Deriving the objective in Hensman et al. 2015
MCMC for big datasets -- faster sampling with JAX and the GPU
A first model to rate Formula 1 drivers
Surface-adjusted French Open seedings
Rally lengths on the log scale
Random effects in Gaussian Processes (featuring rally lengths)
Rating the ATP's best forehands and backhands since 1990
Tracking the depth and skill of ATP tennis with an adaptive Elo rating
Good posterior (co-)variance estimates from variational inference
Deterministic ADVI in JAX
US Open Elo preview
The first serve percentage in tennis is noisy
How good is the UTS format at picking the better player?
ATP Serve & Return skills by surface since 1991
Historical ATP Serve & Return skills
When browsing through tennis stats (as you do…), I was surprised to find that Jeff Sackmann’s Tennis Abstract site has player serve stats – points won on first serve, second serve and so on – all the way back to the early ’90s. If you like, you can have a look at Sampras for example; the stats are shown in the far right columns of the table. I was surprised because my usual source of data, OnCourt, only has these as far back as about 2003 or so.
Grand Slam five-setters to watch
Much like Steph Kovalchik did a few days ago, now that there is no live tennis I found myself wondering which historical matches to watch. We both developed approaches to pick the most interesting ones, and I think the results are different enough that I thought I’d put my selection up too.
Has tennis become more predictable? An initial look with a topic model
In these tennis-free days, I thought it could be fun to look back at the history of tennis and chart some general trends in playing style.