It is draft day, and so I want to take one more chance to discuss Draft Theory while expanding the discussion I started with my piece earlier on Logan Stanley.
So, here is a few more words on what I said before and what you can learn from both history and numbers.
Draft Theory and why statistics matter
In essence, draft theory is really just a combination of understanding simple probability and projecting prospects. The better a player is in junior or college, the more likely they make the NHL and the higher their potential ceiling.
Our friends at Canucks Army created this visual to represent this (that I have edited for example simplicity, but read the article for some great background information on Draft Theory):
The visual is a drastic exaggeration and over simplification (ex: having no bust probability for the players and a normal distribution), but it does help get the basic idea across. Each curve is a different NHL prospect and the scale below shows all the potential possibilities for each player. This could be something that extends from a prospect model like PCS or pGPS.
The non-overlapping areas represent the likelihood that prospect from the right-side curve ends up better than the prospect on the left side. The overlapping area represents the likelihood the opposite. The left curve being less flat represents more certainty on what the player will develop into.
Probability is not destiny. There is a chance that either prospect ends up the best.
Scouting is like a game. You have multiple doors and only one has the best prize that you covet. Each door has a different percent chance of having the prize but you don’t know the odds. You can investigate and figure out hints to what the odds are like. You may choose the right door but for the wrong reasons, and vice versa.
A scout’s job is to maximize a team’s ability to figure out which player gives you the best chance at getting the best player. The scout achieves this by looking at factors like skating, size, strength, vision, work ethic, and defensive play. They then rate and compare these factors to the player’s peers in gauging which players are the best.
Statistics ignore the inputs but look instead at the outputs: performance. Draft analytics then use the player’s statistics and compares them to historically similar performances in gauging which players are the best.
The two are actually very similar and the end goal is the same: acquire information and try to predict who is best.
There are differences though; differences that can allow them to work together and better each other. Statistics shows a player’s performance to estimate how effective a player is, while scouting shows you a player’s make to estimate why they are as effective as they are.
The other benefit to statistics though is in testing. In my Stanley article I showed this graphic:
The first two bars are defensemen drafted between picks #1-25, with the left being those who scored over 0.6 points per game in their draft eligible season and the right being those who scored under 0.6 points per game. The next two bars are 26-50, and the final two are the remainder.
This is a very powerful visual. The first two bars are players that scouts view as consisting of equally rated prospects, but there is a group of players who are better and they just so happen to be statistically better performers too. However, the “Mid scorer” bar is one that statistics would rank as greater than the “Top non-scorer bar”, but statistics are wrong and the two groups succeed about equally as often.
As I noted previously, this tells you that scoring is very important and has been undervalued in the draft by scouts. What it also tells you though is that scouts are able to genuinely evaluate and project which players are better than others in the non-scoring parts of the game.
Logan Stanley is still a NHL prospect
Reading my last article on Stanley could make someone think that I view Stanley as a non-NHL prospect. This is not true. I’m simply pointing out that players like Stanley drafted at pick number 22 end up not being as good of a bet as others that will most likely be available then.
Here is that same graphic again with a few additions:
If the Jets select Stanley at 22nd or 36th overall, or another team does in that general area, this would place Stanley as a talent somewhere between “Top non-scorer” block and “Mid non-scorer” block. On average, a non-scoring defensemen selected around this area makes a NHL impact somewhere between 20-40 per cent of the time. Scoring defensemen selected around the same area make an impact somewhere between 40-75 per cent of the time.
And here’s an interesting caveat to add to all of this: most of the “non-scorers” who are “successes” and became impact players were those that eventually did become “scorers” prior to moving up to the next level. Whether their draft season performance did not match their true skill level (ex: injuries, mononucleosis, shooting percentages, buried on depth charts) or late development, the bulk of the success “non-scorers” had the skill to catch up to the “scorers”.
Shutdown defenders in junior or college almost never become capable shutdown defenders in the NHL with positive impacts on Corsi or goal differentials. In fact, they are almost never NHL regulars. Even players like Dan Hamhuis, Marc-Eduoard Vlassic, Willie Mitchell, and Adam Foote scored at respectable paces.
Back to Stanley, there are those non-scoring factors that scouts like about him. While his skating is a weakness, he did improve upon it immensely since the year prior which is a promising sign. He can skate out the puck and make decent passes. His decision making abilities seems to be solid behind his own blue line with and without the puck, although it disappears as soon as he’s outside of the defensive zone.
He has a legitimate case to why he might be better than the average “non-scorer” and better than his scoring numbers. He has a case to why he might improve in scoring in the future.
The issue is that all those who were in the “Top non-scorer” bar, yet they still on average were overvalued.
The issue is that Stanley could double his point production and would still not pass the somewhat arbitrary line used to separate “scorers” from “non-scorers” in the study.
The issue is not what Stanley could potentially become, but how likely he is to become what his top potential is versus the alternative options that will likely be available.
Stanley is most likely better than those with similar scoring numbers, like J.D. Greenway, Nicolas Mattinen, or Colby Sissons. Stanley is most likely not better than most, if not all, of Jake Bean, Dante Fabbro, Samuel Girard, Frederic Allard, Cam Dineen or Luke Green.
How numbers would have helped the Jets
Long ago I did a “statistical” redraft for the Jets, looking at the best statistical performer (using an ancient version of Florida’s now more modern PCS) within 20 picks after the player the Jets took.
While the Jets sometimes did better than the numbers in the odd pick, the exercise showed clearly how much numbers can improve an organization.
