LeBron James on Sunday turned the thrilling finish between the Portland Trail Blazers and Denver Nuggets in Game 7 of the Western Conference semifinals into a conversation about analytics.
CJ McCollum carried the Blazers to victory, going 17-for-29 on field goals for a game-high 37 points. 26 of his shots were 2-point field goal attempts, leading LeBron to question the current wisdom of analytics.
That’s exactly why I don’t wanna hear all that analytics talk! In PLAYOFF games when it usually come down to 1/2 possession games down the stretch, just get me bucket! The best shot during that possession. And if it’s a Pull-up 2 then so be it cause it was the best SHOT!
— LeBron James (@KingJames) May 12, 2019
The prevailing thought with NBA analytics is that they favor 3-pointers and layups/dunks and outlaw mid-range jumpers. But that’s not true, and anyone who believes that is the case or argument is missing the point.
The idea behind studying the numbers of the game is to create the greatest offense and defense possible. Shot charts, percentages, and math suggest that what produces the highest-scoring and most efficient offenses are taking close 2-pointers, since they are easier to make. They also favor attempting three pointers rather than long twos because the 50 percent point bonus on threes greatly outweighs the difference in shooting percentage. For instance, if a player makes 38 percent of his threes and 44 percent of his 20-foot twos, it makes more sense for him to take threes. If that player takes 100 threes at 38 percent, that would yield 114 points, compared to 88 points for making 44 percent of their 20-foot twos. That’s why over the course of a season, you would want that player taking threes.
But the thinking and arguments do not end there. And that’s what LeBron is missing if he thinks that’s all analytics tell you.
Over the course of 10 shots, that same shooter might not be at their best and make only three threes (for nine points) but they could make five twos for 10 points. In that short run/small sample size, attempting twos might yield the best and most consistent results if we’re only talking one or two games, not 82. Moreover, if you just need to make one shot, you should take the one you make at a higher percentage, which would be the 2-pointer.
When it comes down to it, the objective of an NBA’s offense is to score. If you have five minutes left in a Game 7, you don’t want to hoist threes simply because analytics tell you those tend to be the most rewarding shots over the course of 82 games or 48 minutes. You want to take shots that you’re most likely to make — the “best shot during that possession,” as LeBron says. Analytics aren’t telling you otherwise.
Analytics would also tell you that aside from layups, McCollum’s best shooting area is the mid-range by the free throw line. NBASavant.com says he was 121/270 (44.8 percent) from that area during the 2017-2018 season, which is 5.2 percent better than league average. He’s not too far off from being a 50 percent shooter from that range, compared to being a 37.5 percent shooter on threes this season. If the Blazers just needed a bucket as they did towards the end of the game against the Nuggets, there weren’t many more preferable shots for their offense to take. And that’s what they want — high-percentage shots. So analytics would endorse McCollum’s shooting approach.
The same issue exists in baseball.
There is a thought that analytics say you are not allowed to bunt and that it’s the worst thing you can do. And while giving away outs by sacrifice bunting will limit your ability to have big innings and score runs over 162 games because you lower the runs you’re expected to score, it’s not a blanket rule. If my team is up 1-0 in the top of the ninth and we have runners on first and second with nobody out, a ground ball pitcher on the mound, a slow .230 No. 9 hitter at the plate, and I’ve got a good closer, a sac bunt would make a lot of sense. The run expectancy between having runners on first and second with nobody out (1.44) compared to second and third with one out (1.38) is only lowered by 0.06 and fairly similar. If my nine hitter is up in the first scenario, I’d rather have them bunt to bring up my leadoff hitter, who’s a better hitter and more likely to help the team score.
The real problem is if you take these statistically-rooted recommendations and turn them into edicts, failing to allow for adaptation depending on the situation. LeBron should know and understand that better than anyone.
