Open walshie4 opened 6 months ago
This one looks pretty solid. Lots of stars - https://github.com/swar/nba_api
https://gom-uat.ngss.nba.com/ui/developer
It is quite old... (2017)
~https://geniussports.com/sportstech/data/~
~This company also looks to be pretty heavily used, but has no pricing. Probably would need a demo and pricing details (so maybe further down the road)~
Eh actually this seems more related to producing and managing data not consuming
found another here: https://api-sports.io/documentation/basketball/v1
have a free account I'm playing with to try it out
There are also some older ones that are like these we can investigate a bit more to see if we can piggy-back the nba CDN
Others I've found too:
Most appear to not have box scores at the player level which is a blocker for most off what we want to do
This one has boxscore info but only for each team's leaders in points, rebounds, assists per game
YEAH, HERE WE GO!
Example response of the game summary call:
Also note: we should have people be able to choose favorite teams / players so the first things suggested to them are usually players and teams they're familiar with
We would need to be a bit strategic to stay under the 1k/ month call limit but 1 call an hour for a month uses 744 so we could probably have a few extra calls to get the list of games at the start of day and then have each update at key times throughout the game, and then probably once more before the end of the day to check for any corrections.
Of course we can also pay and increase the limits
Good resource for looking into developing projection system (if we want to go down this path - seems like it would probably be worth it) - https://fivethirtyeight.com/methodology/how-our-nba-predictions-work/
Also looks pretty interesting - https://www.basketball-reference.com/about/projections.html
Some stuff I found digging around some into where we could source our data from:
Would recommend we start w/ some form of scraping before starting to pay more expensive channels (may be beneficial to use multiple sources to avoid errors / scraping detection)
I am imaging this starting as a scheduled batch process, pulling data into some db we own. In the future these could also be supplemented w/ streaming sources if they are available.