On Thursday, we rolled out our new NBA metric, named RAPTOR — the Robust Algorithm (using) Player Tracking (and) On/Off Ratings — along with our new 2019-20 RAPTOR-fueled player and team forecasts, and a thorough explanation of the system from my boss, Nate Silver. But since we released an early, pre-RAPTOR version of our predictions back in July, I wanted to dig into some of the biggest changes that our new method has brought to the team forecasts on the eve of the regular season.
As I mentioned above, we have posted a very detailed explanation of the RAPTOR metric. In short, though, we think this new system is an upgrade over our old CARMELO forecasts (which were powered by ESPN’s Real Plus-Minus and Box Plus/Minus data from Basketball-Reference.com) because RAPTOR uses not just the box score metrics but also player tracking numbers from the NBA and play-by-play data beyond just plus/minus (which it also incorporates).
These modifications mean a comprehensive implementation of shot defense — in the form of our DRAYMOND stat — to go with details like defensive credit or blame depending on which offensive player we estimate a defender was assigned to; the value of an assist and made basket varying by the type of shot; the value of a rebound depending on whether it was contested or deferred; extra value assigned to events outside box scores like offensive fouls drawn; an extra penalty for using up a team’s time of possession without generating a valuable offensive action, etc. The laundry list of small-but-important elements we’ve included in RAPTOR is quite long.
Who was affected?
In practical terms, this has resulted in certain players seeing a boost in value, while others have suffered a decline. I plan to write more on the players who’ve changed the most under RAPTOR in the future, but one big takeaway is that big men are, broadly speaking, going to be viewed as less valuable than they were before, while perimeter players look more valuable. Of the 100 players who lost the most per-possession value in the switch from RPM/BPM to RAPTOR, 60 are centers and 32 are power forwards. Meanwhile, of the 100 players whose ratings improved the most, 43 are point guards, 41 are shooting guards and 13 are small forwards.
Why? In general, RAPTOR gives more credit to players who create shots — for themselves and others — and takes the corresponding amount away from the players who benefited from those acts of creation. (For instance, big men with low usage rates and a high rate of assisted field goals can tend to be less valuable than their personal efficiency metrics suggest.) It also puts more emphasis on floor spacing, so bigs who can’t shoot become less valuable than you’d expect based on the rest of their stats. On defense, rim protection remains important, but RAPTOR has also added more precise measures of perimeter D, meaning good wing defenders have taken away some value from defensive bigs in the zero-sum ratings game. And finally, RAPTOR recognizes that not all rebounds are created equal, reducing the value of some players’ rebounding stats if a disproportionate number of their boards are uncontested.
All of these changes mainly benefit teams whose stars are primarily two-way perimeter players — think Kawhi Leonard or Paul George (hello, Clippers!) — and hurt those whose top players can’t shoot (sorry, Ben Simmons), don’t create much of their own offense (Al Horford) or are overrated by standard metrics on defense (Russell Westbrook … and, dare I also say, LeBron James???). The changes weren’t huge across the entire league — the typical team’s over/under win total changed by +/-1.8 wins on average because of the new metric — but there were a few important cases where a team’s outlook was affected more than a little by RAPTOR.
Winners and losers
The biggest winner from this change has to be the Los Angeles Clippers, whose odds of winning the NBA title have risen from a paltry 5 percent in our way-too-early summer forecast to 17 percent in the official preseason prediction. (They went from ranking fourth in the West alone by championship odds to second in the entire league.) The biggest reason is that RAPTOR really likes Leonard and George more than other advanced metrics do:
RAPTOR loves the new pair of Clippers
Season-by-season plus/minus ratings from various sources for Kawhi Leonard and Paul George, 2017-19
|Kawhi Leonard||Paul George|
RAPTOR was much higher on both players than either RPM or BPM last season, and that’s generally been the rule for the past handful of years. (RAPTOR also thinks Leonard and George will get much better in the playoffs, with each improving their player ratings by more than a half-point per 100 possessions in the postseason.)
Other teams who look better under RAPTOR include the Oklahoma City Thunder, whose core of Chris Paul, Danilo Gallinari, Shai Gilgeous-Alexander, Steven Adams, Dennis Schroder and Andre Roberson all saw a ratings bump when we plugged in the RAPTOR numbers — although at 2 percent to win the West, it probably still won’t matter for OKC. Among those teams that have an actual shot at the title, the Milwaukee Bucks’ championship odds are up by 5 percentage points, the Denver Nuggets are up by 3 points, and the Rockets are up by 2 points.
