pGPS 2016 Draft Extravaganza: Pacific Division

Draft Review - pacific

Image created by the great and wonderful Matt Henderson.

Well, we’re inching closer to the start of the hockey season, but before we get there, it’s time for the Big Unveiling of the full 2016 NHL Entry pGPS data set. I know. Try to contain your excitement.

Taking on the entire draft at once already produces a massive amount of information, but this year’s review is so replete with beautiful charts fashioned by our very own Petbugs, it would simply be outrageous to try to fit it into one article. That’s while we’ll be taking this journey one division at a time, with a league-wide overview at the end.

We’ll start with the Pacific Division and move east. No #eastcoastbias here! Why? Because it needed an arbitrary starting point and this endeavour originated in Vancouver. Plus, the Pacific has the highest volume of Nation Network affiliated teams, so they get preferential treatment I guess. Let’s dive in.

Terminology

This article is chock full of pGPS information. For a full initial review of the system, follow this link. A quick rundown of the columns used in the charts below will also include some quick info on pGPS metrics used in this article.

  • Draft #, PLAYER, POS, LEAGUE – Actual draft position, name, position and primary league in which they spent the 2015-16 season. (Note: a + next to a player’s name indicates that he was in his second, third, or fourth year of eligibility – otherwise known as an overage player).
  • Rank – A ranking derived from a hybrid list of the 2016 CSS North American and European final rankings.
  • Δ Pick – The difference between the rank and the actual selection. (+) indicates perceived steals, while (-) indicates perceived reaches.
  • GP, P – Games played and points accrued during the player’s draft season.
  • pGPS % – The percentage of statistical comparable players that went on to play at least 200 NHL games. The statistical similarity is determined by a Euclidean formula using exact age, stature and production. Matches are determined by a predetermined threshold that the statistical similarity must meet to qualify. In some cases, the threshold must be lower to find a respectable number of matches – in these cases, pGPS is adjusted by a certain factor to account for a potential decrease in accuracy due to slightly lower similarity.
  • Exp. pGPS – The expected pGPS % of the player’s actual draft position, based on all players selected at the 2016 draft.
  • Δ pGPS – The difference between the player’s pGPS % and the expected pGPS percentage at his draft position. (+) indicates value gained on the pick, while (-) indicates value lost, according to the model.
  • pGPS R – A combination of the player’s pGPS percentage and the average NHL points per game of his successful matches (pGPS P/GP). Balances likelihood of success with potential upside.
Three graphs are presented for each team. They visualize the following information:

  • Scouting ranking plotted against actual draft position. A demonstration of steals or reaches based on traditional scouting rankings.
  • Delta (Δ) pGPS plotted against actual draft position. A demonstration of value weighted by where the player was selected.
  • Expected Points (xPts, sometimes displayed as pGPS P/82) plotted against actual draft position. A demonstration of potential offensive upside based on their pGPS cohort.
All three graphs use pGPS R as the bubble size. Therefore it stands to reason that larger bubbles indicate a more positive overall projection in the eyes of the pGPS model.

The Team Overview chart reviews the selections made by a total and makes inferences based on the total and averages of the individual picks.

  • Selections – Number of draft selections, not including goalies. Goalies are not part of the pGPS model and thus are roundly ignored through this whole process.
  • Expected NHLers (pGPS) – The number of NHL players (200+ NHL games) that pGPS expects the team to pull from this draft. Based on the sum of pGPS percentages.
  • Expected NHLers (Pick position) – The number of NHL players that the model expects based on where the team picked, disregarding who the actual selections were. Based on the sum of Expected pGPS percentages.
  • Δ pGPS – The total Delta pGPS of the team’s selections. A metric that is designed to measure value above (+) or below (-) what was expected based on the team’s draft picks.
  • Δ pGPS / Selection – The value gained (+) or lost (-) per selection. Determined by dividing Team Delta pGPS by number of selection.
  • Δ Picks – The sum of player’s pick delta. An overall number indicating how far the team strayed from the standard scouting rankings, indicating tendencies of perceived steals (+) or reaches (-).
  • Overall Rating – The sum of the drafted player’s pGPS Ratings. The bigger the number, the more pGPS liked the draft class.

