Back in December, I looked at scoring rates in tie games, and Tom Tango thought it would be interesting to see the same numbers for all goal differentials. Here goes, from the visiting team's perspective in 10-minute blocks using data from 2005-06 to 2008-09:
The table lists the goal differential per 60 minutes in each game state. I broke out the last 100 seconds of the game because of the huge differences in strategy once teams start pulling the goaltender.
As you can see, most of the most common buckets (tied, +/-1 goal) are slightly negative, which is the essence of home-ice advantage. In particular, road teams are quite likely to lose a 1-goal lead when they have it. Also, the "pulled goalie" effect manifests itself quite strongly in the last few minutes: a team with a 1-goal lead is much more likely to go up 2, while a team that's down 1 is very likely to go down 2.
Here are the raw GFs and GAs by state:
The low scoring levels in the first period of tied games are quite striking - in the first 10 minutes, they're one-quarter what they are in the 2nd and 3rd period. The only other block of time that deviates that much from the mean levels is the last couple of minutes of the game - teams are very likely to get scored on when they have the goalie pulled, which I've discussed in great detail before.