Grantland’s Zach Lowe wrote a brilliant piece Tuesday which focused on the innovations from the Toronto Raptors’ analytics team.
Lowe was able to get a close glimpse of how the Raptors have cracked the code with respect to their analysis of big data. The immediate reaction was twofold. The first was a positive reaction to the article but many also wondered why would the Raptors open up the recipe to their secret sauce?
I see revealing this information as a good thing since it will create a healthy debate. Pressing traditional assumptions, when done right, often creates positive change.
The challenge is, of course, for the Raptors coaching staff to embrace the changes and it’s clear from Lowe’s article that some friction exists between what the analytics team is finding versus what is being implemented by the coaching staff. Some of the analysis goes against conventional wisdom which means the findings will take courage to implement.
Let’s take a look at the two prime takeaways from the analysis.
Shoot as many threes as possible
First, the team argued to take many more three-point attempts per game even if the players have a marginal shooting percentage from beyond the arc. Rucker admits almost all coaches (one could throw in fans and media as well) would not take kindly to the trade off of shooting percentage versus the incremental point from a three-point make.
Why not? It's very clear there is a positive trade off. I believe the answer is something many don't want to admit: it causes more (perceived) variance. Some nights the team would shoot 5-of-25 from long range and the fans/media would scream, "why in the heck would you allow the players to jack up so many threes?"
The reality is, if the player's shots were off that much on a given night, they would have perhaps shot 3-of-13 from deep and five of 12 long twos anyways. The difference? One point. Even shooting a robust 50 per cent from those extra long twos (league wide average is 38.3 per cent from 16-23 Feet) would have only made a one-point difference in the game.
The article explains that almost every coach (and media/fans) would choose a two-pointer with 42 per cent accuracy over 28 per cent from beyond the arc.
They are missing three very obvious pieces:
1. It's a neutral proposition at face value: 100 shots from each distance yield 84 points.
2. However, the 28 per cent shooting from three is clearly superior. Why?
Two things: i) more minor, but the offensive rebound rate off a three point miss is higher (albeit slightly); and ii) the more important point: a team has 72 chances out of a 100 for an offensive rebound versus 58 chances out of a 100. For simplicity, let's use the overall NBA offensive rebound rate of 26.6 per cent. Thus, by shooting more threes, a team will have ~ 4 more (19.2 versus 15.4) offensive rebounds per 100 shots. Since most rebounds are inside the key, I'll use 60 per cent FG% as a rough guide for putbacks (note, in several cases, the player getting the offensive board will be fouled -- the league average for free throws is 75.3 per cent from the line). That, in itself, should add another five or six points per 100 shots.
3. Finally, while difficult to quantify from public data, shooting more threes helps spread the floor more. The coaches should at least embrace this.
The reality is the 28 per cent versus 42 per cent debate example is even more glaring when we look at actual data -- especially for the Raptors. The Raptors shoot 35.1 per cent from beyond the arc. The team shoots 36.5 per cent from 16-23 feet ("long 2") and 41.6 per cent from 10-15 feet ("mid range"). Thus, the effective FG% for three-point field goals of 52.6% is vastly superior to these alternatives.
It should be obvious to anyone the team should maximize the three-point attempts.
Weakside defenders should be more aggressive helpers
The article shows examples of the algorithm the analytics team wrote which calculates the optimal positions of defenders versus their actual position from the game data. The optimal positions (using "ghost" players - see the original article) are tracked in order to gain insight how defensive philosophies can be tweaked.
As Lowe points out:
"The ghosts in Toronto's ideal defense are almost always more aggressive helpers than real players, and that's true across the league, according to the analytics team and Toronto's coaching staff. Teams either haven't realized they should be sending even more help toward the middle and the strong side, and sending that help sooner, or they haven't fully convinced players to behave in this way."
The analytics team recognized that actual players were unlikely able to mimic the "ghost" players rapid movements. However, certainly this analysis is quite valuable to coaches to devise schemes to generally send help sooner and more aggressively than conventional wisdom suggests.
So why the controversy?
A couple responses to the article were along the lines of "we should be worried that the Raptors put too much stock into analytics" and "sometimes we forget the human element."
These statements worry me. When you have a competitive advantage why wouldn't you use it? Why wait until other teams figure it out?
The "human element" is a desire to win. If making adjustments wins you more games and takes you deeper into the playoffs than I'm going out on a limb to say the players will be happier.
Are players going to revolt if they are asked to take more threes and less long twos?
If they are asked to help more aggressively and see a return with more shot-clock violations and steals, is this a bad thing?
Of course not.
So why the resistance?
Coaches, like many of us I suppose, gravitate to lower-risk strategies even if it means trading off much higher rewards.
In the article, Raptors assistant coach Micah Nori, noted the same is true with players:
"It can be tough, say coaches, to sell players on drifting so far from their own assignment. Guys don't want to be embarrassed, or see themselves on TV giving up a dunk or an open 3"
It is human nature. By relying on more three-pointers and helping more aggressively on defense, you introduce the potential for more blowout losses. However, you also introduce the potential for several more wins. If a staff has to option to be "in" more games (i.e. keep the score close) and lose more often or to have a few more blowouts but win more often, they will take the road well travelled (the former). Does in really matter if you lost more games by a smaller margin? You either won or lost. Doesn't matter how close it was.
Yet it's clear many fans and several in the media support conventional wisdom. Because if you adopt higher reward strategies but take on more risk, we will all be the first to criticize the blowout: "Raptors shot only 18 per cent from three-point range, yet the coach continued to have the team bomb away" or "The team gambled on defense leaving the opponents a number of wide open looks". There will be some bad nights.
However, there will also be several more wins over the course of a season. There is no rational answer why a middle of the pack team wouldn't use this information to their advantage.
What do the Raptors have to lose?
Play the last games and work some of these strategies into the game plan. They will likely work over a series of games. Next year do you want be a .500 team with more certainly or adopt new and innovative methods that just may lead to several upset wins?
It takes courage as job security is on line but to not embrace and run with a clear competitive advantage would be disappointing.
Leading a new era of basketball intelligence and winning more basketball games is something I strongly endorse.
The fans and media will follow. But the coaching staff needs to lead.