Anaheim Ducks

Draft # PLAYER POS LEAGUE Rank Δ Pick
GP Pts pGPS Exp. pGPS Δ pGPS pGPS R
24 Max Jones LW OHL 17 +7 63 52 23% 36% -13% 9.3
30 Sam Steel C WHL 36 -6 72 70 31% 32% -2% 17.7
85 Josh Mahura D WHL N/A 2 1 16%
93 Jack Kopacka LW OHL 40 +53 67 43 17% 15% +2% 6.7
115 Alex Dostie+ C QMJHL 276 -161 54 73 19% 12% +7% 9.9
205 Tyler Soy+ C WHL 216 -9 72 85 30% 3% +27% 15.3

London Knights winger Max Jones, Anaheim’s first selection, has an attractive skill set that had him rightfully viewed as a first round pick – the Ducks actually got him a little later than he was projected to go. His 23 percent pGPS is lower than what we’d hope for at that draft position, but given that it’s statistically based, anything that hampers his ability to put up points could influence the model to undersell him. In Jones’ case, he had the unfortunate luck to play the same position on the same team as Matthew Tkachuk. Between Tkachuk, and other Knights stars Mitch Marner and Christian Dvorak, Jones was precluded from spending any meaningful time on the first line or first power play unit. His pGPS comparables include Todd Bertuzzi, Glen Murray, and Zack Kassian.

Sam Steel of the WHL’s Regina Pats was taken by Anaheim’s other first round selection. His pGPS percentage was right around the expected value for that position, and he was ranked by scouts just a few spots later. His closest successful pGPS comparables are Matt Calvert, Scott Gomez, and Brendan Morrow.

Anaheim’s third round pick, Josh Mahura (85), is not present in the pGPS database because he was limited to just two games during the 2015-16 regular season due to injury. The Red Deer Rebels defenceman returned and played 17 games during the WHL playoffs, scoring two goals and adding two assists. He was held pointless in Red Deer’s four games as the host team at the Memorial Cup.

Jack Kopacka, a winger for the Soo Greyhounds, is praised for his work ethic and battle level. His comparables include Mike Fisher, Boyd Devereaux, and Adam Henrique. Gatineau Olympiques centre Alex Dostie put up a pile of points in his second year of eligibility. His 2015-16 season is comparable to the draft-plus-one years of Brad Marchand, Yanic Perreault, and Eric Belanger. Anaheim’s final pick was Cloverdale, B.C. native Tyler Soy, another centre in his second year of eligibility. Close matches include Jan Hrdina, Darren Helm, and Cody Eakin.

ANA Scouting

Anaheim picks were roughly in line with the scouting service rankings, with the exception of Alex Dostie (115), who was projected nowhere near the top 211. Jack Kopacka (93) looks to have been grabbed far later than he was ranked, while Jones (24), Steel (30) and Soy (205) were picked pretty close to where they were thought to go.

ANA Delta pGPS

This efficiency graph shows that Ducks’ selection hovered roughly around their expected value. Interestingly, the 93rd pick (Kopacka) seen as a steal and the 115th pick (Dostie) seen as a reach were roughly equal in Delta pGPS, with Dostie even squeaking out a small advantage. Tyler Soy, who scored 85 points in 72 games this year in the WHL, projects as the Ducks’ best value pick, with a pGPS percentage nearly 30 percent above what was expected in that spot.

ANA xPts

All of Anaheim’s picks are hovering in the 30 to 45 range in terms of Expected Points (pGPS P/82). Sam Steel (30), taken on the edge of the first round, has the highest pGPS percentage, Expected Points, and pGPS Rating of Anaheim’s draft class.

