Finding "The Why"
The first time I tried CrossFit back in 2016, I hated it. I was in graduate school and took Russian kettlebell classes at the campus gym and really enjoyed the sweat and strength I was getting from it. That work seemed so much more rewarding than doing cardio on an elliptical for 30 minutes or cycling through the various isolated muscle machines. But sadly, the gym changed up their staffing and scheduling and I was no longer able to get the same kind of workout, so I Googled "kettlebell" and CrossFit came up. There was a gym that was a few blocks down the road from me, so I signed up for their 6 week on-boarding program. I quit after maybe two weeks. I won't go into the details (at least not in this post), but essentially I didn't see the point of such masochism.
During the pandemic, I started running outside because (1) it was the only way I thought I could get exercise when all the gyms were closed and (2) I was terrified that being as overweight as I was at the time would put me at high risk for severe illness from COVID-19 if I caught it. I ran a few 5Ks, and the last one I did I remember trying to sprint at the end to cross the finish line—my aerobic capacity was there, but my muscles just couldn't push me any faster. I figured I needed to do more strength training and once again, my Google search led me to CrossFit. This time, I thought I'd give it another try because I was in much better shape than I was last time. But it turns out that my initial fitness had nothing to do with why it stuck the second time.
I walked into Rally Point Endeavors (Northbrook, IL) on a Saturday morning while an endurance class was going on. It was chaos and I felt extremely out of place. But someone pointed me to Coach Brad, who kindly took me to the whiteboard and proceeded to give me what I now would classify as a 45-minute crash course on the CrossFit methodology they teach in the L1 certificate program. I think that would turn most people away, but it was exactly what I needed: a "why."
I'm a scientist - naturally more of a mathlete than an athlete. CrossFit was the first time I had ever seen a scientific approach to fitness. I was immediately hooked. What makes CrossFit scientific? Greg Glassman, the founder of CrossFit, wanted to create a definition of fitness that was "measurable, observable, and repeatable," which is exactly what the scientific method is all about. I was so excited when I saw the same physics equations used to describe work and power in CrossFit as I had seen in my physics classes.
The Project
As of 2023, there are estimated to be 5 million people worldwide who do CrossFit. The sport generates an extraordinary amount of data, from athletes tracking their day-to-day training to the exciting spectacle of the CrossFit Games. The season kicks off each year with the CrossFit Open, which I think might be the largest compilation of standardized functional fitness tests in the world, with more than 250,000 athletes signing up in 2026. There is a gold mine of data waiting to be mined, and I'm building this project to do just that.
My vision for Cross-Validated Fitness is to take my curiosity and apply it to the world of fitness. CrossFit is already measurable, observable, and repeatable, but we can do even better. My goal is to provide research on numerous topics related to fitness using data analyzed with statistical and computational methods that can be held to rigorous standards of scientific inquiry.
The Name
Fitness is defined as "work capacity across broad time and modal domains." Other methodologies focus too narrowly on a single domain, like strength or endurance, or a single modality, like running or lifting.
In his first CrossFit Journal article, What is Fitness?, Greg Glassman argues that athletes we might colloquially refer to as "the fittest" -- like marathon and IronMan finishers -- would not necessarily stand up to this definition of fitness. This is why CrossFit training is "constantly varied" -- it is designed to train the body to function in any situation so that when it is tested, it can perform at a high level in general.
In the realm of statistics and machine learning, we also talk about "fitness" in the context of model performance. We want to build models that are not overfit to the data they were trained on, but rather can generalize and adapt to new data. Cross-validation is a method that helps us ensure that our models are not overfit to the data they were trained on. More than just the similarity in the name, cross-validation is a core concept in the CrossFit methodology.
What to expect
I have high standards for what I'll publish here. I'm not interested in "hot takes" or "speculation." I'm interested in building models that can be used to inform decision-making and improve the community and sport of fitness. Cross-Validated Fitness will be:
1) Quantifiable: All science needs data, and data needs collecting. I've only just gotten my feet wet in this endeavor so far, and I foresee this step being a challenge on its own. We don't have nice, clean datasets to work with -- we're going to have to do some scraping from a variety of sources. It's going to be messy, but I'm going to document the process as best I can so others don't have to.
2) Modelable: This is the fun part. From data we can build models, and from models we can make predictions and recommendations. When we train models on the right data, we can eliminate biases that humans can get trapped in. CrossFit has been around for over two decades now, and while it was novel and experimental in the beginning, I worry innovation has slowed and it's struggling to continue growing. I've seen headlines suggesting CrossFit, LLC has struggled to find a buyer at the valuation it wants -- one signal that outsiders aren't seeing its potential value. I believe if there is unseen value in CrossFit, it's hidden somewhere in the data. Let's go find it.
3) Reproducible: I come from an academic background where the reproducibility of research is a virtue. I intend to produce high-quality code that can be run by other savvy enthusiasts to replicate and verify my findings.
Topics
I have a few ideas for topics I'd like to explore:
- Participation & Ecosystem Trends: How has the population of CrossFit athletes changed over time? Can it continue to grow? Can we find points of leakage and prevent churn?
- Competition Modeling & Rankings: Can I build a computer model that ranks athletes in the sport of fitness better than the best analysts? How do we even quantify what "better" means? Can it accurately predict the outcome of major CrossFit competitions?
- Modernizing the methodology with AI: Can we help train new and existing coaches better with tools built using AI?
I also see part of this website as a personal blog for my own fitness journey. I have a unique story to tell in how CrossFit has changed my life, and I know there are plenty of people out there who can either relate to my experience or are waiting to see that someone who has faced hardship when it comes to entering this space can thrive despite the odds. I wholeheartedly believe in the methodology of CrossFit and its ability to improve the health, fitness, and livelihood of anyone who follows it. But I understand that getting fit is more of a challenge for some people than others.
Therefore, I want to dedicate part of this project to finding ways we can make fitness more inclusive and accessible to demographics where exercise and proper nutrition are not a lifestyle we can take for granted or ignore -- they are a luxury some simply cannot afford.