How I grade every NHL draft pick

What the numbers on my draft page actually mean

July 19, 2026 · Keith Wilcox

There's a draft page on this site that gives every pick since 2000 a score from −1 to +1. Late round guy who turned into a star, that's near +1. Top pick who never played, that's near −1. Zero means the pick gave you what that slot normally gives you.

This is me explaining how that number gets made, because a score nobody can explain isn't worth much. Four pieces:

Turning a career into one number
Figuring out what each draft slot is normally worth
Comparing the two
Handling guys who are still early in their careers

Everything below runs on my own database: every draft pick from 2000 to 2026, and every NHL season those players have played.

Step 1: what's a career worth?

A career is just seasons added up, so I needed a number for one season:

season value = 0.4 × (games played / most games anyone played that season)
             + 0.6 × (points / 90th percentile points at your position that season)

The two denominators are doing the real work.

Using the most games anyone played instead of 82 means the short seasons handle themselves. The 48 game lockout year and the COVID years don't need special treatment.

Using the 90th percentile at your position, that season means a defenseman gets compared to defensemen and 2003 gets compared to 2003. One denominator handles position and era at the same time. Roughly: a first line season is about 1.0, an average NHL season is 0.5 to 0.6, a ten game call-up is 0.05.

Here's one famous season run through it, Daniel Sedin's Art Ross year.

seasonid k gamesplayed points v 20092010 3 82 41 0.79 20102011 4 69 56 0.87 20112012 5 71 63 0.95 20122013 6 41 33 0.89 20132014 7 81 79 1.18 20142015 8 82 87 1.23 20152016 9 82 89 1.30 20162017 10 77 69 1.09 20172018 11 82 79 1.13 20182019 12 78 53 0.81 my A = 10.2349 published actual_value = 10.2349 MATCH - you just rebuilt the value metric

Step 2: what is a draft slot normally worth?

10.23 means nothing until you know what pick 129 usually returns. That's the expected value curve: take every pick at each slot across the finished draft classes, average what they actually produced, smooth it out.

Two choices in there I'd defend if someone asked.

Average, not median. Most picks after about 60 never play an NHL game, so the median outcome back there is exactly zero. Use the median and every late round guy who made it looks like the same perfect steal. The average keeps the lottery ticket value in there, which is what you're actually buying.

Step 3: the score

score = (actual − expected) / (actual + expected + 1)

Top part is the surplus. Bottom part squeezes it into −1 to +1 and, more importantly, that +1 fixes an edge case. A seventh rounder who never plays has an actual of 0 against an expected of about 0.2, so he scores about −0.17. Barely negative, which is right. You didn't lose much, because that pick wasn't promising much. Without the +1 he'd score −1.00, the same as blowing a top ten pick, which would be ridiculous.

Benn: A=10.23, E(#129)=0.61 -> score = +0.8119 published: +0.8119 <- should match 2000 #205 Henrik Lundqvist A= 5.97 E= 0.30 -> +0.78 2012 # 1 Nail Yakupov A= 3.40 E=11.57 -> -0.51 2008 # 11 Kyle Beach A= 0.00 E= 4.00 -> -0.80 2010 # 1 Taylor Hall A=10.81 E=11.57 -> -0.03

Step 4: guys who aren't done yet

You can't hold a 2023 pick to a twelve year standard, so there's a separate curve for every stage of a career: what a slot has produced by year one, by year two, and so on. A pick three years in only gets compared to what his slot had produced by year three. Every summer I rerun it and everybody moves up a year.

This is also where something I'd always believed shows up in the data. Wingers arrive first, defensemen take longer, goalies take longest. I never coded that in anywhere. The curves learned it on their own, because they're fit separately by position and by year.

What this doesn't capture

Points aren't everything. A shutdown defenseman's value barely shows up here. Comparing guys to their own position helps, but it doesn't fix it. A WAR style number would.

My goalie number is crude: starts, wins, and a save percentage bonus. Fine for draft accounting, not a goalie evaluation.

The curve is one average across 2000 to 2014. Drafting has gotten better since, so late picks today are probably worth a bit more than the curve says.

No credit for development. If a team turns a fourth rounder into a player, the score gives the pick credit, not the organization. Those are different skills and I'd like to separate them eventually.