Russ Smith has a perception problem.
The Louisville guard has the reputation of a player who takes risks, plays recklessly and takes shots without conscience. All of those statements are true. He’s also awfully good. He led the national championship Cardinals in scoring, and his pressure spearheaded one of the best defensive teams in the country.
The “Russdiculous” label coach Rick Pitino gave him two seasons ago was out of love, but also frustration. Last season, the tag came more out of admiration. Still, Louisville’s most dynamic player on both sides of the court garnered only third-team All-America honors.
He’s also the kind of player you’d expect to be hammered by the statheads.
This story appears in the 2013-14 Athlon Sports College basketball annual. This year’s edition previews every team in the country and includes everything you need to now to prepare for the upcoming season. The annual is available online and on newsstands near you.
College basketball’s most prominent stathead, though, will have no part of the traditional evaluations. Ken Pomeroy and his numbers saw what the national discussion missed.
Pomeroy, whose tempo-free statistics and advanced analytics have become indispensable to coaching staffs across the country, named Smith his National Player of the Year when most other services picked Michigan’s Trey Burke.
The reason Pomeroy diverged from popular opinion had to do with a word not often thrown around with Russ Smith: Efficiency.
The basics of Pomeroy’s metrics are not complicated: Every possession in a basketball game can either end well (a made shot or free throw) or poorly (a missed shot or a turnover). But the number of possessions in a game is not fixed, based on the tempo of the teams involved. His statistics attempt to evaluate, simply put, the rate of possessions that end up with a positive result.
Smith ranked 22nd nationally in Pomeroy’s offensive rating metric, and just as important, Smith had a usage rate of 32 percent. So what does that mean? Let’s start with usage: Smith was responsible for the way 32 percent of Louisville offensive possessions ended, either in a shot from the field, a free throw or turnover. Only 10 players nationally ranked higher. The offensive rating determined Smith accounted for 109 points for every 100 possessions he ended.
For detractors, Smith can say the statistics know what kind of player he is even if they don’t.
“The numbers say you don’t know what you’re talking about if you’re saying I’m inefficient,” Smith says.
And as for Smith’s reputation, Pomeroy writes on his blog, making rational decisions is perhaps an overrated character trait.
More kindly, the traditional metrics of points per game, rebounds per game and field goal percentage are a nice snapshot, but they’re not entirely accurate.
In the last 5-10 years, college coaching staffs have adapted to this way of thinking. Tempo-free statistics have become one piece in the scouting puzzle for assistants across the country. And outside the film room, increased media exposure has made the tempo-free approach and other advanced metrics mainstream among hardcore basketball fans.
Following the lead of Major League Baseball and the NBA, college basketball has immersed itself in advanced statistics and tempo-free analytics.
“(Tempo) can have a profound effect on the stats that are out there,” says Pomeroy, who began publishing his statistics on the internet in 2004. “Scoring stats per game is profoundly effected by how many possessions you have in a game. The tempo-free approach takes out that factor and compares teams on an even playing field.”
Pomeroy owns some debt to Dean Oliver and his Four Factors, which have become one of the foundations of modern statistical analysis on the basketball court. Oliver, ESPN’s director of production analytics and former director of quantitative analysis for the Denver Nuggets, named four distinct statistics which are now essential to determining efficiency:
1. Effective field goal percentage (which puts added weight on 3-point baskets)
2. Turnover percentage (turnovers per possession)
3. Offensive rebounding percentage (percentage of rebounds claimed by the offense)
4. Free-throw rate
Related: The Keys to Becoming a Smarter Fan
Oliver, a former Division III basketball player at Caltech, began charting statistics for his team in 1989 before venturing into graduate school at North Carolina and conducting advance scouting for the Lakers. Oliver wrote for a handful of analytical publications before landing on they payroll of NBA teams as a consultant and then a full-time employee.
“I don’t think it’s underground anymore,” Oliver said. “Some of the stats are part of coaching lingo, where they weren’t 25 years ago.”
The NBA has been several years ahead of the college game in statistical analysis beyond tempo-free, but the most high-profile coaching hire in the pro ranks had an analytical angle to it when the Boston Celtics hired Butler’s Brad Stevens.
Stevens already was considered one of the top minds in college basketball after leading Butler to back-to-back national championship games, but he also was a full-fledged devotee of advanced analytics from an early stage. Last season, Stevens took things further than Pomeroy’s kenpom.com rankings. The Bulldogs coach hired Drew Cannon, a Duke graduate in statistics and Butler MBA student, to conduct statistical research and scouting. When the Celtics hired Stevens, he brought the 23-year-old Cannon with him.
While Cannon was a high-profile statistics expert on a basketball staff, he wasn’t the first to bring advanced analytics to the bench.
