LeBron James Destroyed Our Elo Ratings, But Can He Beat The Warriors Again?

It’s become a rite of spring. Every year — or at least every year since LeBron James returned to the Cleveland Cavaliers — our NBA Elo ratings are skeptical of the Cavs when the playoffs begin. And every year, LeBron and Co. have smashed our poor algorithm to bits.

In 2015, the Cavs entered the playoffs with a lukewarm 1631 Elo rating. That’s perfectly respectable, but the sort of rating you might associate with the Los Angeles Clippers or another 50-something-win team that you’d expect to lose in the second round or the conference finals. Instead, Cleveland reached the NBA Finals, losing to the Golden State Warriors in six competitive games even with a depleted roster.

In 2016, the Cavs had a similarly good-but-not-great Elo rating — 1642 — when the playoffs began. But they blew through the Eastern Conference playoffs before beating the 73-win Warriors to win the NBA title, famously overcoming a 3-1 series deficit along the way. Their Elo rating finished at 1759, ranking them among the top 25 teams of all-time and implying that the system had massively underrated them initially.

This year, Elo had the Cavs pegged lower still when the playoffs began last month. Although the Cavs were our preseason favorite to win the Eastern Conference, they slumped at the end of the regular season — losing 13 of their final 22 games, including their last four — and their Elo rating fell all the way to 1545. That isn’t good; it’s the sort of rating you’d normally associate with a No. 6 seed or some other team you’d expect to lose in the first or second round. Accordingly, the Cavs’ chances of winning the title drifted around in the low-to-mid single digits — variously at 2 percent to 5 percent according to our simulations — as the regular season wound down and the playoffs began.

But the Cavs have gone 12-1 in the playoffs and won by an average score of 117-103. Their Elo rating has climbed by almost 150 points, to 1691. They clinched a return to the finals by beating the Boston Celtics by 33 points on Thursday. It’s been dominating stuff.

So has Elo learned its lesson? Well, maybe not. Cleveland’s chances of winning the finals are just 10 percent according to the more advanced, “Carm-ELO” version of our ratings — and 13 percent according to the simpler, original Elo algorithm. Bookmakers also have the Cavs as underdogs, but not as heavily, implying that they have about a 30 percent chance to beat the Warriors again and repeat as champions.

Giving Cleveland only a 10 percent chance is not the hill I want to die on. Our NBA projections are pretty simple, and sports betting markets are pretty sophisticated. While there are occasional exceptions, I’d usually defer to Vegas in the event of a major disagreement.¹ Still, we’ve gotten a lot of questions throughout the playoffs about why Elo hasn’t given the Cavs a better chance. There are basically three reasons — but the one that matters the most right now has nothing to do with the Cavs and everything to do with the Warriors.

Reason No. 1: Elo doesn’t account for teams such as Cleveland finding a higher “gear” in the playoffs. We covered this point extensively before the playoffs began, so I won’t go into too much detail here. Our Elo projections — and most other projection systems — essentially treat regular-season basketball as equivalent to playoff basketball. But LeBron’s teams have a long history of performing at a much higher caliber in the playoffs than in the regular season.

Maybe this is because James and his teammates conserve their energy; there aren’t a lot of high-leverage regular-season games in the Eastern Conference, as evidenced by the fact that the Cavs could play so crappily down the stretch run and still stumble into the No. 2 seed. Maybe it’s because LeBron is a terrific half-court player, and there’s a premium on the half-court game in the playoffs as defenses tighten up. In any event, the assumption that playoff basketball equals regular-season basketball seems to be pretty wrong in the case of the Cavs. This is something we plan on re-evaluating as we retune our NBA models this summer.

Reason No. 2: Elo ratings heavily weight recent performance. That hurt Cleveland before, although it’s starting to help them now. Elo ratings were originally devised for chess, which doesn’t have any such thing as a “season.” Instead, performance continuously fluctuates up and down over time. Our Elo-based sports ratings mostly work the same way.² The more recent the game, the more heavily it gets weighted.

I’d defend this as being the right assumption to make, in general. The degree to which Elo ratings fluctuate from game to game — which is governed by something called the K-factor — has been tested based on tens of thousands of NBA games. Other things held equal, a game played a week ago ought to tell you more than one played six months ago. Elo can be “smart” about catching cases like the 2014-15 Atlanta Hawks, who started out 40-8 but went 20-14 for the rest of the regular season before being swept by Cleveland in the conference finals.

But for a team whose regular-season performance doesn’t tell you much about how they’re going to fare in the playoffs (like the Cavaliers), there isn’t much benefit to doubling down on recent play. Cleveland played pretty well in the first half of the regular season, but middlingly — sometimes even poorly — in the second half. Elo put a lot of emphasis on that late-season slump as the playoffs approached, and that made it more skeptical of the Cavs.

Elo’s philosophy of rapidly adjusting its ratings is benefiting the Cavaliers now, however. Because of their dominance in the playoffs, the Cavs’ current Elo rating has rebounded. Their 1691 is the highest Elo rating they’ve had since Dec. 25, when they were at 1692 and had a 23-6 record after beating the Warriors.

That’s a very good Elo rating. Since the ABA-NBA merger in the 1976-77 season, the average NBA Finals participant has entered the finals with a rating of 1695. So Elo is saying that despite their regular-season struggles, the Cavs are every bit as strong as the typical conference champion.

There’s just one big problem for Cleveland: Golden State.

Reason No. 3: Elo thinks the Warriors are insanely great — one of the two best teams ever, along with the 1995-96 Bulls.

The Warriors’ current Elo rating is 1850. That’s the highest rating a team has held upon entering the NBA Finals. And it’s the second-highest rating a team has had at any point in the regular season or playoffs; the 1995-96 Chicago Bulls peaked at a rating of 1853 after sweeping the first three games of the finals.³ It’s higher than the peak rating of last season’s 73-win Warriors, who topped out at 1839 after starting out the regular season 24-0.

We’ll be publishing a deeper dive on the Warriors next week, but Elo’s affection for them isn’t hard to explain. They’re 27-1 over their last 28 games. That includes a 12-0 record in the playoffs and an average margin of victory of more than 16 points, which is the best playoff scoring margin of all time. And they’ve done all of this in the Western Conference, which is still a lot deeper than the East.⁴ The Warriors are making it look so easy that they may even be underrated by the “eye test,” which tends to reward teams that triumph in the face of adversity. Other than in Game 1 of the Western Conference finals, the Warriors haven’t faced much adversity because they haven’t let their opponents get close.

To put this in perspective, suppose you took an indisputably great team like the 1986-87 Los Angeles Lakers, who went 65-17 in the regular season and entered the NBA Finals with an Elo rating of 1738. Elo would have given the Lakers only a 20 percent chance to win a seven-game series over the Warriors, assuming that the Warriors had home-court advantage (as they will against the Cavs). Compared with that, the Cavaliers’ 10 percent or 13 percent chance doesn’t seem so bad. Still, I’d put a few dimes down on LeBron at Elo’s odds.

CORRECTION (May 30, 4:05 p.m.): An earlier version of a table in this story gave an incorrect winner for the 1951 NBA Finals. It was the Rochester Royals, not the New York Knicks.