|
|
|
|
|
by jparker165
4259 days ago
|
|
You'll surely be more efficient with handling data, but you don't have digital access to all the data you'd need. For example, the chance that a player plays at all (if they are in questionable health) is one of the most important signals. But you can't just write a regression from past data on that player/team/coach. An intuitive guess from reading several media reports will be far for accurate. This guy is surely manually entering a "%chance of playing" driver into these models. But, if you teamed up with a subject matter expert that fed meta-predictions into your model, you'd likely end up with better results. |
|
It occurs to me that collating said data in an accessible format for a modest fee would be a pretty good low hanging fruit business idea. I'm sure there must be something like that out there but last time I looked into this, 2-3 years ago, I couldn't even get the NBA schedule in JSON or some other programming-friendly format, let alone things like injuries and lineup changes.