In 2010, Mike Lepore was a singular sight in college basketball. Sitting between players and assistants on the Wake Forest bench, Lepore kept his eyes on his laptop. Then the assistant director of basketball operations under Dino Gaudio, Lepore tracked many of the same things Cannon eventually would as well: plus-minus ratings on specific lineups, success rates on offensive sets, key stats on defense. Lepore brought a printer with him on the road to give Gaudio all the data he needed at halftime. However, Lepore eventually gave up the laptop when he learned having it on the bench could result in a technical foul (Cannon kept notes with pen and paper on Butler’s bench).
“I was probably the only person in the country with a computer on the bench,” says Lepore, who is now the director of basketball operations at Saint Louis. “It was really good, detailed information. That’s great information.”
With live, in-game analysis limited, these statistics are most valuable in scouting and evaluation.
On the way to winning the ACC regular-season and tournament titles, Miami, like most schools, made checking kenpom.com a critical part of its preparation. It was one of the first things assistant Chris Caputo read as he evaluated an opponent.
“We try to get a little bit of their DNA statistically,” Caputo says.
“I don’t think it’s underground anymore. Some of the stats are part of coaching lingo, where they weren’t 25 years ago.”
As in any league, the ACC featured an array of styles of play last season. North Carolina was one of the fastest teams in the country at 71.8 possessions per game. Virginia was one of the slowest at 61.5. Nationally, the amount of possessions per game ranged from 58.7 (Western Illinois) to 75 (Central Arkansas).
With that kind of variance, it’s easy to see why points per possession is a more valuable number to a coach than points per game.
Consider this: Team A likes to get up and down the court and score in transition. Team B prefers a more methodical approach, walking the ball up the court and milking the shot clock.
Points per game says Team A has a better offense, but tempo-free statistics tell us Team B is far more
at running its offense. (And in fairness, the numbers will tend to say Team B, with fewer possessions in its games and thus fewer points, has the more effective defense. Team B might not.)
And then consider two teams from last season: North Carolina ranked 16th nationally in scoring at 76.7 points per game, and they ranked in the top five in shot attempts and made field goals per game.
But were the Tar Heels any good offensively? Not especially. North Carolina averaged 107 points per 100 possessions, ranking 56th in the country. North Carolina was Team A, a team that ran more plays but wasn’t necessarily efficient. The Tar Heels made only 46.3 percent of their two-point baskets, a figure that ranked 224th nationally, according to teamrankings.com.
On the other hand, Florida averaged 71.4 points per game, a figure that ranked 75th nationally. The Gators, though, ranked seventh in Pomeroy’s adjusted offensive efficiency rating and eighth in points per possession. The reason for the discrepancy: Florida took its time. The Gators averaged 62.9 offensive possessions per game, ranking 311th nationally. But Florida was more efficient, ranking in the top 30 in shooting percentage from both 2-point and 3-point range.
Virginia assistant Ritchie McKay, who became enamored with tempo-free stats while the head coach at New Mexico from 2002-07, now uses the numbers to dispel inaccuracies about the Cavaliers’ offense. Under head coach
, Virginia is one of the most deliberate teams in the country.
The Cavs ranked 11th in the ACC in points per game at 64.2, but take tempo out of the mix, and the Cavaliers don’t look so inept. Virginia ranked sixth in the league at 1.04 points per possession.
“If I told you we average 66 points per game or 63 you’d think,
,” McKay says. “But if you look at our numbers and the product we’re putting out on the floor offensively, our percentages, we’re decent in terms of national rank.”
None of this is new. Legendary North Carolina coach Frank McGuire noted points per possession in the 1950s. Dean Smith was also a per-possession analysis adopter.
While tempo-free concepts have been around for decades, what has changed is access.
Pomeroy was a meteorologist (the kind that works for the National Weather Service, not the kind that works on television) when he started running basketball box scores and play-by-play data through computer code, first for sports-ratings.com then his personal site.
By 2005, media mentions in
and elsewhere brought new eyes to his efficiency numbers. In the subscription-based portion of his site, Pomeroy presents further in-depth game-by-game statistics to the coaches who pay. And pay they do.
“When I first start getting ready for a team, I’m using that to give me a road map or a broad picture of strengths and weaknesses relative to all the other teams in the country,” says Kevin Kuwik, an assistant at Dayton and former video coordinator at Ohio State. “That’s one piece of it. I’m using that to pick out some tendencies on how that team might play.”
Kuwik will watch film on opponents’ recent games like any scout. But through kenpom.com, he’s looking at statistics for tempo, indicating how much of a factor transition defense might be. Or he’ll look at assist-to-field goal rate, which may indicate whether a team likes to go one-on-one or prefers to pass. Pomeroy’s statistics also may indicate games in which the opponent struggle in offensive or defensive efficiency earlier in the season.
Then it’s on to the video where Kuwik, like many scouts, turns to Synergy Sports. A video service with archives of college and NBA games, Synergy allows scouts to break down film by player or by situation to isolate habits or tendencies.
