Skip to main content
Menu
The Warriors’ Improbable Comeback

As the Oklahoma City Thunder learned Monday night, it’s really tough to eliminate a defending NBA champion. After opening up a 3-1 series lead in the Western Conference finals, the Thunder were poised to knock the Golden State Warriors out for good — they led for almost the entirety of Game 6 before collapsing down the stretch, and they took a 6-point lead into the second half of Game 7. None of that mattered; the finishing blow was always just outside of OKC’s reach.

But, man, did the Thunder come close. In fact, according to our quarter-by-quarter analysis of seven-game playoff series since 1984,1 the Thunder — with a peak win probability of 95.8 percent going into the fourth quarter of Game 6 — came the seventh-closest to a series victory of any modern team that went on to lose. And in the conference finals, only the Portland Trail Blazers’ infamous collapse against the Los Angeles Lakers in 2000 was more unlikely.

(The least likely comeback overall? Last year’s Houston Rockets, who went down three games to one against the Clippers before storming back to win three straight games — just like this year’s Warriors team.)

LOWEST WIN PROBABILITY MOMENT
YEAR TEAM OPPONENT ROUND GAME QTR ODDS
1 2015 Rockets Clippers 2 6 3 0.3%
2 2000 Knicks Heat 2 6 2 1.8
3 2000 Lakers Trail Blazers 3 7 3 2.0
4 1994 Rockets Suns 2 3 1 2.4
5 1995 Rockets Suns 2 5 1 3.5
6 2013 Heat Spurs 4 6 3 4.2
7 2016 Warriors Thunder 3 6 3 4.2
8 2001 Bucks Hornets 2 6 2 4.7
9 1997 Heat Knicks 2 6 1 6.6
10 2005 Mavericks Rockets 1 3 3 7.5
Greatest best-of-seven NBA playoff series comebacks

Win probability calculated at the end of each quarter or overtime, based on FiveThirtyEight’s Elo ratings

Source: basketball-reference.com

Here’s how we crunched those numbers: First, we gathered quarter-by-quarter line score data for every NBA playoff game since ’84 and used Wayne Winston’s method of converting point margins to win probabilities to generate a win expectancy number for each team after every quarter. (Given the data we were working with, we couldn’t go deeper to, say, the play-by-play level to find the most dire circumstances within a quarter.) We then plugged those game-level probabilities into the formulae for winning a series, which allowed us to incorporate both the results of games already played and the expected outcomes of future games based on Elo.

In terms of series win probability, the Warriors rode a wild roller coaster against the Thunder. At the start, our Elo-based projections gave them a 70 percent chance to win the series. However, by the end of Game 4, that had dropped to just 17.5 percent. A solid Game 5 win helped their odds, but by Game 6 the Thunder had them on the ropes. Entering the fourth quarter of that game, the Warriors stood only a 4.3 percent chance of winning the series.

paineflowers-warriors-roller-coaster-1

But the Dubs pulled out a dramatic win in Game 6. Their reward: a Game 7 at home that had them as 66 percent favorites to win the series.2

The rest is history. The defending champions are returning to the Finals, in a rematch against LeBron and the Cavs.

Check out our NBA Finals predictions.

Footnotes

  1. The year the NBA playoffs expanded to 16 teams.

  2. It’s worth noting that the betting markets (as measured by Inpredictable’s in-game odds) were never quite as bullish on OKC’s upset chances as the statistical models were. Perhaps the markets were factoring in the invisible aura of Golden State’s championship experience?

Neil Paine is a senior sportswriter for FiveThirtyEight.

Andrew Flowers writes about economics and sports for FiveThirtyEight.

Comments