Statistically, which game is the most important in an NHL playoff series?

I was dead wrong. I had an idea for a study and went into it with a concrete, seemingly well reasoned hypothesis. It could not have been further from the truth. To summarize: my father always had spouted the idea that the even-numbered games in a playoff series were the most important, and through years and years of watching the Stanley Cup Playoffs with him, I internalized that thinking. He convinced me that Games 2, 4, and 6 are the most important (excluding Game 7, obviously).

But I had never considered fact-checking that idea until I got on this analytics-induced kick of putting every narrative, every closely held belief (such as this one) to the test with objective, numbers-driven, non-anecdotal analysis.

It was this tweet and subsequent exchange that got me thinking about this topic again:

…and I decided, once and for all, to do some digging. I found to be a valuable resource, but that it lacked the precise information I was interested in. So, using this series-by-series breakdown, I started inputting data into spreadsheets. First, I came up with the information for this post about how the average NHL series has been getting longer over time.

That’s when the real work got started. I proceeded to enter the win/loss sequence of every best-of-seven playoff series since the dawn of the Original 6 Era (1942) through the 2014 playoffs. I wanted to determine which games in a series were the most important. I did so by determining what percentage of series the Game 1 winner went on to take; what percentage of series the Game 2 winner went on to take; and so on. Naturally, having been influenced by my dad’s theory, James Mirtle’s tweet, and hockey fan intuition, I figured the winner of Game 4 would win the series more often than the winners of most other games (again, not including Game 7, of course).

What I found might not be statistically significant. Despite 615 Stanley Cup Playoff best-of-seven series from 1942-2014, that might not be a big enough sample size to clear out the random noise or statistical variance in the data. However, it would appear that my and my dad’s theory was dead wrong.

Here’s what I found:

Games 1 and 2

  • Out of 615 best-of-seven series, the winner of Game 1 has won 422, or 68.6%.
  • The winner of Game 2 has also won exactly 422/615, 68.6%.

Games 3 and 4

  • It seemed natural that Game 3, being later in the series, would take on slightly added importance. So I was not surprised when I found the winner of Game 3 has won 433/615 best-of-seven series, for 70.4%.
  • And then we get to Game 4. Again, this data could be victim to random variance or small sample size. But the winner of Game 4 only went on to win 65.8% of series. Which is even more surprising, considering that 111 out of the 615 total series were sweeps, meaning they ended in Game 4. So, despite being spotted a 100% number on 111 series, Game 4’s winning percentage is the lowest of all seven games. In non-sweep series, the Game 4 winner has won only 294/504 times, or 58.3% of the time.

Games 5 and 6

  • Obviously, there hasn’t been a Game 5 in all 615 series—there have only been 504 Game 5s since 1942. The winner of those games has a 72.8% winning percentage in the series. However, when a win in Game 5 did not clinch the series (458 times), the winner of Game 5 only holds a 45.8% winning percentage in the series.
  • There have been even fewer Game 6s since 1942—only 346. But the winner of Game 6 has won 77.4% of those series. And when Game 6 did not end the series, the winner of Game 6 is only 72/150, for a 48% series winning percentage.

Game 7

  • 151 out of our sample of 615 best-of-seven series from 1942-2014 have gone seven games. And this is probably the biggest shocker of them all: the winner of Game 7 has an ASTOUNDING 100% winning percentage! 151-0.
  • All kidding aside: only 24.5% of series have gone the distance. Less than one-in-four.

So, what does this information tell us? Again, it’s possible this data is subject to too much random variance and is not statistically significant. However, if I were the GM or head coach of a team with home ice advantage heading into a playoff series, and if given the option, I’d prefer to eschew the traditional Games 1, 2, 5, and 7 at home. Instead, based on the information presented here, I’d prefer to host Games 2, 3, 5, and 6.

Why? I’d rather not waste one of my four home games on Game 7, which historically only has a one-in-four chance of happening. I’d definitely want Game 3 at home over Game 1, because of its slightly higher winning percentage. And I’d avoid hosting Game 4, because it appears to be slightly less important than the other options.

Here’s one final piece of information: only 308 out of 615 series ended in either Game 5 or 7. That means the team with “home ice advantage” only played more games at home in 50% of series. Therefore, if you’re the team with “home ice advantage,” and if given the option, you might also consider hosting Games 3, 4, 5, and 6. That would give you about an 82% chance of hosting more games in the series than your opponent.

(Stick tap to my friend Elliott, who helped me sift through and understand this data).

About Trevor Kraus

Born and raised in St. Louis, Trevor is a diehard fan of all the major sports (and even the non-major ones), but particularly hockey. He plays goalie in a local hockey league and is striving to become a hockey broadcasting pioneer: the first play-by-play announcer to incorporate advanced stats into his broadcast.