Don't Draft What You Can Buy at the Dollar Store

Earlier this week I was reading an excellent piece by the Falconer about where star players are drafted. I think that his work offers pretty compelling evidence that the "Best Player Available" approach may not be the best way to go. Neither is drafting for need. In his analysis, teams tended to select almost all of the best forwards in the first round. High-end defenders, on the other hand, tended to be available more often in rounds two through seven. In my view that's really the entire point of drafting. You need to pick up pieces in the draft that you won't be able to find cheaply on the free agent market.

The Falconer's study, however, was limited to only one season's worth of data and (by design) looked only at teams that made the playoffs. As such, I thought that I would do a similar study that takes more years into account and uses slightly different criteria. In my view, the NHL version of the Dollar Store is anything you can pick up for 2% of the cap or less on the UFA market (with a 55M cap that works out to 1.1M). We're basically talking about bottom six forwards, bottom pairing defenders and back-up goaltenders or marginal starters. Drafting a player that fills one of these roles is better than drafting a player that completely busts but not nearly as good as drafting a top end player. As such, I thought I'd take a look at the drafts from 1994 to 2003 and determine where the top end talent was drafted. Since the data we have is rather limited I decided to call a successful pick any forward that puts up at least o.50 points per game in at least 200 NHL games. For defenders, it becomes a bit more complicated to measure since defensive defenders often play in the top four and don't put up as many points as the Marc-Andre Bergeron's of the world. Still, defensive defenders tend not to be as expensive and I would like to use criteria other than just games played. As such, I decided that defenders should play at least 200 NHL games and put up at least 0.20 points per game. For goaltenders, I considered them worth drafting if they played at least 200 NHL games and had a career save percentage of .910 or better. Of course, the story for many of these players is still being written (especially the 2001-2003 draft classes) but I thought it worthwhile to use data close to the present draft in order to gauge more current trends. Further, the time that these players present a bargain price is often on their first and second contracts and all of the players from 2003 should now be on their second contract. This analysis was helped along by the use of the database created by Daoust which you can find a link to and download yourself in this post. Without further ado, here is a chart that represents where the top players were taken in the draft:

From what I can tell there are a few acceptable draft strategies given this assessment. First, the goalies, and this isn't negotiable. Of the four goalies taken in the first round that have worked out so far, two were taken in the first four picks. Some might argue that goalies are hurt most by including the drafts from 2000-2003. That's true. There are some goaltenders who could still qualify and bring the total number of good drafts up. Goaltenders like Mike Smith, Josh Harding and Marc-Andre Fleury (qualified in games played but not save percentage). But that's part of the problem with taking a goalie. Fleury was another top four pick and so he may indeed work out. Smith? Traded. Harding? Probably traded. Sure they might bring back something of value but they haven't contributed much value to their clubs in the mean time. Furthermore, the fact that goalies were hurt by the recent drafts the most doesn't explain why only 25% of the best goalies drafted from 1994-2003 were drafted in the top 100 picks. Fully 75% of the best goalies were still available after that! That's crazy! It also makes it look like a very good bet to take a goalie late in the draft. The chance of success is actually higher between 101 and 220 then it is for forwards and defencemen. Did you have a pick between 5 and 100? Did you take a goalie? If so, that was really (really) stupid. Moving on.

From here there are two possible draft strategies that make sense to me. The first is having a bias toward forwards in the first 100 picks. The reason for this is that, of the best forwards in the draft, 78% were taken in the first 100 picks compared to only 63% of the top defenders. Plus, you have a much better chance of getting a top defender later in the draft than you do a top forward (it's almost twice as likely). The problem? You have a better chance of getting a top defender than a top forward almost any time in the draft. As such, you have a better chance of drafting more top notch players by taking more defencemen than forwards. What to do?
.

Obviously a team needs to procure both forwards and defencemen. As such, I think that I would take whichever player I thought was the best available in the first round of the draft without regard for position (unless it's a goalie, I would never never never draft a goalie in the first round). The fact that I have a better chance at being right on a defenceman is balanced out by the fact that there will be more good defencemen later on than there will forwards. With my other picks in the top 100 I would have a strong bias towards drafting forwards. This is the only point in the draft where I have as strong a liklihood of drafting a good forward as I do of drafting a good defenceman. If I have two players ranked closely, one forward and one defender, I'm taking the forward. In the remaining rounds I would have a strong bias toward defenders and goaltenders since I'm almost twice as likely to get a good player by taking defenders and goalies. I would need to have a forward well clear in my rankings in order to be convinced to take him. The typical seven-pick draft would look something like this:

1st round - forward/defenceman (34.5% chance at one strong forward or 45.2% chance at one strong defenceman)

2nd and 3rd round - two forwards (13.4% chance at one strong forward)

4th through 7th round - three defencemen and a goalie (17.2% chance at one strong defenceman and 6.9% chance at one strong goalie)

This isn't drafting for need and isn't exactly drafting the "best player available" in the traditional sense either. It's using statistics to help me discover who the best player is. Teams are, apparently, not good at identifying who the best players are. They tend to overrate forwards and undderrate defencemen. By inserting these biases into the draft strategy I have a much better chance of drafting the *actual* best player available.