Two Week Windows: A Basketball Evaluation Manifesto
USA Today Sports
We’re proud to share this guest post from Brad Stenger, a Rochester, New York, journalist and researcher. You can find more from Brad at nbagraphs.tumblr.com and on Twitter, @nbagraphs.
Improvement never ends in the NBA. Each improvement, and they come daily, is a force-multiplier for the organization tuned to evolving from the team they are to the team they can become. A lot of analysts dismiss the improvement and call it “recency bias” or “small sample size,” expecting and assuming a “regression to the mean,” but I think there’s plenty of information to draw solid evidence-based insight. In fact I believe that two weeks is the right time frame for analyzing and understanding basketball.
The range of possible improvements by players, coaches and management is a long, long list. Players improve their athleticism by training and they steadily lose fitness if game schedules preclude the recovery time to develop endurance. Players work on their games, for offense and defense. Coaches invent tactics to create player and lineup advantages. Team chemistry can make a team something greater than the sum of its parts or something less. Management adds, subtracts and incentivizes personnel. Improvement comes in a thousand different forms.
Players and teams also endure a brutal schedule that taxes their physical and mental well-being. Decline, the flip side of the improvement coin, is also everywhere you look.
Take a look at this game graphic from Brooklyn-Charlotte last night. An algorithm transforms play-by-play data to show the second-by-second +/- for each player along with shots, free throws, assists, rebounds, turnovers, steals, blocks and fouls. (Game graphs are posted daily at nbagraphs.tumblr.com and on Twitter, @nbagraphs.) See the Nets’ 11-man rotation at work during the first half. Tell me that’s not some crazy stuff.
More orange means more + on the graph. Pierce and Williams are thriving as Kidd goes long rotation for Brooklyn. And Jefferson was suffering for Charlotte, as Garnett, Blatche and Plumlee took turns on him. This is just one 48-minute game. The Nets have played 7 of these in the past 14 days.
That’s a lot. I’m saying that 14 days is a consequential time period for basketball players and teams, especially with the demanding NBA schedule, and doubly so if they are sincerely interested in improving, which pretty much means the same things as winning. The two weeks represent a significant change in the player’s health and fitness. It’s also enough time for confidence to emerge in a successful new skill and, realistically, enough time for opponents to recognize that success and/or confidence and adapt a response.
For basketball players in-season health and fitness is ever-changing and it’s a challenge to maintain at consistently high levels given the ongoing game schedule and the burdens of travel. Most rotation-level players lose 20-30% of their peak fitness in a matter of weeks whether that peak comes at the start or in the middle of the season.
It’s a fine line between working hard to gain fitness and strength (more accurately, keep from losing it) and overworking in a way that increases injury risk and can decrease actual athletic performance. There are bound to be peaks and valleys, usually within the two week window.
Every once in a while though the peaks come like waves, one after the other. Often it’s a young player who keeps hitting new peak performance levels. Here two weeks is a good timeframe for gauging whether the ascending athlete can sustain whatever production gains he is making. (It happened a few weeks ago with James Johnson. It might be happening again right now with Jimmer Fredette.)
Two weeks is also a good time window for examining improvements that players have made in their games as a result of new or improved skills. It’s ample time for the basketball universe to absorb and reflect a significant adaption by a player.
What’s challenging from an analyst’s perspective is monitoring the evidence, and noticing the change as it happens. To help with that I have two things to rely on.
One is a proper set of tools to manage the data and spot the changes. Game-by-game data is collected second-by-second for interactive visualization, like the Nets-Bobcats game graph. For the 5-8 games that a team plays in two weeks, a rotation player will get 1500-2500 seconds on the court, plenty of evidence for producing sound insights.
The data is also not reduced to statistics, whenever possible. It’s annoying to hear “small sample” in article after article. Samples come into play for statistical analysis but not all evidence-based analysis is statistical. I created the visualization for second-by-second +/- to completely represent game data without any of the information loss that comes with statistics.
Take a look at Houston-Washington, also from last night. See the mostly all orange band for Dwight Howard and also see the smaller gray and dark gray sections where Omer Asik subbed for him. The confident Howard had a great game and Asik is probably still bummed about his career misfortunes.
This next graphic shows Howard’s second-by-second +/- for the last two weeks and it’s clear that he has plenty of reason to feel confident. His minutes trend strongly positive game in, game out. (An interactive version of the static graphic is at http://bit.ly/1iS6v0I)
Compare that to Howard’s late-December graph (interactive) when he and the inconsistent Rockets were failing to produce on a consistent basis. At the time I called Howard an every-other-game All Star.
The second thing is a key operating assumption which doubles as my central hypothesis. It’s the idea that sustained improvement (or decline) is multi-faceted. Teams improve when players improve. Players improve when their skills advance and their physical athletic ability increases. When players improve coaches have more options and opportunities to innovate tactically. And management can innovate with what they do, adapting and incentivizing personnel.
Teams and players improve relative to other teams and players, something that adds a cognitive creative dimension to the two-week window. Individual and collective intelligence will also get better (or worse) as short-term gains progress (or declines occur) to become long-term improvements. Pay attention to getting better over the long haul and you’re smart, like the Spurs or Pacers or Heat. Fail to really give a shit and you’re the Cavaliers or Bucks or Pistons.
The simple version of the idea: Everything gets better (or worse) together. There are physical improvements, skill-based improvements and cognitive/creative improvements at every level-player, team, coach, management-as well as declines, and they don’t occur in isolation. And in many cases, the improvement (or decline) is visible in the two week window.
This is Deron Williams‘ last two weeks’ graph (interactive). It should be easy to spot when Kidd started with the long rotation. Things are getting better in Brooklyn.
Things aren’t looking up for every NBA team. Look at what’s going on with Philadelphia and Evan Turner (interactive). A cautionary tale, this is what happens when a team does none of things it could do to get better. Things gets worse, much much worse.
So I have an eye on this moving 14-day window throughout the NBA season. The season has its phases.
Players and teams figure things out at the start and then hit their stride (probably) a few weeks in. Inevitably the schedule and travel takes its toll. Players decline or get hurt and the teams deal with it. Yet some players improve, take on larger roles and, in the best cases, establish their NBA credentials.
With the new year comes mid-season, the dog days when players and teams grind it out, trying to improve but rarely practicing. The strata of teams is set. Losers mostly lose. Mid-levels mostly beat the losers and mostly lose to the Elites. Elites playing other Elites is usually exciting. And then there’s the last wildcard strata, teams that improve or decline as they bounce between strata.
With the All-Star break and the college game’s March Madness, the end is in sight. Elites are likely confident in their playoff situation and make March and April into a second training camp to prepare. Improving Mid-levels experiment, looking for something that will make them Elite. For me this is the best part of the season. The best teams have a key question to answer: Can Indiana find bench lineups that don’t lose points during the 2nd quarter? How will the Thunder and the Clippers do bringing back Westbrook and Paul? Can Miami and Portland settle into player rotations that put them in position to win games but don’t overwork their best players? Will San Antonio’s best players ramp up properly in March to play their best in April, May and, they hope, June?
Two weeks is about 10% the season, enough time to fit the window within a phase of the season, or to capture the transition from one phase to the next.
The plan is use this window to identify trends for NBA players and team lineups and write up the most interesting cases. I hope it’s something I hope you’ll make time for.
Thank you for reading my manifesto.