This is a phenomenon which I find interesting, and which I feel I've been learning the hard way. I went into my first two seasons with the same thought that because the two goalies accounted for many cats while 10 skaters accounted for only a few more that emphasizing goalies was the key to success. I now agree more with DerekReese and Penguin that this isn't true.

The problem is that there seems to be more variance in the goalie stats, especially stats like W, GAA, SV%, and SO. I just tried to go into my league's stats from last season but apparently CBS does not have old weekly stats available. In any case, greater variance is what you'd expect if a stat is coming from far fewer players. All it takes is one of your two goalies to get shelled and two out of three cats (in my league's case) are gone regardless if your two goalies are superior to the other team's. With a bigger sample size the better goalies will likely come out ahead, but head to head isn't so much about the long term. A top goalie in save percentage will probably have fewer shellings than a meiocre one, but part of their elite status usually comes from the fact that the good weeks are much better than the good weeks of the weaker tender. But beating some teams by 50% while losing to others in other weeks by a few percent doesn't make a difference in H2H: a cat is a cat.

This is compounded by the fact that much of the season depends on the playoffs. You only need to do well enough to make the playoffs, and then you need some luck. Cats with more variance will depend more on that luck, so having stronger offensive cats would give you a better shot at #1 than being weaker at offense but stronger in net. Strength in net might give you a better shot at being in the top half of your league (but see above), but the difference between top three or four and winning it all is less likely to be due to strength in goal.

Another factor that has converted me to this way of thinking has been that goaltending is more amenable to grabbing a goalie off the wire (or having one from a late pick available) to ride a hot streak. One week of strong stats from a third tier goalie (and all that takes is an injury to the starter, a favorable schedule, a surge of confidence) can make him valuable, and you only need him to be valuable for a week. This is especially true with ratio cats, which can be fantastic even with only one or two starts.

I know many will object to this strategy, and I know there are prima facie arguments against it. However, after my (albeit limited) experience of the last few years, I will need convincing evidence to pull me back to the goalie-focus camp.

P.S. I think this would be a much stronger post with some data for me to lean on, and if anyone knows where I can get such I'll be happy to try to improve it.

__________________
H2H 10 GMs Keeper: keep up to 4 at cost of draft pick (round in red) G, A,+/-,PIM,SOG,PPP; Wins,GAA,Sv% Start 2C 2R 2L 1 Util 4D 2G, weekly starts
C: Tomáš Plecanec (10), Jonathan Huberdeau (11)
LW: Patrik Eliáš (14), Pascal Dupuis (10) Иlлья Kovalchuk (1), Brandon Saad (10)
RW: Jarome Iginla (3), Phil Kessel (2) Jakub Voráček (9)
Util:
D:Ryan Suter (7), Mike Green (4), Brent Burns (10), Mark Streit, Justin Schultz (12)
G:Craig Anderson (6), Дeвaн Dubnyk (17) Cepгeй Bobrovsky (9), Niklas Bäckström (8)