Here are the results of what blind numbers from a simple model would take, without much input from scouting to improve upon it:
Please excuse any misspellings in the list due to quick typing.
- Adam Lowry, who suffered from mono that season which impacted his PCS (his PCS% through junior was 15.2, 5.9 (mono), 12.0 (broken wrist), and 18.7). Blind numbers take Michael Paliotta (17.2 PCS%) who put up 23 points in 68 games in his AHL debut last season. He already has 2 NHL games and one point under his belt as well. Lowry may be the better player but Paliotta already looks very promising as well. Still, win goes to Chevy. 1-0
- Numbers get a forward with 102 NHL games instead of a defender who could not even earn a Entry Level Contract. Shore has been a dominant scorer in the AHL. In the NHL his scoring still lags, but he’s put up very solid shot metrics. Huge win for numbers. 1-1
- Both numbers and Jets missed on Franki Corrado, who is the best player taken right after. Numbers slip up on this one a bit worse with Eddie Wittchow over Austen Brassard, but neither do well. This really shouldn’t be a win for the Jets, but whatever. 2-1
- Blind numbers miss on the next one due to Patrick Daly actually quitting hockey. Still, no one is arguing to use blind numbers. Just showing how they stand up well, so you should use scouts to make numbers not blind and get the best of both worlds. Harstad is essentially the same thing in value though… 2-1
- Numbers dodge a huge bullet in taking Jake McCabe, a young, solid left-handed defender the Jets really need, instead of Lukas Sutter. Huge win for numbers. 2-2
- Honestly, imagine if the Jets defensive core had Jake McCabe and Shayne Gostisbehere in addition to Dustin Byfuglien, Jacob Trouba, Tyler Myers, Josh Morrissey, and Toby Enstrom. Ridiculous. Scott Kosmachuk might be nice, but he wasn’t contending for the Calder trophy this season. This really should count as 2 wins. 2-3
- Numbers get a 10 point AHL defenseman instead of a 10 point AHL forward. Call it a draw. 2-3
- Nic Petan is a nice player. 2-3
- Oliver Bjorkstrand becomes a huge step up from Jimmy Lodge. While both were skilled players in junior, Bjorkstrand has been the one to transfer into pro better. 2-4
- Matt Buckles is still in University, and his numbers came from a very small sample size of cohorts (read: low confidence). It also gave Buckle’s comps a very low P/GP upside so really I’d say that numbers taking Buckles may be a stretch, and it might go for JC Lipon instead, depending if you go for the higher probability or the higher upside statistical player. Draw or win depends on what you set numbers to go for. 2-4
- Ben Harpur is actually a statistical cohort to Logan Stanley, for what it’s worth. The defender played in all three of the ECHL, AHL, and NHL last season, although his numbers suggest AHL player to me. The numbers are blind in this exercise, but if you take Copp’s last half of the season (after pulled up from fourth line) he is the statistical favourite. Chevy wins. 3-4
- Jan Kostalek was a pretty good junior player in his final year but struggled in the AHL. Will Butcher just finished a near point per game performance as a defender in the NCHC (NCAA division). It’s still too early to tell, but this could be another win for the numbers. 3-4
- Tucker Poolman is a statistical anomaly because he’s a big scorer who keeps hanging around at each level for far longer than really he needs to. He muddies the system. Eric Roy meanwhile had his numbers inflated by Ryan Pulock. Educated numbers go for Poolman but blind numbers go for Roy. 4-4
- Kichton stays the same. While he may never be a NHL player, he has been an elite AHL scoring defenseman and that is worth value to the organization and far more than one would expect from a random 7th round defender. 4-4
- Marcus Karlstrom had a really good first half of his post-draft season. Other than that, he has not really done anything. Numbers get a good ECHL player who may provide value as an AHLer, instead of someone who sadly doesn’t look fit for North American hockey. 4-5
Blind numbers big misses are Copp, Poolman, and Lowry, but all three of them would likely be selected by a team that supplements their numbers with scouting. Copp would get picked up due to his scoring after being elevated from fourth line. Poolman’s numbers wouldn’t be used for a player who is in an abnormal situation. And, Lowry performance reasoning for recovering from mononucleosis would be accounted for.
The only miss that even fact-checking numbers would pass on that ended up a win over numbers is Austen Brassard.
Numbers though add two NHL left-shot defenders in McCabe and Gostisbehere and two AHL scorers pushing for NHL jobs in Shore and Bjorkstrand. Imagine adding those four to the Jets current talent pool. Quenneville over Karlstrom would be an improvement, but only enough to negate the loss of Brassard.
So, yes, while the Jets have been a good drafting team, even they could use some numbers to supplement their scouting.
Numbers are important. Hockey is a game where the most goals win. Everything, from size, character, skill, and other factors, are merely a means to the ends.
In drafting you want to get the best possible players available. The players that will help your organization build the best possible team, especially with cost controlled contracts. You want to use all tools available in the best possible manner to optimize your drafting. Measuring performance, scouting a player’s composition, knowing their character traits, and understanding probability are all important facets that can improve a team’s scouting.
You want as much information as possible to make the most informed decision possible.
I would draft Stanley. I would use a 7th round on him. I would use a 6th round. A 5th, a 4th, or a 3rd. I would not use a late 1st or a high 2nd on him though. If the Jets draft him at 22nd or 36th, it will not be the end of the world or the worst pick. I just doubt it will be the best pick.
He may end up being worth a 2nd. He may end up being worth a late 1st. Only time will tell.
Anything is possible, but not all is most probable.