Internet of Kilos
How conversations turn into passion projects.
A while ago, while having a conversation with my friend Ruben, he suggested that I should look into developing a new method for the Sinclair coefficient in Olympic Weightlifting. If you are not too familiar with the Sinclair formula, what it does it that it allows for a comparison between two lifters in different weight classes. This formula was developed by a Roy Sinclair, a Canadian Mathematician who was, like myself, very interested in the sport of Olympic Weightlifting.
The only problem with the Sinclair is that it uses as a reference the world record holder of the heaviest weight class, so it creates a certain bias. My approach has been to use all current eight world records for the model.
The picture above shows data scraped from the International Weightlifting Federation’s website, from all international competitions in 2015 up until October. Those vertical bars that formed from the data, represent the top of end of each weight class. As you can see from the graph, every weight class is trapped in a of more than 11 kilos but the 105+ is scattered, with entries all the way up to almost 190kg. Not only that, but the 105+ is the category with fewer lifters, which means, less data.
The new method I’m testing consists on approximating the logarithmic curve that best fits all eight world records and a comparison method based on how far off they are from the curve. Whoever is closer is better.
Back in November I launched a web app that allows you to compare yourself in real time with other lifters. Before you take a look, please don’t judge my HTML skills based on this, I was keeping it simple because I had not clue of how to integrate Python to it. This was just a test and you really need to keep reading to find the latest version which looks much better. http://nel-sinclair.herokuapp.com
I was learning a lot of Machine Learning tools at the Data Incubator this past Winter so at that time I decided to improve upon my previous work and add a couple of things. I ended up creating a whole new website which contains an interactive database with records in weightlifting going back to 1917, as well as some analysis about the distribution of such records, a Machine Learning app that gives lifters an idea for what they should be aiming for on their next training cycle (work in progress, still), an interactive world map with the amount of records broken by countries and also the comparison method that was introduced on the previous web app (still also work in progress also), all wrapped in a neat front end.
With this new website I plan to create a platform for all sorts of Olympic Weightlifting analysis. Here’s the link in case you’d like to check it out. http://iokilos.herokuapp.com
I’m also going to leave you my Github, in case you’d want to check out my code. There’s a few repos about weightlifting analysis and two others containing the code that generates each website. https://github.com/nelabdiel
If you have any ideas for web apps, Data Science projects or some cool sports related analysis that you’d like to see, let me know and I will look into it!
Cheers,
El Nel
Like this post? Press the heart shape button just below to recommend it and/or share via social media. Feel free to highlight any part of it you like and/or leave me comments on it.
