Jump to content
Sign in to follow this  
RazorStar

Predicting Series Winners Based on Game Results (NHL)

Recommended Posts

Hey guys, I wanted to sate my curiosity about something for a while, and I finally decided to buckle down and get the data together to find out. I wanted to see what the impact of winning a single game in a series would mean for a team's chances of winning the whole thing. I decided to look at the NHL from the 2000 to the 2017 playoffs to get my sample size of 255 series, and I would work from there. Naturally, a team winning game 7 will win the series 100% of the time, so I just ignored that for the most part. What I'm interested in is seeing what games 1 through 6 mean for a team.

 

My hypothesis was that winning game 1 would lead to a lot of series wins, dropping gradually in games 2, 3, and 4, until slowly rising with games 5, and becoming even more likely with a game 6 win (since you either win a series or force it to 7 with momentum.) The results were a little more interesting than that.

 

Types of Series:

4 Game Sweeps: 29/255 (11.4%)

5 Game Series: 65/255 (25.5%)

6 Game Series: 85/255 (33.3%)

7 Game Series 76/255 (29.8%)

 

Sweeps aren't actually common in the NHL, usually only happening once or twice a playoff season. 6 game series happen with a bit more frequency than any other type, which is about what you would expect on average.

 

Overall Series Winning Percentage by Game Winner:

Game 1 Winner: 173/255 (67.84%)

Game 2 Winner: 178/255 (69.80%)

Game 3 Winner: 152/255 (59.61%)

Game 4 Winner: 163/255 (63.29%)

Game 5 Winner: 156/226 (69.03%)

Game 6 Winner: 122/161 (75.78%)

Game 7 Winner: 76/76 (100.00%)

 

So this result didn't quite meet my expectation, namely I didn't think the winner of game 2 would be a greater predictor of series success than the winner of game 1. It's entirely possible that it's just a quirk of small sample size, and that the results would normalize themselves if I had 100x the amount of series to analyze, but it's still interesting to note. I should have expected Game 4 to be a little better than game 3 because of series sweeps adding to the tally. Still, it's interesting that the winner of game 3 has a less than 60% chance of winning the series. But anyway, we can further break this down by the amount of games in a series, so let's look at that.

 

5 Game Series Winning Percentage by Game Winner:

Game 1 Winner: 55/65 (84.62%)

Game 2 Winner: 51/65 (78.46%)

Game 3 Winner: 42/65 (64.62%)

Game 4 Winner 47/65 (72.41%)

 

This tells us that if a team is going to play a series in 5 games, they're most likely to lose the third game of the series. This makes sense to me, because usually the home team will win their first two games, and the opponent will feel their backs against the wall with home ice advantage. If all things were equal, you'd expect each game winner to have a 75% chance of victory for this amount of games in a series, because the winning team would need to be up 3 games to 1 before winning game 5.

 

6 Game Series Winning Percentage by Game Winner:

Game 1 Winner: 59/85 (69.41%)

Game 2 Winner: 51/85 (60.00%)

Game 3 Winner: 46/85 (54.12%)

Game 4 Winner: 45/85 (52.94%)

Game 5 Winner: 54/85 (64.53%)

 

This is sort of the overall trend I'd expect in general. Winning Game 1 gives you the best indicator of winning a 6 game series, while the opposing team is likely to win games 3 and 4 (probably on their home ice) to keep serve. If all things were equal, you'd expect each game winner to have a 60% chance of victory for this amount of games in a series, because the winning team would need to be up 3 games to 2 before winning game 6.

 

7 Game Series Winning Percentage by Game Winner:

Game 1 Winner: 30/76 (39.47%)

Game 2 Winner: 47/76 (61.84%)

Game 3 Winner: 35/76 (46.05%)

Game 4 Winner: 42/76 (55.26%)

Game 5 Winner: 37/76 (48.68%)

Game 6 Winner: 37/76 (48.68%)

 

Now this really baffled me. I would have figured that the team holding serve by winning in game 1 had an advantage in a long series, but that doesn't seem to be the case at all. Winning Game 1 seems to be a significant disadvantage if the series drags on to 7 games. What's really interesting is that winning game 2 has a much higher chance of success than any other game, going well above average in terms of chances of winning the series. Game 3 is a little lower than expected, game 4 a little more, and game 5 and 6 are a virtual 50/50 with this data size. Perhaps the results normalize to a 50-50 shot if more data is gathered, but for now I have to go with what I see. If you're looking to pick the winner of a game 7 this year, your best odds are on the team that won game 2.

