If our NBA model could talk, here’s what it might say about the NBA Finals:
Bleep, bleep, bloop. Based on their accomplishments over the past few seasons, the Golden State Warriors are better than the Toronto Raptors at full strength. However, the Warriors will start the NBA Finals without Kevin Durant, and possibly also without DeMarcus Cousins. To state the obvious, being without those guys makes them a worse team. Meanwhile, Toronto is also a really good team, and its regular-season record somewhat understates its performance because its current lineup is stronger than it was for most of the season. Plus, the Raptors have home-court advantage. Run the numbers, and the Raptors come out as slight favorites in the series. Bloop, bloop, bleep.
Make sense? Well apparently not, at least not to those of you who are wagering your hard-earned income on the series. Betting market prices imply that the Warriors are about 72 percent favorites to win the championship.
We think our NBA forecasts, in their current, improved form, are pretty smart, but we also think sports betting markets tend to be really smart. (Note: This isn’t true for political betting markets, which are mostly pretty dumb.) So we wouldn’t suggest that you go out and wager all your loonies on the Raptors to become the first Candian team to win a title in a “Big 4” sport since … to the chagrin of literally every Canadian NHL team … the 1993 Toronto Blue Jays.
Still, it’s interesting to see the series through our model’s eyes. So while we also talked about our forecast on this week’s edition of Hot Takedown, I want to go through it in more detail here. Basically, I’m going to work through everything in the italicized paragraph, starting with the least controversial claims and moving to the most contentious ones.
“Run the numbers, and the Raptors come out as slight favorites in the series.”
OK, so that’s actually the most contentious claim — we’ll loop back to it at the end. But I do want to point out that “slight” really does mean “slight” in this instance. The Raptors are merely 55 percent favorites in the series, at least based on our current understanding (as of early Wednesday morning) of the injury prognosis for Durant and Cousins. In our election forecasts, we’d label a race like that as a “toss-up.”
“Plus, the Raptors have home-court advantage.”
In a seven-game series between two equal-strength teams, the home team should win about 54 percent of the time, according to our model. So basically, the entirety of the Raptors’ very small edge in the series is because a Game 7 would be played in Toronto. If the Warriors had won two more regular-season games and had home-court advantage instead, they’d be roughly 53 percent favorites to win the series, per our model.
And if anything, our model might be understating the impact of home-court advantage in this series. The Warriors are generally regarded as having one of the biggest home-court advantages in the league, and Toronto is 40-11 at home between the regular season and the playoffs.
“Based on their accomplishments over the past few seasons, the Golden State Warriors are better than the Toronto Raptors at full strength.”
Our NBA team projections are derived from our CARMELO player projections, which use data from the past three seasons plus the current season.
That’s a good thing for the Warriors, because if you based the projections based on this season’s data alone, the Raptors would be more substantial favorites. Three of their five starters — Pascal Siakam, Danny Green, Marc Gasol — have significantly outperformed their preseason CARMELO projections. (Reserve swingman Norman Powell has also outperformed them to a lesser extent.) For the Warriors, conversely, Cousins significantly fell short of his preseason projections — no doubt because of his injuries — and his current projection is probably too optimistic. Klay Thompson and Draymond Green also slightly underperformed their projections, although Green has been great in these playoffs. Andre Iguodala and Kevin Looney have outperformed their projections, but overall, the Warriors are helped by the fact that we’re looking at multiple years of data.
So even though both teams played about equally well this year — the Raptors went 58-24 to the Warriors’ 57-25, but the Warriors had a slightly better point differential and played a slightly tougher schedule — our model would have the Warriors as 65 percent favorites if each team was at full strength to start the series (or 69 percent if the whole series were played on a neutral court).1 This reflects the Warriors’ accomplishments over the past several seasons in addition to having more playoff experience, a factor that our model accounts for — although the Raptors, with former NBA Finals MVP Leonard as well as Green (118 career playoff games), Kyle Lowry (80) and Gasol (77), have plenty of experience of their own.
In other words, our model takes some countermeasures to the fact that veteran, championship-driven teams like the Warriors tend to lollygag their way through the regular season. It looks at longer-term performance, and it accounts for playoff experience, as well as the increased playing time that’s given to top players in the playoffs, which helps top-heavy teams like Golden State. Is it doing enough to account for those factors? Maybe not, and our model has had plenty of challenges with teams like the Warriors and LeBron James’s Cavaliers in the past. But it’s at least aware of these issues, and it doesn’t hold the Warriors’ good-but-not-great regular season all that much against them.
“Toronto is also a really good team, and its regular-season record somewhat understates its performance because its current lineup is stronger than it was for most of the season.”
