On the Dangers of Single-Number Metrics

I don’t want to say single-number metrics are completely useless, but the people who create them think they’re completely useless.

Of course, they don’t think the one they created is completely useless. They just think all of the other ones are completely useless. So collectively, they all think that all but one of the single-number metrics are completely useless.

Hyperbolic? Not entirely. The search for the Perfect Number that entirely encapsulates a player’s value is an impossible game of statistical Whack-a-Mole for a few fundamental reasons.

By addressing these issues we can hope to stop using them irresponsibly and look at the few ways where there actually is value.

They are Metrics, not Statistics

People will argue that the “metrics” are stats, but they aren’t. They’re opinions disguised as numbers.

Single-number metrics work take various stats such as points, assists, blocks, steals, field-goal makes and attempts, perhaps some on/off numbers and so on, assigning different values to those things.

Those values are based on opinions.

How much is an “assist” worth? If a player drives the lane, collapses the defense and kicks the ball out to a 3-point shooter on the corner, is that the same as when a player runs through four screens, loses his man and gets open while the passer just stands there?

How much does a missed field-goal cost? Some will discount a missed field-goal a full point with the logic that it is a lost possession, but what if that missed field goal is snatched by a teammate? One out of every three or four misses doesn’t actually cost the team a possession.

What if the shot clock was down to five seconds and it was the best shot the player was going to get?

These are not absolute factors. Nothing is.

See the problem? There are variations in opinions on these things, and even those things aren’t static. There is no one value for all assists or for all missed field-goals or rebounds or any of those things.

Metrics aren’t statistics; they’re giant box-score, cluster-balls of opinions. Don’t get fooled by the number.

And never, ever, settle an argument over who a better player is by showing how one guy has a higher opinion-number than the other guy and saying #FACTS!

They Create Blinders

















Because they are inclusive and mush everything together, single-number metrics tend to take the differences between players out of the discussion, and that can actually hurt your ability to appreciate those differences.

Stephen Curry can spot up and shoot from anywhere on the opponent’s side of the court at a moment’s notice. Coupled with his ability to put the ball on the floor and get to the bucket with impressive finishing ability, he makes life easier for everyone around him.

James Harden exploits every rule to the thinnest margin, drawing fan-enraging fouls by conning defenders to reach in—or hitting legal crossover pull-ups that make your eyes believe he traveled.

And if you think you have him stopped, he’ll just turn his dribble into a pass in one almost undetectable motion to a teammate parked at the corner for a 3, (or a skying Clint Capela for an alley-oop).

LeBron James is a point guard in a power forward’s body, the most dominant physical force the NBA has ever seen. He’s cultivated that physique through work and practice into being one of both the current and historic league’s most skilled players as well.

As different as their styles of play are, they are statistically similar: the only three players with 25 points, six assists, five rebounds and 60 percent true shooting over the last two seasons, per Basketball-Reference.com.

Players averaging 25 points, 6 assists and five rebounds since 2017-18
Rk Player From To Tm PTS TRB AST TS%
1 James Harden 2018 2019 HOU 30.1 5.4 8.5 .616
2 LeBron James 2018 2019 TOT 27.7 8.5 8.8 .620
3 Stephen Curry 2018 2019 GSW 27.0 5.1 6.1 .676
Provided by Basketball-Reference.com: View Original Table
Generated 11/19/2018.

The point here isn’t just about the speciousness of lumping them together because they’re statistically different; it’s about how fans can fail to see the specialness of each one. When we get too worked up by MVP debates and player value, we fail to appreciate the uniqueness and greatness of each player. If you can’t enjoy Harden, Curry and James, you’re just not watching basketball right. Put down the single-number metric long enough to appreciate the generational talents that move the game forward.

They Overlook Team Aspects

A player-ranking device extracts the player’s performance from the team’s, as though everything about it is going to be an exact duplicate on a different roster.

Even when you’re talking about something as simple as assists—which are a bit easier to come by when you’re passing the ball to Klay Thompson—there are differences from team to team.

Russell Westbrook gets a lot of uncontested rebounds because it makes sense for him to snatch the ball and initiate the break. But that doesn’t mean his rebounds are the same as Andre Drummond’s, who leads the league in contested rebounds, according to NBA.com.

A scorer like Zach LaVine might be able to average around 25 points if he’s on the Chicago Bulls where there aren’t other scorers, but pair him with a second star, and his scoring will go down as his attempts take a hit.

However, more of those attempts are going to be the bad ones than the good ones, so his shooting percentages are likely to go up.

Defense, more than anything, is impacted by teammates. Adjusted real plus-minus stats attempt to alleviate some of that, but it’s not perfect and is more an estimate of the player’s value on that team than a stand-alone absolute.

Who you’re playing with matters, and so does who you’re playing against. Bench players racking up big Player Efficiency Ratings might not be so impressive if they had to put the same numbers against starters.

All of these complications add up but get lost in the single-number metric.

They Can Be Used Responsibly

Single-number metrics do have a place; it’s just not in ranking players or absolutes.

They are useful in terms of broadly gauging a players’ value. If you’re not using it as a lock-tight ranking structure, they do have the benefit of including all the variables so that players of different types can be roughly compared. As long as we aren’t making these into end-all, be-all arguments, it’s helpful to know which general “bucket” that players go in.

Second, they can help us find the underrated players and glue guys.

For example, Robert Covington’s value gets lost in traditional stats, but he’s near the top of ESPN’s Real Plus-Minus year after year because of his defense. This year, he’s 24th. Last year, he was eighth. In 2016-17, he was 24th. Does that mean he’s a top-25 player? No, and if you’re arguing it, you’re giving the stat too much credit.

Does it mean, however, that Covington is better than his box score numbers suggest? It most definitely does.

When you see something “weird” like that, it’s a good time to evaluate, “Is that the stat or the player?” Eye test and stats aren’t the “either/or” too many try to make them. They’re “both/and.”

Don’t accept the number as gospel, but don’t just throw it out either. Watch the player play. Study him.

On a personal note, I’ve learned a lot just by trying to reconcile numbers with player reputations. Few things do a better job than the numbers when forcing you to set aside confirmation bias, open your eyes and really watch a player.

Using single-number metrics wrongly dumbs down the conversation, but using them responsibly can elevate it.