Over the course of the season and into the NCAA Tournament, those little edges can make a big difference.
Iowa State has been one of college basketball’s best overachievers in the last two seasons, which have included two NCAA Tournament appearances and two top-four finishes in the Big 12 standings. A major reason is fourth-year coach Fred Hoiberg, who is well-versed in advanced analytics due to his time as Vice President of Basketball Operations with the Minnesota Timberwolves.
Hoiberg’s main focus has been shot selection on both sides of the floor. Naturally, it makes sense to attempt high-percentage shots while forcing an opponent to take low-percentage shots, but Hoiberg and his staff throw some statistical weight behind it. The most valuable shots in basketball are around the rim and 3-point shots. The least valuable is a mid-range 2-pointer — it’s a tough 2-pointer to make and it lacks the reward of a 3-pointer to make it worthwhile.
While NBA teams are installing cameras in arenas in order to analyze the data of how efficient individual players are from certain areas from the court, Hoiberg is adapting the same concepts, albeit at a lower budget.
“We chart what the highest true percentage shot is,” Hoiberg says. “In the NBA it’s the corner three. You want to create as many corner threes as possible because that’s the shortest 3-point shot on the floor. In college it’s not the corner three, but if you can get an uncontested catch-and-shoot three or a shot at the rim, you know you’re accomplishing what your offense is supposed to get you.”
Still, there’s a human element. Hoiberg is reluctant to declare certain areas of the floor off-limits, but charting shots helps him tailor practices and workouts.
“Some guys go strictly by the stats,” Hoiberg says. “If you shoot a low-percentage in the mid-range, they just flat out say you can’t shoot that shot. I don’t go that far with it because I don’t want to take a player’s confidence away.”
Coaches also use advanced stats to play on another human trait: Motivation.
Once Caputo assembles his own data and scouting report for Miami coach
, he’ll condense a few key points for the players.
“There’s nothing stronger than showing them not only that number, but where that number ranks in the country,” Caputo says. “That’s key. When you can say they’re No. 3 in offensive rebounding percentage, you’re not just telling them they’re a great offensive rebounding team, you’re telling them that they’re one of the best in the country, which means hopefully your guys will be more aware that the emphasis needs to be on blocking out or whatever your game plan is.”
The integrated approach has been perhaps one of the reasons basketball — both pro and college — hasn’t had the protracted battle between stats and scouts as Major League Baseball did during the Moneyball era.
Not only has advanced analysis been in use in MLB and the NBA for several years, in college there’s not a rift between the people doing the scouting and the people emphasizing statistics. It’s more a product of function than culture.
“First and foremost, the assistants are the ones who have to do the deep dive and call out every significant little nugget,” Kuwik says. ”As it’s become more prevalent, the assistants were the ones who used it the most. You have some younger head coaches who are a little more aligned to when that started happening in the last couple of years. You’re going to see more and more head coaches be attuned to it.”
That’s the certainly the case for Kuwik and other Ohio State colleagues. Thad Matta’s staffs are an example: Stevens worked under Matta at Butler. Kuwik’s boss at Dayton, Archie Miller, coached under Matta at Ohio State. Illinois’ John Groce, another former Ohio State assistant, is a believer. So is the new Butler coach, Brandon Miller, who worked for both Stevens and Matta over the years.
The use of statistics has spread so much that Pomeroy left his day job to concentrate on analysis full-time.
One the one hand he has his subscription-based site, but he’s also consulted for a handful of college teams including Iowa State and Baylor, plus the Houston Rockets.
“Three or four years ago, it became mainstream enough to see it on graphics in an ESPN broadcast where they don’t have to explain those numbers and exactly what they mean,” Pomeroy says. “That’s when the corner got turned.”
The emphasis is there and it’s spreading. But the prospect of Drew Cannons at every school is a long way off.
“In college you don’t have a front office,” Hoiberg says. “It’s the coaching staff. That’s a pretty big difference.”
The budget to hire a statistical expert on the staff may be the least of the barriers to advanced stats in college basketball. Data collection and play-by-play and shot-tracking data, especially for mid- to low-major programs is not as consistent as that in the NBA. The sample size of 82 games each year in the NBA versus 30 or so games in college gives the pro ranks a reliable sample size.
Another Drew Cannon may be the most accessible part of the equation.
“You’re going to see more and more schools doing it — I’m positive you are,” Oliver says. “Frankly, students are not very expensive. You have colleges with computer science departments, stat departments, math departments, kids who love basketball and they want to contribute.”
For now, knowledge of the numbers and analytics is a skill for a coach, not all that different from drawing up a play or recruiting.
“You can find that one advantage that’s going to help you to win close games you wouldn’t win otherwise,” Hoiberg says. “Is it the ultimate factor? No. But it certainly is a piece of the puzzle when you’re putting everything together.”