The box score is a new feature of the draft this season, so nobody really knows.
BB introduced the release of FA last season, and I could look at the stats for a player and tell whether he was more of a rebounder, or a shooter, or a passer. But that was based on season-long stats.
My guess is that the box score was generated as equivalent to a simulated game, and thus might not be overly reliable, especially since box scores are heavy on offensive stats, and can't show if a player held a 26 PPG player to 5 points on 2-19 shooting.
So I have used the star rating as the primary rating. I then pretty much went with potential, but I also might use a box score from a good game as equivalent to good potential or a good letter grade. I also considered position to some extent. I have had good PG as my 1st pick the last two seasons. I don't need a 3rd PG, so I've tended to go for inside players.
You should probably put all the 2s and 1s at the bottom, below the unknowns. The average player is probably a little better than a 3. So I mixed my unnowns with my 3s, moving 3's with good potential, a good letter grade, or a good game up in the ratings.