Analyzing the Top 25 Under 25: Does potential or progress matter more?
What matters more, games played or a player’s ceiling?
Every year, we have the same debate about the Top 25 Under 25: how do you compare players with high potential to players who have actually player NHL games? On the extremes, this is easy. Everyone would put Erik Brannstrom ahead of Nick Paul, even though the latter has 54 more NHL games. But it becomes harder within certain ranges. Is Brannstrom ahead of Brady Tkachuk, even though Tkachuk has proven he can at least be a top-six NHL player and Brannstrom has proven nothing in the NHL?
Everyone evaluates these things differently. I was curious if we could find a statistical link between the results and those input factors. Obviously, this isn’t a very scientific exercise. For starters, statistical evaluation of rankings is bad because typically ranks aren’t equally spaced (the difference between 1 and 2 is not necessarily the same as the difference between 19 and 20), but analysis assumes they are. But considering this isn’t a very scientific exercise anyway, it’s probably fine. And as you'd expect, the results are inconclusive. So if you like simple concrete answers, maybe stop reading now.
First, let’s take a look at NHL experience. Here’s a player’s rank plotted as a function of their NHL games played:
It’s not a great correlation, but you can kind of see the trend towards the right as the dots get lower. In other words, playing more NHL games puts a player at a higher rank. This makes sense — good players generally get more NHL games, and more NHL games also gives us more chances to evaluate a player as fans. The biggest exception is probably players with 0 NHL games played, some of whom end up ranked highly because of their potential. The other major outlier is Andreas Englund, who didn’t even finish in the top 25 despite having 9 games of NHL experience. On the flip side, someone like Anthony Duclair probably benefited from having 287 NHL games played in earning a top-10 ranking.
The next thought is potential, and I evaluated this by looking at draft position. Here’s the relationship between draft position and ranking:
There is a general trend of higher draft position correlating to a worse rank, but again, it’s not very consistent. Instead, what I noticed was that there seemed to be a distinct region in which the correlation was high: the first round. Players drafted in the first round did well in our rankings with one notable exception — Morgan Klimchuk (28th overall, 2013). The implication here makes sense, in that players drafted in the first round are expected to do well in the NHL, and until they fail to stick in the NHL, they’ll be ranked highly. This probably helps someone like Jacob Bernard-Docker get ranked much better than Angus Crookshank, despite both having had decent rookie NCAA seasons.
The last thing I looked at was age, just to see if there was a trend towards older or younger players. Here’s the birth year vs. rank graph:
Now this, my friends, is a plot without correlation. Anyone who sees a trend in this graph is trying to sell you something. But I think the reason for this is that age plays different ways into different players. Being young is a benefit for Lassi Thomson, who gets a boost from his high draft position, but is not helpful for Maxence Guenette, who has not had any time to defy the expectations of being a 7th-round pick. In the same way, age has helped establish someone like Anthony Duclair, who has played 287 NHL games, but is a problem for Morgan Klimchuk, who at 24 has only played one NHL game. What would likely be a more useful statistic would be the age/draft position and age/NHL games played combined effect.
Even though this isn’t a great way of looking at things, I attempted to make a model that predicted the effect of these three variables, and then evaluated how close the predictions were to the final answers. The player who was closest to where he was predicted was Nick Paul, who with a moderate 56 NHL games played, a modest draft position (101st overall), and being among the oldest players on the list (born in 1995), should have appeared at 20th on the list. The biggest exception (other than the bottom of the list players) was Erik Brannstrom, who has played fewer NHL games than other players around his age like Drake Batherson and Brady Tkachuk, and doesn’t have the high draft pedigree to warrant his high position on our list. Basically, he’s beaten his draft position in the two years since he was drafted, and now we rank him really highly.
So overall, there isn’t a strong mathematical relationship here. More what we see is what we already know: experience matters, but pedigree can make up for lack of experience. Experience seems to matter more than potential, unless you were picked in the first round of the draft, in which case that matters more for the first five or so years after you’re drafted. If you’re wondering who’s gonna jump to the top of the rankings next year, look for a) whoever sees a big jump in NHL games played, or b) whoever the Sens take with a very high draft pick in June 2020.