A remarkable part of the Miami Heat’s run to a 3-0 Eastern Conference Finals series lead over the Boston Celtics has been how little faith ESPN Analytics’ Basketball Power Index has had in them. That included giving the Heat just a three percent chance of winning the series before it started and just a 35 percent chance of doing so after it started. And even after Miami’s dominant 128-102 win Sunday, ESPN Analytics is still giving the Celtics a chance as high as 31 percent of advancing. That would be remarkable, as the NBA is the only North American sport with best-of-seven playoff series where a team trailing 0-3 has never come back to win.
Why “as high as 31 percent”? Here’s the 31 percent number, from ESPN’s playoff schedule page (and which is in sharp contrast to the Caesars Sportsbook odds they also provide on that page):
However, the actual BPI page (which provided the previously-cited 65 percent chance) provides the Celtics with just a 25.9 percent chance of making the Finals (and a 15.7 percent chance of winning the title). So it’s unclear which is “the BPI number.” Both of those numbers are very high relative to the sportsbook odds. And some of that is because ESPN’s BPI (at least, their past public explanations of it) factors in everything from preseason expectations to the score of every regular-season game, so it’s much more a season-long metric than a “where these teams are right now” metric.
By contrast, the metrics at corporate sibling (for now) FiveThirtyEight give the Celtics an 11 percent chance of making the Finals based on their RAPTOR player metrics, and an eight percent chance based on an Elo forecast. That’s still maybe high but looks a little more reasonable. But there’s history with their projections too, as they once gave the Celtics an 83 percent chance of winning the NBA Finals ahead of the start of that series last year, and the Golden State Warriors won it in six games. Even more egregiously, they tweeted in 2020 that the eliminated New Orleans Saints had an 11 percent chance of winning the Super Bowl. And their predictions on NHL expansion haven’t looked good either. So they’re far from immune to predictions blame.
There are a couple of things to note with these models’ projections, and with the ESPN one in particular. A pick of any one team as more likely to win a series than its opponent before the series starts is defensible. And there’s merit to these models and to their consideration of a full season of sample size (and in ESPN’s case, factors beyond that). And models with a large regular-season component should absolutely be expected to pick the Celtics (57-25, second in the East) over the Heat (44-38, eighth) preseries. Unlike analysts deciding who to pick, these models aren’t making spur-of-the-moment decisions, they’re turning numerical inputs into outputs.
But it’s still possible to question the models’ underlying data and how it’s weighted. And that’s especially true when it comes to their odds in the middle of a series, where there’s certainly an argument for valuing recent head-to-head data more. That would likely produce numbers that look more reasonable. But ESPN’s models also have been producing very high numbers with low potential for uncertainty for some time, and heavily promoting them on social media and their broadcasts. And they’ve been roasted for it, including with Will Levis. And that hasn’t led to any change yet. It will be interesting to see if any change comes as a result of this Heat-Celtics series, unless the Celtics do somehow pull off a feat unprecedented in NBA history, which ESPN figures they have a more than 30 percent chance of doing.