Size Does Matter: Defencemen
By Josh W
9 years agoIn MoneyPuck’s most recent article, he analyzed the NHL success rate of CHL forwards based on their size and point production.
We know that offensive production is not the single determining factor as to whether an NHL player is good from an analytics perspective, but MoneyPuck demonstrated that point production in the CHL is a good indicator of NHL success. Similarly, a player who is taller than average
also increases their chances of success. If you combine both factors positively in a young CHL forward, their chance of success rises greatly.
We know that offensive production is not the single determining factor as to whether an NHL player is good from an analytics perspective, but MoneyPuck demonstrated that point production in the CHL is a good indicator of NHL success. Similarly, a player who is taller than average
also increases their chances of success. If you combine both factors positively in a young CHL forward, their chance of success rises greatly.
In our previous experiment, we
split forward and defencemen data, as they are two different
types of players and their success rates are different. In this article, I will continue where MoneyPuck left off
to present you the success rates for CHL defencemen, broken down by their height and CHL scoring rate.
split forward and defencemen data, as they are two different
types of players and their success rates are different. In this article, I will continue where MoneyPuck left off
to present you the success rates for CHL defencemen, broken down by their height and CHL scoring rate.
Background
It is important to start off with a refresher of the important factors in this experiment.
There are many ideas in hockey that we assume are true and are taken at
face value. People used to believe in
many ideas, now considered crazy, until science debunked them while discovering
the truth (e.g. the world used to be thought of as flat). This is the motivation for our work; we can assume that high point production (a ballpark indicator of more talented players) and bigger players (average NHL size is greater than average North American male size) have a better success rate, but we cannot be sure until we examine the data.
There are many ideas in hockey that we assume are true and are taken at
face value. People used to believe in
many ideas, now considered crazy, until science debunked them while discovering
the truth (e.g. the world used to be thought of as flat). This is the motivation for our work; we can assume that high point production (a ballpark indicator of more talented players) and bigger players (average NHL size is greater than average North American male size) have a better success rate, but we cannot be sure until we examine the data.
For players in our data, we use height as a proxy for size. We’ve evaluated player’s weight and found it has little correlation with success, and human weight is constantly in flux, so we opted to focus on height. Height rarely changes (except
when players flub their numbers) and we’ve seen correlations with height and future success and with NHL draft position, especially in MoneyPuck’s last experiment.
when players flub their numbers) and we’ve seen correlations with height and future success and with NHL draft position, especially in MoneyPuck’s last experiment.
We focused on
the CHL as it is the largest pool of NHL prospects; CHL data is readily available and the players skate in high number of games a season so point production is less likely
affected by PDO. We can assume if the logic for this experiment is true for the
CHL, it likely holds true for other development leagues (i.e. USHL/NCAA/Superelit).
the CHL as it is the largest pool of NHL prospects; CHL data is readily available and the players skate in high number of games a season so point production is less likely
affected by PDO. We can assume if the logic for this experiment is true for the
CHL, it likely holds true for other development leagues (i.e. USHL/NCAA/Superelit).
We use the
200 NHL GP benchmark as an indicator of success for a few reasons. This standard was used by others evaluating the NHL draft including TSN’s Scott Cullen who determined the likelihood of a draft pick reaching the 200 games played mark. MoneyPuck graphed the data which can be seen above.
200 NHL GP benchmark as an indicator of success for a few reasons. This standard was used by others evaluating the NHL draft including TSN’s Scott Cullen who determined the likelihood of a draft pick reaching the 200 games played mark. MoneyPuck graphed the data which can be seen above.
Drafting is difficult, and with only 8% of CHL
forwards becoming regular NHLers, a given player reaching 200 games very unlikely. It is rare that a given player plays
200, let alone 600 games, and scores 0 points without management and coaches seeing him
provide positive value to the team. Using
different values for NHLers (i.e. GVT, Corsi, Pts/GP, Raw Points etc) will likely
change the exact numbers of past “success rates,” but I doubt you will see a different
trend.
forwards becoming regular NHLers, a given player reaching 200 games very unlikely. It is rare that a given player plays
200, let alone 600 games, and scores 0 points without management and coaches seeing him
provide positive value to the team. Using
different values for NHLers (i.e. GVT, Corsi, Pts/GP, Raw Points etc) will likely
change the exact numbers of past “success rates,” but I doubt you will see a different
trend.
Data
The data
for this experiment was the defencemen in the years defined previously. We looked at over 2,400 defencemen in their
17 year-old data in the CHL who played 10 or more games in that season. All data is from EliteProspects.com.
for this experiment was the defencemen in the years defined previously. We looked at over 2,400 defencemen in their
17 year-old data in the CHL who played 10 or more games in that season. All data is from EliteProspects.com.
