The Hidden Reality of the 2015 NBA Finals

Anyone who’s driven a car knows the unease of when a small animal darts in front of the vehicle and escapes being hit by a split second. Many animals, especially birds, seem to be so wholly ignorant of the roadway that they dash out in a suicidal swerve, but that’s a human-centric, ignorant viewpoint. Recent research has found that small animals, mostly those with high metabolic rates, perceive time in finer grain details than humans do. A bird can dodge a car with ease because the car is effectively a lot slower and the path around a lot safer. We, as humans, are fine with crossing roads as long as a car is far away so we have enough of a buffer between the moving vehicle, and it’s the same with birds except that because of how they see time they only need a smaller safety buffer (and, naturally, they’re quicker too.)

The perception of time — an individual’s own “frame-rate,” if you will — is an added dimension in understanding how we all interact with our environments. This is something that varies across people too, and you can make that connection and realize how this is vital to athletes. A pass that seems impossible to most viewers is normal to an athlete who can experience time in smaller chunks and process the information quicker. Even if you don’t physically move faster, having a faster reaction time is extremely useful in sports where decisions are made at a high rate while everything is moving.

Around the time the aforementioned research was released, the NBA had SportVU tracking cameras installed and every arena. Critics may decry the slow domination of statistics into the game, seeping into every corner and affecting some of the most powerful teams and players, but there’s something to using a tool that can handle things we cannot. An NBA fan can track a couplet on the floor — a ballhandler and his defender — fairly well and can also keep an eye on a roving defender or two, but it’s entirely different to track every single player on the court along with the ball every 0.025 seconds.

This is a new way of understanding basketball, and it’s going to take a while to use this information appropriately.

SportVU Individual Player Tracking

Using a method I’ve outlined before, I’ve collected nearest defender stats for Curry and LeBron James, showing the effectiveness of their defenders by points per possession allowed. I do this by thinking of a possession in a granular way — for example, if you guard a player for half the possession you accrue 0.5 possessions guarded and 50% of the points scored by the player guarded, if applicable. There are surely better adjustments, but one reason I like this method is that it gives help defenders a bit of the credit if things work well and a bit of the blame if the player scores. Also, it’s one of the few stats that includes ball denial, which is usually ignored because it’s tough to measure it directly and virtually impossible with box score stats. (I use two separate methods, by the way. Adjusted possessions tweak the nearest defender calculation by adding a penalization for not being in front of the player.)

The table below shows every defender (with at least ten possessions) for LeBron James in seven games, as it includes one regular season game. Iguodala won the finals MVP partly for his ability to hold LeBron to low shooting numbers, and it holds up here — he had the second best points allowed rate and guarded him for the majority of the total possessions. Strangely, Curry has the best points allowed rate. If you want to double-check my results, here are similar stats from SportVU stats.NBA.com released for one game and go through the shotlogs on their site: LeBron was 3/10 when Curry was the nearest defender. Of course, Curry only guarded roughly 35 to 40 possessions, so there’s some noise here clouding his real value, but he also did a good job helping off a poorer offensive player.

Table: LeBron’s defenders, finals + Feb. 26th game

Turning to Curry’s defenders, the other storyline of the finals was Dellavedova’s defense. The numbers reflect this: Curry did not score at high rate when the Aussie was nearby, but it might be overstated. Shumpert had good numbers too, and so did Irving in the regular season and the first game of the finals, highlighted by his game-saving block. No one else, including LeBron, had strong numbers, and Tristan Thompson was decimated. Interestingly, smaller defenders did significantly better, and big men were more often a nearest defender than with LeBron’s set of players, which might indicate the amount of movement and switching involved with Golden State’s play.

Table: Curry’s defenders, finals + Feb. 26th game

There’s some work that needs to be done with this defensive measure — it’s tough to see how much power this has without some sort of out-of-sample test showing how reproducible the numbers are. You can sloppily defend someone and be bailed out by a few missed open shots, but over a large enough set of possessions those issues will be (partly) counteracted. That will take a lot of data, and probably another season to definitively answer the questions. But the stats could be used to great effect in concert with our subjective knowledge of individual defenders — it’s best two use both.

What Happened, Only Happened Once

Too many people succumb to the “one world” fallacy where the results we see on the scoreboard mean that’s the only possible outcome and we stumble through post hoc analysis to explain all the events that apparently led to what happened. Everyone who knows basketball well understands the role of luck. Sometimes the shots don’t fall (for either side), and sometimes there are extraneous factors like injuries — you can’t control everything. If it takes Golden State six games to eliminate a hobbled Cleveland team, then people discount them as true champions and ignore the nearly 100 games before that. But what do you actually know?

Using the method outlined here, I completed the numbers for the disparity between the points you’d expect from both teams on jump shots and the points they actually produced based on shot distance, defender distance, whether or not it’s a catch-and-shoot attempt, and the shot clock. This is also adjusted by how every player shot over the past two seasons with the 2015 season weighed twice as heavily. Now here’s the usual caveat: it’s not perfect, it’s missing some factors, the shotlogs aren’t perfect, etc.

Looking at all six games, the Warriors scored 17.3 fewer total points than expected on jump shots. However, the Cavaliers had them beat: they scored 46.3 fewer points for an average of 7.7 per game. That’s enough to swing a series. This means, essentially, that the Warriors weren’t just putting the Cavaliers in worse positions to score — the Cavs just missed. You can see in the below table that it wasn’t just LeBron, whose jump shot never showed up for the finals; it was everyone. J.R. Smith, who by most standards was the most valuable offensive player outside of their big three for most of the season, was completely taken out of the series and Dellavedova received praise for his defense but killed his own team on offense.

