Revamped my draft rater yesterday, and while it still has some flaws (most notably that the sample size of 300-odd players is on the small side, and doesn't include data past the 2013-14 season), it's substantially better than the previous version in a few ways.
-It's free of all interaction terms (things like Rebounds*Assists, etc.), which makes it much more stable for outliers like Trae Young.
-It's asymptotically linear in all variables, which also makes it more stable for outliers.
-It's divided into offensive and defensive components, which makes it more stable in some ways (e.g. defensive rebounds are only allowed to affect a player's defensive rating, three pointers are only allowed to affect a player's offensive rating) and also makes it easier to interpret. One important note on this topic: while total ratings are directly comparable across positions, offensive and defensive ratings are not. For instance, a defensive rating of 0 would be pretty bad for a center, but very good for a point guard. Conversely, an offensive rating of 0 would be pretty bad for a point guard, but very good for a center.
I won't pack this post with any more details (feel free to reply or pm with questions though, and I'll explain in any level of detail). I'll post all the ratings (for lottery prospects) tomorrow afternoon sometime once sports reference has updated to include the results of today's games. So far I've added Jackson and both Bridges (who don't play today). You can also see full historical ratings (most players from 2012 draft and earlier should be in there) under the Historical Ratings tab in the bottom left. I'll also get around to posting estimated uncertainties in the ratings soon.
I'm also happy to calculate ratings for other interesting prospects in this draft and in other drafts post-2012; I'll make another tab for these if there's enough interest.
https://docs.google.com/spreadsheets/d/1h98LyhYUQHnGQTSMJTj3i_0Lj4iF03oHGeyDDtjWgV8/edit?usp=sharing