The Warriors hold the Western Conference Champion trophy their Game Five win against the Dallas Mavericks at Chase Center on May 26, 2022 in San Francisco.

This post also appears on Indignity.

Last time we checked in with the FiveThirtyEight basketball prediction machine, in the early part of the NBA playoffs in April, the Boston Celtics were its overwhelming pick to win the title. This week, the Celtics take the floor in the NBA Finals as Eastern Conference champions, just as numbers-buffet proprietor Nate Silver's projections said they had a 59% chance of doing. 

At the start of the playoffs, I made fun of the FiveThirtyEight numbers for being noisy and absurd — fluctuating wildly off the results of single games and illogically rating the Celtics' chances higher than those of all the other top contenders combined. Yet the Celtics did make it out of the East. Great news for the machine! Less good news: On the other side of the tipoff, representing the Western Conference, are the Golden State Warriors, whom FiveThirtyEight gave only a 22% chance of making the Finals and a 9% chance of winning compared to the Celtics' 39%. 

Unshaken by Golden State's success (or by the fact the Celtics had to survive two game sevens to escape the East), the computer is affirming that ratio even now that other teams have fallen away: FiveThirtyEight currently gives the Celtics an 80% chance of winning the Finals to the Warriors' 20%. 

This does not mean that Silver's computer model got here by being steady or consistent. FiveThirtyEight boasts that its "forecasts are based on 50,000 simulations of the rest of the season" and that the system ignores teams' wins and losses in favor of "our NBA player projections, which estimate each player's future performance based on the trajectory of similar NBA players." 

Yet that big data set, rather than generating imperturbable views of each team's essential quality, somehow leaves the machine more fickle than the loudest TV take artist, ready to dump all 50,000 simulations' worth of its previously held opinions as soon as it sees who wins Game One. In early May, one week after FiveThirtyEight had the Celtics at 39%, it panicked and slashed Boston's title chances to 18% — half the chance it gave the Phoenix Suns. The Suns did not make it to the Western Conference Finals, let alone the Finals. 

The computer was, however, resolute in discounting the Warriors. While Golden State was breezing through its opponents by a cumulative playoff record of 12-4, FiveThirtyEight kept the Warriors' championship odds in single or low double digits. At the start of the Western Conference finals, the computer predicted the Warriors should lose to the Dallas Mavericks by 73% to 27%. The Warriors brushed off Dallas in five games by an average score of 115.2 to 106.6.

Even now, the FiveThirtyEight computer insists the Warriors don't belong. By its numeric rating system, it still considers them only the eighth-best team in the NBA this year, behind six teams that have already been sent home — including the Brooklyn Nets, who were swept out of the first round, and the Los Angeles Clippers, who couldn't get past the play-in games. 

As of today, Nate Silver's model likes the Celtics to beat the Warriors even more strongly than it liked the Mavericks to. Just like when he ducked the challenge to defend his numbers back in April, of course, Silver doesn't really believe this; he won't bet $400 on the Celtics against $100 on the Warriors. The DraftKings sportsbook has the Warriors as the solid favorites, paying $162.50 on a $100 bet, while $100 on the Celtics would pay $235. I have a friend in Vegas standing by to hedge my bet if Silver is willing to defend his machine, so I'd make money either way. 

Tom Scocca is the editor of Indignity, a twice-weekly newsletter of essays, reporting, and commentary.