Table: Expected points on 10-30 ft. shot attempts compared to actual points, CLE

Although Curry did not start hot, he eventually found his jump shot and scored at a pretty high rate considering the stage and all the extra attention he got (my method doesn’t cover everything, including things like double teams.) Both Klay Thompson and Green shot under their historic norms, but at least Golden State had Iguodala and Barbosa performing well. Iguodala’s jump shot has been infamously shaky for a long time, but he hit enough jumpers to help seal the deal and clinch the championship. Side note: for a team known for its shooting, they really only have two good shooters.

Table: Expected points on 10-30 ft. shot attempts compared to actual points, GSW

Defenses don’t have a complete and large influence on jump shots, especially open ones on the perimeter, so it’s good to take a step back and look at how players or teams have historically shot with the same type of attempts. Remember, earlier in the series, Golden State was under-performing, and they completely turned it around. (There’s a decent chance Cleveland simply ran out of steam with LeBron doing everything possible, Dellavedova cramping, and JR Smith being … well, who knows with him.) A good defense does well long-term by controlling what’s actually possible to control, not missed free throws and open three-pointers.

With the recency bias people have, there were immediate questions about Golden State’s legitimacy in the “all-time great teams” pantheon. But that’s why we have an 82 game season and multiple playoff rounds. We should have learned from the previous games that this was a truly great team. I’d much rather judge a team based on 100 games, along with knowledge from previous seasons, than a mere six games. But the league has a “title or bust” philosophy, and we’re conditioned to dismiss other games. If you’re wondering if the finals were truly representative of a team, look back and learn from the past.

Historic Footnotes

If you want to explain to a skeptic why the finals actually proved the Warriors were elite, just say they found a test in LeBron James playing at an unprecedented level, the opposing defense having the right strategy for them, and their shooters struggled to open the series. But they were smart and deep enough to turn things around — that’s a great sign of consistent long-term value. And, of course, if you worry about their smallball approach against bigger teams, let’s just note that they beat Anthony Davis, Marc Gasol with Zach Randolph, Dwight Howard, and Mozgov with Tristan Thompson to win the title.

Golden State’s road to the title wasn’t particularly tough, but those roads usually aren’t, which is part of why those teams win. It’s not even close to the easiest title road — in 1987, the Lakers played through teams with 37, 42, and 39 wins, respectively, to make the finals where they faced a Boston team with an injured Kevin McHale (broken foot.) Yet we still regard that LA team highly. As the years go by, we forget all the details in the season that contextualize championships including major injuries and other factors. The Warriors are dismissed because they didn’t play the defending champions, but the Spurs were fairly healthy and just plain lost to the Clippers, who in turn lost to the Rockets — and Golden State beat them. In fact, it’s fairly common that a defending championship team is taken out of the race in the playoffs the following year by not having their best player. It happened in 2009 (Garnett injury), 2000 (Duncan injury), 1999 (Jordan retirement), 1994 (first Jordan retirement), 1989 (Magic injured during series), and 1978 (Walton injury) — not to mention all the other injuries that affect a season like Westbrook in 2013.

Curry’s praised has dwindled a bit in the finals compared to his coronation during the regular season, but he performed pretty well against some tough defenses and was the driving force on offense to eliminate the Cavaliers at the end of the series. Looking at offensive box-score plus/minus, there were only four playoff seasons with higher individual ratings  — one from LeBron and three from Jordan. By that box score measure, which includes a trade-off between shot volume and efficiency, he had a better run than LeBron.

That’s also a prelude to the popular argument that Curry deserved the finals MVP — and it’s easy to see why, given the conventional stats. Using my finals MVP model detailed here, Curry’s odds at winning were at 61.5% based on a game score model that included box score stats with a boost for shot volume and playmaking. LeBron was second at 23.3%, while Iguodala lagged behind at 7.7%. Models are great at working with data they understand, but defensive team-first players are tough to judge through basic stats. Wes Unseld, for example, actually had the lowest calculated odds at 4.9% when he won in 1978. There’s also the power of the narrative, as the series was about LeBron James.

However, if you argue Iguodala has this extra value not seen by conventional stats, the same is true of Stephen Curry, king of the new SportVU stat known as “gravity.” Using my SportVU data for the finals, whenever Curry was 10 to 30 feet from the basket and at least 10 feet from the ball, defenders on average were 5.7 feet away, which is quite close for off-ball, away-from-the-basket defense. Iguodala’s defenders, by contrast, were 10.0 feet away — and they even hid Mozgov on him so their big man could stay close to the rim. That’s compounded by the fact that Iggy was more often playing off-ball in these scenarios, as Curry handled the ball more, naturally. This gums up an offense and gives the opposing team more freedom.

There’s a depth to basketball we’re trying to uncover with all these techniques, from the revolutionary plus/minus stats that searched for indirect influence to player tracking. A lot happened during the series we’ll probably never fully understand, even though it was thoroughly analyzed — that’s the hidden reality to the finals. We can watch the game and follow who’s guarding whom, and how well they’re doing it, but we don’t have the perceptive tools to pick up on everything. The next time you see a squirrel or bird dart across the road, remember that they process time much quicker than you can, and it completely changes how they move through the world. So what are you missing when you watch an NBA game?