We binned
heights to see sized bins. We can already see a trend that the
bigger a player is the more successful his cohorts have been in becoming regular NHLers. Surprisingly, the number of CHL defencemen below 6’0″ was very limited. Below 5’11”, the numbers become so sparse the bin sizes had to be expanded to include more players.
heights to see sized bins. We can already see a trend that the
bigger a player is the more successful his cohorts have been in becoming regular NHLers. Surprisingly, the number of CHL defencemen below 6’0″ was very limited. Below 5’11”, the numbers become so sparse the bin sizes had to be expanded to include more players.
| Apprx Inch | Height (Range) cm | n | Success | % |
|---|---|---|---|---|
| < 6’0″ | [163-182.5) | 484 | 10 | 2.07% |
| 6’1″ | [182.5-185) | 387 | 16 | 4.13% |
| 6’2″ | [185-187.5) | 559 | 26 | 4.65% |
| 6’2″ | [187.5-190) | 408 | 31 | 7.60% |
| 6’4″ | [190-192.5) | 327 | 27 | 8.26% |
| 6’5″ | [192.5-195) | 194 | 20 | 10.31% |
| >= 6’5″ | [195-205) | 121 | 16 | 13.22% |
Similarly to our first look at this, we found the higher a players points/game statistics are, the more likely they player is to become a successful NHLer.
This is no
surprise as we found when performing the “Sham Sharron” experiment that we were able to
significantly beat multiple NHL teams using just points/game. Rhys also found similar results when looking at CHL defencemen and future NHL games played. We can summarize this best from a two-year old quote from Rhys:
surprise as we found when performing the “Sham Sharron” experiment that we were able to
significantly beat multiple NHL teams using just points/game. Rhys also found similar results when looking at CHL defencemen and future NHL games played. We can summarize this best from a two-year old quote from Rhys:
…while scoring in junior doesn’t guarantee NHL success, not scoring in junior more often than not predicts NHL failure.
| PPG Range | n | success | % |
|---|---|---|---|
| [0-0.1) | 697 | 3 | 0.43% |
| [0.1-0.2) | 660 | 9 | 1.36% |
| [0.2-0.35) | 573 | 28 | 4.89% |
| [0.35-0.5) | 271 | 31 | 11.44% |
| [0.5-0.75) | 192 | 38 | 19.79% |
| [0.75+) | 87 | 37 | 42.53% |
Results
Similar to the forwards results, here are the results of the CHL defencemen binned by height and points/game production.
| [163-182.5) | [182.5-185) | [185-187.5) | [187.5-190) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| n | success | % | n | success | % | n | success | % | n | success | % | |
| [0-0.1) | 108 | 0 | 0.00% | 105 | 0 | 0.00% | 149 | 1 | 0.67% | 111 | 1 | 0.90% |
| [0.1-0.2) | 119 | 0 | 0.00% | 100 | 0 | 0.00% | 154 | 0 | 0.00% | 108 | 2 | 1.85% |
| [0.2-0.35) | 131 | 2 | 1.53% | 91 | 2 | 2.20% | 123 | 2 | 1.63% | 91 | 3 | 3.30% |
| [0.35-0.5) | 51 | 0 | 0.00% | 44 | 3 | 6.82% | 70 | 3 | 4.29% | 49 | 8 | 16.33% |
| [0.5-0.75) | 50 | 3 | 6.00% | 29 | 5 | 17.24% | 46 | 12 | 26.09% | 34 | 7 | 20.59% |
| [0.75+) | 25 | 5 | 20.00% | 18 | 6 | 33.33% | 17 | 8 | 47.06% | 15 | 10 | 66.67% |
| [163-182.5) | [182.5-185) | [185-187.5) | |||||||
|---|---|---|---|---|---|---|---|---|---|
| n | success | % | n | success | % | n | success | % | |
| [0-0.1) | 108 | 0 | 0.00% | 105 | 0 | 0.00% | 149 | 1 | 0.67% |
| [0.1-0.2) | 119 | 0 | 0.00% | 100 | 0 | 0.00% | 154 | 0 | 0.00% |
| [0.2-0.35) | 131 | 2 | 1.53% | 91 | 2 | 2.20% | 123 | 2 | 1.63% |
| [0.35-0.5) | 51 | 0 | 0.00% | 44 | 3 | 6.82% | 70 | 3 | 4.29% |
| [0.5-0.75) | 50 | 3 | 6.00% | 29 | 5 | 17.24% | 46 | 12 | 26.09% |
| [0.75+) | 25 | 5 | 20.00% | 18 | 6 | 33.33% | 17 | 8 | 47.06% |
For a more
visual depiction here is a heat map of success rates based on the
size/production cohorts. It’s clear the
trend shows that the bigger and the higher producing a player is, the higher
their success rates of their cohorts has been.
visual depiction here is a heat map of success rates based on the
size/production cohorts. It’s clear the
trend shows that the bigger and the higher producing a player is, the higher
their success rates of their cohorts has been.
