the sport are able to push much harder in this kind of workout, hence the distribution is shifted a bit to the left.

20.3

This is where things get a little more interesting in terms of the workout complexity and the distribution of the scores. This workout was a short workout with a 9 minute time cap. The workout was 21-15-9 of deadlifts with ascending weight and handstand pushups coupled with 21-15-9 of deadlifts at a heavier weight with a 50ft handstand walk between rounds.

We can think of this workout as being in two parts. There is a lot of high skill involved in this with both handstand pushups and handstand walking. While still high skill, many athletes can do handstand pushups. Handstand walking would be considered a more advanced movement. Also the weight of the deadlift could be a factor here being 143kg on the final bar, with many reps already completed, this could be a factor in performance in the second part.

20.3 Score Distribution

Looking at the above graph we can see something very interesting here. On the left side we see a few spikes, these line up with people being able to do a few deadlifts and some handstand pushups. The drops after each spike show where people went to failure on their handstand pushups and couldn't get any more. We then see a large drop off where athletes could not do any handstand walking. The final spike as we move right is the people who were just about able to finish in the time cap.

The final fall off at the end is show us the top athletes squeezing their times faster and faster down to the best finisher.

20.4

We are getting close to the end of the open now. As expected this 4th workout would introduce another high skill element that was designed to test the best athletes. This workout was a clean & jerk ladder with ascending weight and descending reps. The first 3 rounds were punctuated with 30 box step ups or jumps. In the next three rounds, the box jumps were replaced with pistols. This was the first year we saw pistols in the open.

This workout had two clear separators, the weight of the barbell increased to a degree to only be attainable by the best athletes. The other clear separator was the pistols. Pistols are a very high skill movement which would be considered very difficult by many athletes.

20.4 Score Distribution

Here we see a few interesting peaks. The first one shows us that most athletes were able to get through the first few rounds. The fall off after the first peak is where pistols come in. The following few peaks show us that for the athletes who can do pistols, the issue became the weight of the bar. We see the final peak being where athletes were able to get to the last bar. And we can see, it's not very many. Again, as above, the final fall off is the top of the to athletes doing this faster than the rest of us.

20.5

We are at the end of the open! Bucking the trend this year, there were no thrusters in the final workout of the open. The final workout had a 20 minute time cap and the workout was to complete 40 ring muscle ups, 80 calories on the rower and 120 wall balls in any rep scheme you like.

This workout is more or less the most transparent of any of them in terms of having a separator. Ring muscle ups are considered one of the most technically challenging exercises in the sport. It is no surprise that the number of athletes who can do these is relatively small. With the choose your own adventure nature of this workout, it was going to be the case that many athletes completed the 200 reps of the combined rowing and wall balls under the time cap, and that was the end of their workout.

20.5 Score Distribution

The graph here is dominated by a huge spike. This spike matches to the 200 reps we disgussed earlier. We can see that almost all athletes got this far in the workout. The slight bulge to the right is everyone who can do some muscle ups and managed to get a few reps. The final point when the graphs tails off is everyone who finished the workout in the time cap before it once again tails off to show us our top athletes.

Conclusion

As a disclaimer, this data is far from perfect. The original idea behind it was more of a though experiment to see what we could find out if we looked at the data and tried to make sense of it. We don't claim this is perfect, but we hope you enjoyed our look at the data we collected.

With that said, we do see some interesting trends in the data which seem to match our feeling that the later parts of the open contained workouts that involved higher skill and each one had a key moment which appears to be a separator for the top of the leader board. This is hardly a revelation for those who know the sport and did the open. As ever, it's still interesting to look at the data and see how it matches with our ideas.

Ian Carey
Better. Every. Day.

Articles