The Raptors were often without the services of what our model regards as their two best players. They played 22 regular-season games without Leonard, who was frequently rested for “load management,” as well as 17 games without Lowry. In addition, they only acquired Gasol in February, and he’s a significantly better player than the center he replaced, Jonas Valanciunas, according to our model. It also took some time for Toronto to take full advantage of Siakam, who played fewer minutes and took fewer shots early in the season.
Thus, the current version of the Raptors is associated with an Elo rating that would peg them not as a 58-win team, but somewhere in the mid-60s instead.
Until recently, however, that elite version of the Raptors existed mostly on paper. The Lowry-Green-Leonard-Siakam-Gasol lineup played only 161 regular-season minutes, although it was highly effective when it did play, outscoring opponents by 12.2 points per 100 possessions. That group has now played 314 minutes together in the playoffs, and — somewhat remarkably given that the Raptors have been playing extremely tough competition in the Milwaukee Bucks and Philadelphia 76ers — it’s still outscoring opponents by 12.1 points per 100 possessions.
So Toronto’s starting lineup has begun to prove itself — you have to be really good to win four straight games against the Bucks, who were the NBA’s best regular-season team. And I should probably mention that our model also had the Raptors slightly favored against Milwaukee2 despite the Bucks having been heavily favored in Las Vegas.
“However, the Warriors will start the NBA Finals without Kevin Durant, and possibly also without DeMarcus Cousins. To state the obvious, being without those guys makes them a worse team.”
Durant is out for at least Game 1, with no clear timetable for his return. Cousins is questionable for Game 1, but from the tone of the Warriors’ comments, he looks highly likely to return at some point in the series.
You can see the impact of Golden State’s injuries in the evolving point spreads that our model establishes for each game of the series. In Game 1 — with Durant out and Cousins 50 percent likely to play (based on the very rough science of translating the Warriors’ vague injury guidance into probabilities) — the Raptors are 6-point favorites at home, per our model. In the event of a Game 7 in Toronto, by which point we assume that Cousins is definitely back and Durant is 80 percent likely to play — they’d be only 1-point favorites, conversely. Toronto would also be 4.5-point underdogs on the road in Oakland in Game 6. If the Raptors don’t strike early in the series, the odds will shift dramatically against them.
|Likelihood of playing|
|Game||Date||Location||Durant||Cousins||FTE point spread|
|1||May 30||Toronto||0%||50%||Raptors by 6|
|2||June 2||Toronto||10||60||Raptors by 5|
|3||June 5||Golden State||30||70||Warriors by 2.5|
|4||June 7||Golden State||40||80||Warriors by 3|
|5||June 10||Toronto||50||90||Raptors by 2.5|
|6||June 13||Golden State||70||100||Warriors by 4.5|
|7||June 16||Toronto||80||100||Raptors by 1|
You’d think that all of that seems pretty reasonable. Our model is saying that having Durant and Cousins healthy-ish instead of injured-ish is worth about 5 points per game to the Warriors.
But that’s not the narrative surrounding the Warriors at the moment. Instead, the stat you’ve probably heard is this one: 31-1. That is, the Warriors are 31-1 in their last 32 games without Durant but with Stephen Curry playing. This has lead to plenty of talk-radio chatter about whether the Warriors are better off without Durant, who has an option to become a free agent at the end of the season.
Like most narratives, that one leaves out some of the messy details. Our ESPN colleague Kevin Pelton has a long, detailed breakdown of the Warriors’ play with and without Durant. I’d suggest you read the whole thing. For one thing, that 31-1 record overstates the case somewhat, since it arbitrarily ignores the first six games that the Warriors played without KD (counting those, they’re 34-4) and since their point differential wasn’t quite as strong as their record in those games would suggest. Those games were also played against a fairly easy schedule.
Perhaps more importantly, Pelton finds based on game-by-game data that being without Durant lowered the Warriors’ ceiling. With both Durant and Curry in the lineup, the Warriors had so much firepower that they could take possessions off against mediocre teams, especially on the defensive end. In the NBA Finals, however, the Warriors will presumably be playing every possession at close to maximum effort, with or without Durant. So they’re deprived of a top gear they would have had with him healthy.