Another heat map showing the cohorts that have achieved below (red) and above
(green) defencemen success rates. Of the
2,480 defencemen, we looked at only 146 played in 200 NHL games or more,
translating to a success rate of 5.89%. Only 13% of CHL defencemen played in 1 NHL game or more.
(green) defencemen success rates. Of the
2,480 defencemen, we looked at only 146 played in 200 NHL games or more,
translating to a success rate of 5.89%. Only 13% of CHL defencemen played in 1 NHL game or more.
Conclusion
The same
conclusions drawn from our look at forwards can be applied to defencemen.
The larger the players and the higher their production, the more
likely the player is to achieve NHL success.
There’s a clear trend that both factors positively combined indicate
success, while it is clear that smaller players have a tougher battle to
face.
conclusions drawn from our look at forwards can be applied to defencemen.
The larger the players and the higher their production, the more
likely the player is to achieve NHL success.
There’s a clear trend that both factors positively combined indicate
success, while it is clear that smaller players have a tougher battle to
face.
However, this could
be a chicken and the egg scenario, and a type of self-fulfilling prophecy.
Bigger players are more likely be selected higher in the draft, and drafted players, especially those selected in the first round, are typically
given more opportunities than non-drafted players or later round picks.
That small gap in development aid may be what increases their production and NHL success rates, coupled with the fact that smaller guys with similar skillsets just don’t get the same number of chances to land a roster spot in a competitive industry.
be a chicken and the egg scenario, and a type of self-fulfilling prophecy.
Bigger players are more likely be selected higher in the draft, and drafted players, especially those selected in the first round, are typically
given more opportunities than non-drafted players or later round picks.
That small gap in development aid may be what increases their production and NHL success rates, coupled with the fact that smaller guys with similar skillsets just don’t get the same number of chances to land a roster spot in a competitive industry.
Size
does matter, and so does scoring. Analytics have never suggested you should not
select the bigger player. Both factors,
amongst others, need to be considered to try and maximize your success at drafting
assets. You can find bottom-6 forwards
and bottom pairing defencemen in free agency, but you should be using
analytics with traditional scouting to try and maximize your success in finding
those impact players.
does matter, and so does scoring. Analytics have never suggested you should not
select the bigger player. Both factors,
amongst others, need to be considered to try and maximize your success at drafting
assets. You can find bottom-6 forwards
and bottom pairing defencemen in free agency, but you should be using
analytics with traditional scouting to try and maximize your success in finding
those impact players.
Just
remember, scouts are not infallible alone.
remember, scouts are not infallible alone.
Application
Here are current ISS/CSS ranked D-men and the past success rate of their comparable peers, based on this exercise. The bracketed number in each rank column is their rank among North American defensemen.
| ISS Rank | CSS Rank | PLAYER | Position | P/Gm | Height | Success Rate | Cohort Rank |
|---|---|---|---|---|---|---|---|
| NR | 31 (7) | PILON, RYAN | D | 0.857 | 188 | 66.67% | 1 |
| 27 (4) | 45 (10) | MELOCHE, NICOLAS | D | 0.773 | 188 | 66.67% | 2 |
| 17 (2) | 16 (4) | CARLO, BRANDON | D | 0.488 | 196 | 50.00% | 3 |
| 10 (1) | 10 (1) | PROVOROV, IVAN | D | 1.025 | 183 | 33.33% | 4 |
| NR | 23 (5) | ROY, JEREMY | D | 0.975 | 183 | 33.33% | 5 |
| NR | 91 (21) | LOIK, LEVEILLE | D | 0.804 | 183 | 33.33% | 6 |
| NR | 38 (9) | JUULSEN, NOAH | D | 0.727 | 185 | 26.09% | 7 |
| NR | 67 (18) | CAPOBIANCO, KYLE | D | 0.628 | 185 | 26.09% | 8 |
| NR | 196 (25) | BELL, JASON | D | 0.609 | 188 | 20.59% | 9 |
| NR | 15 (3) | CHABOT, THOMAS | D | 0.605 | 188 | 20.59% | 10 |
| 29 (5) | 11 (2) | ZBORIL, JAKUB | D | 0.579 | 188 | 20.59% | 11 |
| 24 (3) | 51 (13) | SPENCER, MATTHEW | D | 0.561 | 188 | 20.59% | 12 |
