In back-to-backs, maximize rivalry points

Lundqvist has been out since Jan. 31 with a vascular injury in his neck. (Jae C. Hong/CP)

A little over a week ago, Alain Vigneault’s New York Rangers played games on consecutive nights—the first against the Toronto Maple Leafs and the second versus the Edmonton Oilers. Both opponents were fresh coming into the contests, so the Rangers and Leafs were on even footing, while the Oilers had the advantage of rest on New York. That matters, because rested teams facing opponents that played the night before have had a distinct advantage over the course of modern NHL history. New York’s coach, however, had a gamble to try and even the odds.

Vigneault opted to dress backup goaltender Cam Talbot on the road against Toronto, doubtless thinking that, with a rested, team he could afford to make a sacrifice in net. That decision allowed him to play a rested Henrik Lundqvist against the Oilers in New York, where the benefit of home ice advantage and a franchise goalie would give his Rangers stood a decent chance of beating a weak Edmonton squad, particularly with that club down Taylor Hall. Rather than try all-out for the win over Toronto, Vigneault decided to give his team a shot at both games (Why didn’t the Rangers’ coach simply start Henrik Lundqvist in both games? Because goalie performance dips dramatically when a player dresses for consecutive contests).

Unfortunately for New York, Vigneault’s gamble backfired spectacularly. The Rangers weren’t brilliant against Toronto, but if not for Cam Talbot’s lousy .839 save-percentage performance they probably would have got away with the win; as it was they fell 5-4. As expected, Lundqvist was excellent against the Oilers the next night, but it didn’t matter; despite their goalie turning aside 30 of 32 shots the Rangers lost because their skaters couldn’t compete with Edmonton. Had the goalies’ appearances been switched, it’s likely New York would have gone 1-1 rather than 0-2 on the weekend.

That doesn’t definitively mean that Vigneault made the wrong choice, though. While the gamble didn’t work, it’s possible that over time it would. So it’s worth asking which strategy works better for NHL teams. Does it make more sense fto start a No. 1 goalie in the first of back-to-back games to maximize the possibility of winning that contest? Or is a team better off starting its backup in the first contest in the hopes of winning both games?

What do most NHL coaches do? Eleven teams last season had a No. 1 goaltender who started 70 percent or more of their games: Pittsburgh, San Jose, Dallas, Tampa Bay, New York, Arizona, Colorado, Philadelphia, Montreal, Boston and Columbus (interestingly, all but one of those teams qualified for the post-season, with the Coyotes missing by only two points). In the aggregate, how did the coaches of those teams deploy their goalies in back-to-backs?

  • In 41 of 95 sets of games (43 percent) the coach followed Vigneault’s strategy, starting his backup goalie in the first game and his starter in the second contest
  • In 54 of 95 sets of games (57 percent) the coach did the opposite, starting the team’s No. 1 goalie in the first game and playing the backup in the second
  • Quality of competition mattered, but not as much as might be expected. Of the 190 back-to-back games in this sample, the starting goalie faced a team that ultimately made the post-season on 53 occasions while the backup faced playoff teams 49 times

Back-to-back situations matter a lot for teams with a workhorse starter because (at least last year) they accounted for better than half of all non-injury related games started by the backup. It’s worth taking the time to examine which strategy works best for coaches.

In 2013-14 most coaches chose to maximize their chances of winning that first game by dressing their starter, but the teams that followed Vigneault’s approach actually did better last season, going 52-30 (63.4 win percentage). Teams that started their No. 1 goalie in the first contest went 58-50 (53.7 win percentage). But that’s a figure badly skewed by sample size, as we can see when we look at the save percentages of the goaltenders in each contest:

  • Starter, first half of back-to-back: .915 save percentage
  • Starter, second half of back-to-back: .941 save percentage
  • Backup, first half of back-to-back: .912 save percentage
  • Backup, second half of back-to-back: .917 save percentage

It’s possible that teams in the second half of a back-to-back play a more defensive game, driving up save percentages, but I’d need to see a much longer set of data before I put stock in that. We’re only looking at less than 1,300 shots here, so I’d suggest it’s an aberration; it seems ludicrous to believe that the difference between being an average NHL starter and being Dominik Hasek is playing the first or second game of a back-to-back.

With that crazy save-percentage spike skewing the numbers, we can’t rely on last year’s records to indicate whether it makes more sense to start the backup or the starter in Game 1. There is one (slight) advantage to playing the starter in the second game; because teams tend to allow a few more shots in the second half of a back-to-back (generally, between one or two more than they do in the first half) it makes sense to play the better goalie in the contest where he’ll see the most shots. As a practical matter, however, we’re talking about such a small difference that over a period of time as short as an NHL season it’s unlikely to make much difference.

Unfortunately, it’s impossible to develop a blanket strategy that makes sense in all situations. Tired teams playing against rested teams see their chance of collecting points fall by six percent and their odds of winning the game fall by seven, but if the team in question is sufficiently excellent (or the opponent sufficiently awful) it may be worth maximizing the chances of winning that second game. Alternatively, a team that desperately needs to make up a gap in the standings would be well advised to focus on winning that second game, while a team with a comfortable lead on a playoff spot might need only to make sure it gets a win in one of the two contests.

A good rule of thumb would seem to be focusing on the game that matters the most to the team. That suggests Vigneault’s decision was mistaken. Points collected by Edmonton, a Western Conference team, have no impact on the Rangers’ playoff hopes, while Toronto isn’t a division rival, but could plausibly be in contention with New York for one of the wild card berths in the East. It would have made more sense to maximize the chances of winning the contest against a potential playoff threat than it did a team that doesn’t really matter for the Rangers.