Expected points is basically the average amount of points (ie. standings points) scored by teams over the last 3 years by game state.  The basic model I created takes time, goal differential, home/away, and strength (eg. 5v5 or 5v4) into account.  Thus as the game state changes because of goals etc. we can see the expected (or average) points a league average team would accrue at the current game state.  This is a very crude beta version.  The charts are constructed in excel by hand, maybe if I have time I can get an integrated model going eventually. See the last 3 games after the jump...

Vs. PIT- I love that Jamie McGinn's goal accounts for about 1.1 points.  It is by far the biggest goal of the past 3 games.

Vs. NSH- Obviously a bit of a blown lead in the 3rd. It's interesting to see that Couture's goal to put the sharks up 3-1 is about as valuable as that 15min of play in the 3rd. Also, that goal scored by Suter was worth way more EP to NSH than SJS lost.  This is the effect of giving 1 point  for losing in OT.  A GA late in the 3rd isn't as costly to the leading team as it is valuable to the trailing team.

Vs LAK-  This game had a lot of penalties but the interesting thing here is that the late penalties in the 3rd, especially those taken by LA had virtually no impact on expected points.  Thus team's that are behind, with the game out of reach can whack away at the winning team without any real consequence (other than a fine I suppose).

The model is based on Alan Ryder's win probability paper, fitted to game data from the last 3 years.  The idea of win probability is based on various models such as Brian Burke's Advanced NFL Stats, and FanGraphs. As always you can contact me for more details.