We can also look at the Warriors’ lineup data over the past three seasons (including both the regular season and the playoffs), which accounts for their performance on a possession-by-possession basis with various combinations of players. With both Curry and Durant on the floor, the Warriors outscored opponents by a dominating 15.2 points per 100 possessions. With Curry only, that number falls to 11.8 points per 100 possessions. That’s still a very good number — Curry is impossibly good! — but it’s in the same ballpark as the Raptors’ current starting lineup, and the Raptors have more depth and home-court advantage.
|Lineup||Curry||Durant||Off. Rating||Def. Rating||Net Rating|
|Curry + KD||✓||✓||115.0||99.8||+15.2|
The most obvious conclusion from the lineup data, in fact, is that Curry is a lot better than Durant. With Durant but not Curry playing, the Warriors outscore opponents by a pedestrian 1.7 points per 100 possessions. But that’s not the same as saying they’re better without Durant. That’s especially true on offense, when there don’t appear to be any diminishing returns from having both Durant and Curry in the lineup at once.
Dig a little deeper, and you find that while Curry and Durant work just fine as a tandem, there may be some diminishing returns from playing Durant and Thompson together. Lineups with Durant and Klay playing but Curry off the floor have been mediocre, perhaps because Durant doesn’t like to pass and Thompson relies heavily on assisted field goals. Furthermore, lineups with Curry and Durant but without Thompson have been better than lineups with all three together. The Warriors give up a bit of offense in those lineups, but they make up for it with superior defense by having players such as Iguodala on the floor.
|Lineup||Curry||Durant||Thompson||Off Rating||Def Rating||Net Rating|
|Curry + KD + Klay||✓||✓||✓||115.9||100.9||+15.0|
|Curry + KD||✓||✓||111.7||96.1||+15.6|
|Curry + Klay||✓||✓||109.0||97.0||+12.0|
|KD + Klay||✓||✓||101.2||100.6||+0.6|
|None of them||91.6||100.5||-8.9|
But the thing is, our projections actually account for all of this on-court/off-court data, at least to some extent. One of the metrics we use to fuel our projections, Real Plus-Minus (RPM), is largely based on the lineup data. So the fact the Warriors have played quite well with Curry but without Durant is accounted for in their respective ratings. Our forecasts think that Curry is quite a bit better than Durant — if Steph were injured instead of KD, it would really have Golden State in trouble.
You can also go too far in looking at the on-court, off-court stats. They can be noisy, and there are also a lot of technical complications in evaluating so many five-player lineup combinations together. In fact, we’ve found that RPM (which itself is a blend of box score statistics3 and lineup data) actually goes slightly too far in using the lineup data, so we hedge against it by blending it with another metric based on box score statistics called Box Plus/Minus or BPM.
“Run the numbers, and the Raptors come out as slight favorites in the series.”
So do I — Nate as a basketball fan, not as a model co-designer — really buy what our model says?
I mostly buy the part about the Raptors being better than they’re given credit for. Their current starting lineup has been very good, and I can imagine the betting public sleeping on it a bit because it’s involved several fairly subtle changes (e.g. upgrading Valanciunas for Gasol, Leonard playing every game, etc.). Nor do I see any obvious flaws with the Raptors, who can work effectively as either an up-tempo team (perhaps with Gasol off the floor) or in the half-court, with Leonard draining midrange jumpers and corner threes.
Leonard’s health is a concern, however, particularly insofar as it could affect his ability to effectively defend Curry, a tempting matchup for the Raptors.
As for how the model is evaluating the Warriors, I’m less sure. As I mentioned, the metrics behind our model (RPM and BPM) don’t actually like Durant that much; while he was repeatedly going off for massive games against the Clippers and Rockets, a few of us were complaining that the model underrated him. But there are a couple of things that worry me. First, although we have a few tricks to try to account for the Warriors’ variable effort level, their indifference during parts of the past few regular seasons may be contaminating the data to some degree. Second, our model tends to assume that building a lineup is a fairly linear process, when it isn’t. The Warriors are insulated against the loss of Durant to some degree because Thompson functions better as Curry’s Splash Brother sidekick than as a third wheel in Curry-Durant lineups.
The handful of minutes each game that the Warriors play without Curry on the floor are liable to be a disaster, however, and if Leonard somehow can bottle up Curry the same way he did Giannis Antetokounmpo, the Warriors are probably toast. And I think our model actually overrates Cousins, who isn’t likely to play up to his projections while recovering from his multiple injuries.
Overall, I think our model is mostly right about the Raptors, but more wrong than right about the Warriors. Since it only has the Raptors as extremely narrow favorites, that might be enough to tip the balance slightly in Golden State’s favor. But I find it hard to contemplate how the Warriors can be as heavy as 3-to-1 favorites, as they nearly are in Vegas. There is, if nothing else, a lot of uncertainty about how well the Warriors can play against a top-level team without Durant — I’m sorry if I don’t regard the Portland Trail Blazers as a top-level team — and the Raptors are good enough that the Warriors will probably have to bring their A-game.
Bleep, bleep, bloop.
From ABC News:
Check out our latest NBA predictions.