 

Here's the Googledoc for now, I could find out more things with this data, but for now I'm taking a break.

  • Upvote 2

Share this post


Link to post
Share on other sites

Wow, the 7 game series data is extremely interesting. I wonder how much adjustments and momentum account for the winning % due to game 2 and 4 looking to be the best indicators(?).

 

Overall, I am not too surprised. I think game 2 has a higher win percentage because splitting the first two leaves a series likely to go far or become more even (higher seed losing home ice) and a sweep pads those stats as well. It makes sense that game 2 would be more valuable.

Edited by BJORN

Share this post


Link to post
Share on other sites

So let's take a look at a series and see if we can guess the results. Just for shits and giggles, we'll look at Pittsburgh and Washington. They're tied at two games a piece, but the order of their victories indicate that one team might have a better chance. Washington lost games 1 and 4, and Pittsburgh lost games 2 and 3, so let's look at all the data points in which those game's losers have won a series, and calculate it out.

 

Pittsburgh vs. Washington

All Potential Results:
Washington wins in 6: 5/22
Pittsburgh wins in 6: 8/22
Washington wins in 7: 7/22
Pittsburgh wins in 7: 2/22
So this says the odds slightly favour Washington at this point, they've got 12 wins to Pittsburgh's 10. However the rate changes drastically depending on who wins game 5, so let's take a look at that as well.
If Washington wins game 5:
Washington wins in 6: 5/10
Washington wins in 7: 4/10
Pittsburgh wins in 7: 1/10
If Pittsburgh wins game 5:
Pittsburgh wins in 6: 8/12
Pittsburgh wins in 7: 1/12
Washington wins in 7: 3/12

 

So losing game 5 is virtually a death knell for the Penguins, as only one team has pulled out a game seven victory after losing games 2, 3, and 5. However, Washington still has a 25% of pulling off the series victory if they lose game 5, a rate you would expect if games were simply decided by coinflips. In any case, the winner of game 5 will likely decide this series, there's nothing strange about the data here.

 

Winnipeg vs. Nashville
Potential Results:
Winnipeg wins in 6: 11/43
Nashville wins in 6: 7/43
Winnipeg wins in 7: 7/43
Nashville wins in 7: 18/43
If Winnipeg wins game 5:
Winnipeg wins in 6: 11/23
Winnipeg wins in 7: 3/23
Nashville wins in 7: 9/23
If Nashville wins game 5:
Nashville wins in 6: 7/20
Nashville wins in 7: 9/20
Winnipeg wins in 7: 4/20
The Golden Knights and the Sharks face the same pattern as the Jets and the Predators respectively, so if you're a betting man, take the Sharks and the Predators to win their series, probably in the full 7 games. The Jets and the Knights odds are only 60% of winning the series even if they take 5, which is surprisingly low.
Lightning vs. Bruins
Possible Results:
Tampa wins in 5: 10/49
Tampa wins in 6: 11/49
Boston wins in 6: 8/49
Tampa wins in 7: 13/49
Boston wins in 7: 7/49
As Tampa has a series lead at the moment, the odds clearly favour them, giving them wins in 34 of 49 scenarios, or about 69.4%. Nonetheless, let's see how things change depending on who takes game 4.
If Tampa wins game 4:
Tampa wins in 5: 10/27
Tampa wins in 6: 6/27
Tampa wins in 7: 6/27
Boston wins in 7: 5/27
If Boston wins game 4:
Tampa wins in 6: 5/22
Boston wins in 6: 8/22
Tampa wins in 7: 7/22
Boston wins in 7: 2/22
As you can see, Boston winning would give the same series outcome as the Capitals/Penguins series from above, while if the Lightning win, they are pretty favoured to take the series, as you'd expect from a team with a 3-1 lead. Even if the Bruins beat the odds twice to force a game 7, it's still less than a 50/50 shot for them to win it all.
Edited by RazorStar
  • Upvote 1

Share this post


Link to post
Share on other sites

With the Ptt vs Washington series, you are forgetting one very important detail: It's the Caps. They will lose in 7 if they win game 5.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

  • Chatbox

    Load More
    You don't have permission to chat.
×