| NR | 65 (17) | LAUZON, JEREMY | D | 0.548 | 188 | 20.59% | 13 |
| NR | 33 (8) | VANDE SOMPEL, MITCHELL | D | 1.111 | 179 | 20.00% | 14 |
| NR | 84 (20) | ANDERSSON, RASMUS | D | 0.953 | 182 | 20.00% | 15 |
| NR | 60 (16) | CARRIER, ALEXANDRE | D | 0.809 | 180 | 20.00% | 16 |
| NR | 47 (12) | DERMOTT, TRAVIS | D | 0.784 | 180 | 20.00% | 17 |
| NR | 68 (19) | BOURAMMAN, GUSTAV | D | 0.667 | 183 | 17.24% | 18 |
| NR | 133 (23) | BEAR, ETHAN | D | 0.659 | 182 | 17.24% | 19 |
| NR | NR | BOURQUE, SIMON | D | 0.578 | 183 | 17.24% | 20 |
| NR | 118 (22) | LALONDE, BRADLEY | D | 0.571 | 183 | 17.24% | 21 |
| NR | 53 (14) | DUNN, VINCE | D | 0.568 | 183 | 17.24% | 22 |
| NR | 46 (11) | WOTHERSPOON, PARKER | D | 0.543 | 183 | 17.24% | 23 |
| NR | 30 (6) | BRISEBOIS, GUILLAUME | D | 0.467 | 188 | 16.33% | 24 |
| NR | NR | PERILLAT, DUSTIN | D | 0.536 | 180 | 6.00% | 25 |
| NR | 59 (15) | GALIPEAU, OLIVIER | D | 0.342 | 185 | 1.63% | 26 |
| NR | 137 (24) | DAVIS, KEVIN | D | 0.488 | 180 | 0.00% | 27 |
A few names stand out on that list that I wish to touch on. By just his past height/points-game cohorts Brandon Carlo has a very high chance of success (even though Rhys doesn’t like him). This is because he is a very tall player and produces at an acceptable rate. Together, those attributes will catch the eye of an NHL team, likely cause him to be drafted in the top-30, and subsequently be given all of the opportunities his team can provide to increase his chance of becoming a successful.
A name much lower on the list, that is likely to have a much higher success rate compared to the other guys he’s rated around is Rasmus Andersson. Comparing him with only his past height/points-game cohorts suggests a mere 20% success rate (which is still good relative to the 5.89% success rate of all defencemen). What this data lacks is his past success as a 15 and 16 year old playing in the senior Allsvenskan league, and even playing on the J18 team as a 13 year-old.
Finally, the last intriguing name here is Jason Bell, who this exercise indicates is one of the top CHL defensemen available for this upcoming draft, but is hardly rated at all by Central Scouting Services. It’s a mystery as to why this is, especially considering he’s 6’2. Maybe he will climb in the draft rankings in the latter portion of the season since he’s been traded to a better team.
(Editor’s Note: this list only includes half of the draft eligible CHL D listed by NHL Central Scouting in their mid-term rankings. Players like Lethbridge’s Andrew Nielsen [193 cm tall, 0.447 Pts/GP, 16.67% cohort success rate] will be included once we finish our Canucks Army draft rankings sometime in the next few weeks.)
If we focus on current Canucks CHLers we see their success rates.
| Player | Position | Pt/Gm | Height | Success Rate | Cohort Rank |
|---|---|---|---|---|---|
| EVAN MCENENY | D | N/A | 188 | N/A | N/A |
| JORDAN SUBBAN | D | 0.75 | 175 | 20.0% | 1 |
| MACKENZE STEWART | D | 0.164 | 191 | 4.12% | 2 |
| ANTON CEDERHOLM | D | 0.225 | 188 | 3.30% | 3 |
| MILES LIBERATI | D | 0.214 | 183 | 2.20% | 4 |
It comes to no surprise that Stewart, Cederholm and Liberati all have below average success rates, given how infrequently they contributed in their CHL careers. As they are all 19 years old and none of them have really taken that next step in junior, the possibility that any of them are NHL talents is extremely remote.
Jordan Subban is an interesting case; because his height is a few standard deviations away from the median he has no comparables close to his height. Combined with his high scoring rate of 0.75 points/game in his draft year he has no one to be compared to. As he is part of the wider bucket of heights he sees his success rate jump to 20%.
Finally, Evan McEneny is a prospect that does not get enough attention. He was a very good free agent signing for the Canucks as he is big and has always scored at a very good rate. He went undrafted because he was injured as a 17 year old and only scored 2 points in 2 games, but tallied 34 points in 65 games (0.523 Pts/GP) the very next year. Defensemen that also stand 188 cm tall have a success rate of 7.6%, however it’s probably safe to assume that McEneny’s draft year production would have been in the 0.35-0.5 Pts/GP range, so I imagine his success rate is likely around 16%.
Recent articles from Josh W
