On assists, and other stats that sort of matter

What is an assist? Section 5, page 9 of the offical NCAA basketball statistician's manual defines the term: "A player is credited with an assist when the player makes, in the judgment of the statistician, the principal pass contributing directly to a field goal."

Fourteen words into the official definition, and we're already running into problems. "In the judgment of the statistician..." This is a stat unlike points or rebounds that each have distinct, measurable occurrences. The assist relies on a scorekeeper making a judgment during the live action of a game. On any night there could be as many as 150 Division I college basketball games across the country, which means there are 150 different scorekeepers making 150 different sets of judgments. If you think those scorekeepers maintain a reliable level of consistency, you're wrong.

Despite the unreliability in the charting of the stat, fans still love assists. And what's not to love? An alley oop. A drive and dime to the big man. A kick out for a three. We can all point to those things and nod our heads in agreement that we've just seen a good basketball play.

But do they matter? And if they do, then what about that play matters? And do assists capture that value?

Denver, of the Summit League, is remarkable at generating assists. They do it on 70% of their made baskets. The only team in the nation that is better is Wyoming, and they assist on 72% of their baskets. That's great, right?

Well, not really. Wyoming has the 93rd rated offense in the nation, and Denver is 110th. The best two scoring offenses in the nation (Duke and Notre Dame) have assist rates of 82nd and 205th.

The evidence that assists result in efficient offense is murky at best. Instead, assists are a measure of "style." If you are a coach who wants assists, then coach a Princeton style offense. If you are a coach who wants to score, then recruit good shooters who don't turn the ball over, and have them rebound and get to the line. Duke, by chance, has the 2nd best eFG% in the nation, and is in the 93rd percentile at not turning the ball over, the 95th percentile at grabbing their misses, and the 81st percentile at getting to the line.

For the longtime readers of Tomahawk Nation, do those last comments about Duke sound familiar? If so, they should. This is as good time as any for a reminder about the Four Factors.

Dean Oliver, who is a lot smarter than me, kicked off a revolution in basketball statistics when he introduced the Four Factors. I won't recap the entire thing here (this is a good place to start your holiday reading) , but essentially the things that matter, in descending order of value, are shooting, turnovers, offensive rebounding, and free throw rate.

So what to do with assists?

Xavier Rathan-Mayes has 25 assists in the past three games. He's fantastic at getting inside a defense, and he has good vision. Unfortunately, he can still make decisions as if he were playing against a high school level of competition, as evidenced by 15 turnovers in that same span. If you are an assist:turnover fan, then you can say aha! and point to his 1.7:1 a:to ratio and say that needs to improve. But all you would really be doing is putting a very important stat (turnovers) as the denominator in a formula. So yes, the 25 assists are nice, but the 15 turnovers are alarming. And the 15 turnovers are far more important standing alone than they are in some conveniently assembled ratio.

The assists are more interesting than important. The interesting part is that in the past three games FSU's assist rate has averaged 65%. These are the three games FSU has prepared for and played since they found out Aaron Thomas was gone. In the prior games the average was 44%.

Is this a long term trend indicating a change in style, or a short term statistical anomaly? If it is a long term trend, any theories about what the offense is doing differently?