Back-to-back games in NHL - Is rest day overrated?

It's often assumed that teams struggle when playing without a rest day between games. But is that the case? If rest truly affects performance, how much of an impact does it have, and which teams are most suffer from them?

So, how big portion of games are back-to-back games?

Since 1990-1991 season, teams in average have played 17.8% of games without rest day since previous game, which is 14.6 games in average per season.  When we look at the data from each season, we can see that the number of back-to-back games has been decreasing in the recent years though - in the past five seasons, the percentage of back-to-back games is 15.5%.

Quick clarification - The numbers include only second game of back-to-back games. So the first game is not included.

After looking at the amount of back-to-back games, I was wondering it is common that two teams play against each other in consecutive days. Looks like it was quite common in the early 90's, but since that not very common. Expect in 2020-21, which was a special season because of the pandemic.

Which teams benefit from the scheduled back-to-back games?

So who suffer or benefit the most from back-to-back games? In this section, I have ignored games where only other team is playing back-to-back games. It looks like that eastern conference teams have slightly more games without rest day against opponent with rest days.

So in average, teams in east play 11.36% of games (9.31 games in 82 game season) without a single rest day, where teams in west play "only" 10.53% (8.63 games).

Well, what teams benefit from playing against team that had game previous night? Before even looking at the data, I would guess that California teams will be on top of the list.

  

No surprises here - Anaheim Ducks and Los Angeles Kings are both in top 3.

Also the difference between conferences is quite small in this comparison. In average, teams in east play 10.39% of games (8.51 games in a season), and teams in west play 11.55% (9.47 games).

So who gets the best benefit of the back-to-back games? We can calculate the difference between games played against teams with back-to-back games and back-to-back games played. So the bigger the number is, the better:

Sorry Philly fans, schedule is not really in your favor. You play a lot of back-to-back games, but don't get the best benefit of them - playing against team that are playing on consecutive nights while. But we haven't yet seen if the back-to-back games affect performance at all, so maybe the bad schedule don't matter at all.

Does lack of rest day affect affect performance?

First thing to look at, is the winning percentages. With big number of games in scope, winning percentage should be 50%, or at least very close to it.

I selected all games where a team had played the day before, while the opponent had at least one rest day since 1990-1991, and I found 9467 games. That is a fairly big number of games. So what does the winning percentage look for a team without a rest day? 

 

Ouch, not very good. Winning percentage is only 42.36%. 

I was wondering if the number is bigger nowadays, with better recovery methods/habits and so on. If that was the case, we would see it when comparing decades.
 

 

Seems like that winning percentage has been pretty much the same since 90's.

But we saw earlier that about 76% of latter games of back-to-back games are road games, which probably also affects the winning percentage.

The difference in winning percentage is 5% in road games and 5.5% in home games. So based on this, I think we can make a conclusion that lack of rest day affects winning probabilities by ~5%.

Quite often backup goalies play the second game of back-to-back games. So maybe that is also effecting the performance? Let's see.

Goalie strategies in back-to-back games

I didn't find any good list of number 1 goalies of teams in the past, and while that can be done manually by going through each team year by year to get precise results, I cut corners here and automate it. 

For each team, and for each season, I defined primary goalie based on who goalie played most games for the team in the season, and others were defined as backup goalies. That means that in some seasons, starting goalie is defined as backup goalie in this data, if he was injured and therefore played less games than whoever is considered as backup goalie.

Most likely that does not affect the results significantly, but it does have a minor impact.

First it is good to verify, if the hypothesis of "Starter plays first game and backup goalie the second game of back-to-back games" is true.

So its not just me, it really is the most common goalie strategy for back-to-back games. 

I think it is worth mentioning here that in this data, "Backup playing both games" does not mean that the same backup goalie played both games. It can also mean that 2nd and 3rd goalies were playing, if starter was injured.

Also, when we compare the goalie strategies in each decade, it seems that starter playing both games has become less common. in 2000-2009, in 35.05% of back-to-back games, starter goalie played both games. That is quite high number, considering that in 2020's, only in 6.55% of back-to-back games starter plays both games

 

So what about the winning percentages in second game of back-to-back games? We already know that team with no rest days have won only 40% of games, but can we avoid that by not playing backup goalie, if possible? Well, according to data, looks like that it has minor effect to the probabilities, but still the winning percentage is far from 50%. By playing starter in both games instead of backup goalie in second game, there is ~2% increase in winning percentage.

 

So backup goalie is not only one to blame for bad success in second game, although it has some impact. 

Does physical first game affect the second game?

When I think about other factors that might affect the team performance in second game, one of the first things that come into my mind is physicality of first game. 

I use game log data as data source, and NHL provides that data only since season 2009-10. So that reduces the games in the scope a bit.

I calculated winning percentages of second half of back-to-back games by number of hits (both teams) in first game.

Seems like that more hits in first game might lead to lower chance of winning in second game. However, games with < 25 or >= 75 hits are quite uncommon; Only 175 games where total hits of previous game was under 25, and 124 games with 75 or more hits.

In lower volumes, random chance has bigger impact on the results. Category ">= 75 hits" has only 124 games. What if most of the teams in question happen to be underdogs, and therefore have lower winning percentage by default? 

We can do some adjustments to data. We have winning percentages of teams, so we can utilize that to compare expected and actual winning percentage.

For each game, I took teams winning percentage of the season, increased it by 5% if the team was playing home, and decreased it 5% in away games.

So if Team A (home team) with 57% winning percentage in the season played against Team B with 60%, the winning probability of Team A was calculated this way:

Team A winning probability = (57 + 5) / ((57+5) + (60-5)) = 0.5299 --> 52.99%

I compared the expected wins to actual wins:

 

Based on this, it seems that amount of hits in first game has impact to the winning probabilities in second game.

Although the method to define winning probabilities (and therefore expected wins) is quite simplified and could be improved, I'd still say it is way better than just using the actual winning percentages.

Conclusions

Clearly lack of rest day between games has impact on performance.

The schedule is favoring some teams more than other, but the impact in big picture is not remarkable. If we take two extremes Ducks and Flyers, statistically Ducks benefits ~0.5 wins per season from the scheduled back-to-back games.

But that does not necessarily mean that game schedule favors Ducks more than Flyers. There are probably other things in schedule that affect performance, such as length of road trips, traveled distance, etc.