Fantasy Hockey update, plus a curious observation

Fantasy hockey expert Greg Gibeault recently posted a comment on youlikeadajuice.  (Well, not so recently actually…sorry Greg, we’ve been busy.)  He requested an update on my Gibeault-approved fantasy hockey strategy of drafting two goalies early.  Well, Greg, here’s your update:

My team sits in first place with and 85-45-20 record, fifteen points ahead of the team in second place.  Admittedly, it’s not a very competitive league; of the twelve teams, only five have checked in 2009.  I managed to overcome my disadvantage at the forward positions (my first four picks were goalies and defenders) by picking up Mikko Koivu, Jiri Hudler, Bobby Ryan and Kris Versteeg.

My head-to-head record breaks down as follows:

• In 6 skater categories (Goals, Assists, +/-, Penalty Minutes, Power-play points, shots on goal), I’m 47-33-1.
• In 4 goalie categories (Wins, Goals-Against Average, Save Percentage), I’m 36-11-11.

So in other words, the Gibeault-approved strategy of drafting two goalies early seems like a winner.

I have one other observation: there seem to be plenty of solid forwards available on the waiver wire, but no defensemen.  The highest rated defender on the season who’s available is Fedor Tyutin; Yahoo!’s rates 32 forwards higher than Tyutin.  This could be because Yahoo!’s player rater is bad, or because of the strange nature of the league I’m in.  Or it could say that defensemen are undervalued in the draft and that you need to draft four good ones to make sure you don’t get stuck.

Are goalies undervalued?

Every fantasy season I build a model in Excel to evaluate draft choices.  It’s a simple model that uses some dubious assumptions, that’s generally proven pretty useless for fantasy baseball and football, but highly effective for fantasy basketball.  Last weekend I built the model for both hockey and basketball.  I’ll discuss my basketball findings in a later post, but right now I’m more interested in the very surprising results from the hockey model.

When I ran the calculations, and built a draft board, the model said that the top 10 picks–that is, the 10 most valuable fantasy players–are all goalies.  This clashes with conventional wisdom from the fantasy community, where you generally see just a couple goalies in the top 10.  I was concerned about an error in the model, so I started thinking about what it was doing, and whether its results made any sense.

The model I use compares players to a replacement-level player at their position, a projection of the best guy available on the wavier wire.  There are 30 teams in the NHL and as best I can tell, you only want to start the primary goalie on each.  So there are potentially 30 possible starters.  In our 12-team league, each team starts 2 goalies, for a total of 24 starters.  Additionally, teams will draft backups.  This doesn’t allow for much flexibility.  If you don’t land two top goalies, you’ll be stuck playing a bad one, or a backup, or only starting one goalie.  These all seem like really bad ideas.  Category math also works in goalies’ favor.  Fantasy teams start 2 goalies and 10 skaters.  The goalies account for 4 statistical categories, the skaters for 6.  So with two top picks, you can gain a huge edge in 40% of stats. 

In my draft this weekend, I drafted Evgeni Nabokov 9th overall, the second goalie off the board.  I followed with Niklas Backstrom (no, not Nicklas Backstrom) at 33 overall and Pascal Leclaire at 112.  According to the model, all three of these guys should have gone in the first round.

I’ll reevaluate this strategy as the season plays out.  There are a couple points against it.  For one, it’s a head-to-head league and one of the goalie categories is shutouts.  Shutouts are very rare, so I suspect many matchups will end 0-0, 1-0 or 1-1, meaning it’s a bit of a crapshoot anyway.  I accounted for this by devaluing shutouts by 50% in the model.  A second downfall is it’s hard to predict which goalies are actually the top goalies.  I went entirely based on last year’s numbers, since I don’t know enough about hockey to predict anything.  This article seems to support my thinking, and argues against this objection.  Greg Gibeault: “I can’t say that I’ve grabbed a sleeper goaltender and had him actually perform in recent memory…I don’t have enough fingers to count all of the random players who I’ve grabbed at other positions who have performed well.”

Overall, I’m really skeptical about my chances in fantasy hockey, but I’m optimistic about my goalies.