Deep Learning on Cubing Statistics
Speedsolving is an interesting hobby because of it's nascency. Rubik's cube world records are broken every year and often times, multiple record are broken at a single competition. Thus, little focus has been placed on more effective training regimes for improvement. Whereas more established sports and hobbies such as baseball or chess have routines professionals use for practicing, speedsolving lacks a common method for practice. Methods that are common in the community have also not been quantified by research.
Luckily, we have enough people participating in speedcubing these days to start crowdsourcing data and ask questions about what aspects of training are most important for people to improve. These approaches are increasingly common in industry settings such as by 23andMe and Verily.
This page is an ongoing development. Please click the link at the top to view the project on Github in the meantime!
Luckily, we have enough people participating in speedcubing these days to start crowdsourcing data and ask questions about what aspects of training are most important for people to improve. These approaches are increasingly common in industry settings such as by 23andMe and Verily.
This page is an ongoing development. Please click the link at the top to view the project on Github in the meantime!
Below is a screenshot of the pygame window. We see that the approximation is already quite good with relatively few points.