I originally posted this in comments to the thread about starting Boucher on Tuesday, but then I realized that you east coasters are all asleep, and since I did all that work, I wanted someone to see it. (Though I notice the comments sections on fan posts are pretty slim -- do you guys look here?)
There have been a bunch of reasons floated for playing the backup, and I've looked into some stats on a couple.
1) The starter won't be as effective on the second day of a back-to-back anyway, so you might as well rest him.
I took every goalie who's started at least 20 games this year and sorted the games by how much rest they had. It included 159 games played on back-to-backs and 1552 on more rest. I was worried about it skewing, either to the good side because only the best/hottest/fittest goaltenders are allowed to go twice in a row or to the bad side because the team in front of him is tired too and makes him look worse. But here are the numbers:
On back-to-backs, goalies face 28.6 shots per game, saving 91.8% and allowing 2.50 goals.
With rest, goalies face 28.3 shots per game, saving 91.2% and allowing 2.61 goals.
I don't see any obvious signs of fatigue, either in the goalies or in their teammates.
2) The starter will get worn out down the stretch if they play too much earlier in the year.
I took stats from the last two years. In one bin, I put the 10 goalies who had 70+ starts. In the other bin are the 10 goalies who had 60-65 starts (enough that they're a clear #1, but low enough that they clearly got more rest).
The 70+ start guys:
In their final 10 games (100 total games, 10 from each goalie), they averaged allowing 2.42 goals per game, saving 91.1% of shots.
In the rest of their games, they averaged allowing 2.47 goals per game, saving 91.1% of shots.
The 60-65 start guys:
In their last 10 games, they averaged allowing 2.86 goals per game, saving 90.1% of shots.
In the rest of their games, they averaged allowing 2.54 goals per game, saving 91.2% of shots.
Again, no compelling evidence that the guys with lots of starts got worn out.
3) The backup needs playing time or he won't be ready or effective when called on.
Unfortunately, I'm tired of working on this right now, so for now I'll have to settle for questioning the need to rest the starter. But the need to work the backup seems statistically testable too...