Who’s gained ground since summer?
Most improved teams according to change in 2019-20 NBA championship odds between July 16 and today, including new RAPTOR forecasts
|Proj. Wins||Finals Odds||Title Odds|
But for every team whose fortunes improved under our new metric, there have to be others who look worse based on the updated forecast. Chief among those is the Philadelphia 76ers, who were championship favorites in the summer with a 27 percent chance of winning it all — but now they are only fourth in title odds, at 11 percent.
What exactly happened to the Sixers? Joel Embiid still rates as a star; his RAPTOR wins above replacement of 13.3 ranked him among the NBA’s top 10 players last season. And newcomer Josh Richardson looks a lot better in RAPTOR than in more conventional advanced stats. But RAPTOR likes the rest of Philly’s heavy hitters a lot less, including Tobias Harris (whose projected plus/minus decreased by 0.2 points per 100 possessions under RAPTOR); Mike Scott (-1.0); Horford (-1.5); and, of course, Simmons, whose decline of 2.4 points per 100 possessions under RAPTOR ranks him as one of our new metric’s biggest victims.
Why does RAPTOR hate Simmons? His lack of floor spacing ability (zero career 3-pointers — not including the preseason) is held against him on offense, as are his many turnovers and his high time of possession relative to points generated. At the same time, very few players saw their defensive value drop more under RAPTOR than Simmons, who ranked 15th among point guards in defensive RPM but actually graded as a well-below-average defender according to RAPTOR. His poor defensive plus/minus plays a role there, but the “box” portion of defensive RAPTOR — which includes shot defense, opponent production, adjusted rebounding and offensive fouls drawn — hurt his rating just as much. More than perhaps any other current player, Simmons might be the poster child for somebody whose advanced metrics look worse and worse as you add more and more factors from player tracking and play-by-play data.
Who’s lost ground since summer?
Most diminished teams according to change in 2019-20 NBA championship odds between July 16 and today, including new RAPTOR forecasts
|Proj. Wins||Finals Odds||Title Odds|
Beyond Simmons and the Sixers, the Los Angeles Lakers are the other team whose forecast is much more bearish after our switch from CARMELO to RAPTOR. And the reason is shocking: LeBron James looks worse in RAPTOR than other metrics! Specifically, it thinks James has lost a lot of value at the defensive end over the past few seasons:
Is LeBron still a dominating defender?
RAPTOR ratings based on box score (“box”), on-vs.-off court data (“on/off”) and an overall composite for LeBron James by season, 2016-2020
|Raptor “Box”||Raptor On/Off||Raptor Overall|
Unlike in an earlier version, our new forecast is also calling for James to suffer an age-based decline, like many of his most similar historical comparables did in their age-35 seasons. Whatever the reason for King James’s lack of RAPTOR love, the Lakers are second only to the 76ers in terms of teams that lost the most championship probability and projected wins between our old model in the summer and our new model now. (It’s an especially bitter pill because, again, the crosstown rival Clippers were the team who benefited the most from the switch.)
Even so, we give the Lakers the seventh-best title odds of any team, which at least places them among a deep group of contenders in the West, headlined by our new championship favorite — the Houston Rockets. (RAPTOR is down on newcomer Westbrook but very high on James Harden.) The Bucks are the new favorite in the East, though the Sixers are still right behind them in terms of finals odds (30 percent to the Bucks’ 39 percent) despite being hurt by RAPTOR.
In other words, there are a lot of intriguing teams out there! Going into the season, our shiny new forecast gives five different clubs — the Rockets, Clippers, Bucks, Sixers and Warriors — at least a 10 percent chance of winning the NBA title, with six more at 1 percent or better and nobody over Houston at 26 percent. For comparison, the Warriors went into last season with a whopping 49 percent title probability, and only two teams were in double-digits; two seasons ago, Golden State was at 39 percent with three teams in double-digits; three seasons ago, the Warriors were at 55 percent (!) with only two teams in double-digits.
So no matter what happens, this should be one of the most wide-open NBA seasons in recent memory — and RAPTOR will help you keep track of the best players and teams all year long.
Check out our latest NBA predictions.