TEAM OVERVIEW

TEAM Selections Exp. NHLers

(pGPS)

Exp. NHLers

(Pick pos.)

Δ pGPS Δ pGPS /

Selection

Δ Picks Overall Rating
Anaheim Ducks 6 1.19 0.97 +22% 3.6% -118 58.9

On the whole the ducks had a good draft in terms of value for where they picked. pGPS estimates that they should produce at least one NHLer (200+ games) out of the group, which a chance of adding another. They seemed to have done reasonably well for where they picked. As always, the best way to increase the total number of NHLers pulled from a draft is to increase the number of picks that you make.

Arizona Coyotes

Draft # PLAYER POS LEAGUE Rank Δ Pick GP Pts pGPS Exp. pGPS Δ pGPS pGPS R
7 Clayton Keller C USDP 12 -5 62 107 89% 54% +34% 90.1
16 Jakob Chychrun D OHL 7 +9 62 49 50% 42% +8% 20.0
68 Cam Dineen D OHL 48 +20 68 59 50% 20% +30% 21.9
158 Patrick Kudla++ D OJHL 50 66 0% 7% -7% 0.0
188 Dean Stewart D MJHL 187 +1 42 22

Arizona rocked this draft, especially with its first three picks. Clayton Keller‘s successful matches are Patrick Kane and Phil Kessel, high scoring former wingers of the U.S. National Team Development Program.

The Coyotes managed to get two of the better defencemen in the draft, both of which were apprently heavily undervalued by other teams. Jakob Chychrun, once thought to be able to challenge Matthews for top spot, was still in most people’s top ten lists before the draft. He sunk like a stone on draft day, but GM John Chayka and company were willing to trade up to get their hands on this sturdy defender. His comparables include Drew Doughty, Bryan Berard, and Zach Bogosian.

Cam Dineen is clearly offensively talented, but warts in his defensive game had him ranked outside the first round. 68th overall is a steal for Dineen (the Nation Network had him ranked 39th), whose comparables include Trevor Daley, Dennis Wideman, and Michael del Zotto.

Patrick Kudla, already 20-years old, spent last season, his draft-plus-two season, in the OJHL. Next season, he’s slotted to head to the USHL for his draft-plus-three year, and the NCAA after that. Development this delayed does not project a very bright future. Dean Stewart spent his 2015-16 campaign in the Manitoba Junior Hockey League, clearly with the intention of retaining his NCAA eligibility – he’s committed to the University of Nebraska-Omaha for this season.

ARI Scouting

The Coyotes selected more or less within the range of what the scouts expected. Cam Dineen at 68th was a steal by CSS standards, and even more so by pGPS standards – he was among the highest projected defencemen in the entire class when viewed by statistical measures. Meanwhile, Patrick Kudla (158) wasn’t even ranked, while Dean Stewart (188) isn’t charted because he has no pGPS data.

ARI Delta pGPS

Arizona’s draft class has the look of one that was paying close attention to statistical measures. They gained huge value on their first three picks, especially on Keller (7) and Dineen (68). Kudla (158) doesn’t look so bad here, only 7 percent below expected value – however, expected value was only 7 percent, so that’s not good.

ARI xPts

The Coyotes are looking at a strong draft in terms of offensive upside, carried mostly by Clayton Keller. Despite some warts in other areas of his game, Dineen had some of the best offensive numbers of anyone available.

TEAM OVERVIEW

TEAM Selections Exp. NHLers

(pGPS)

Exp NHLers

(Pick pos.)

Δ pGPS Δ pGPS /

Selection

Δ Picks Overall Rating
Arizona Coyotes 5 1.89 1.23 +66% 13.2% +183 132.1

Arizona rocked the draft in every way measured by this project. They had the highest Delta pGPS and Rating per selection, and by a large margin. If they had made a couple more picks, they would have blown every other team out of the water.

As it is, pGPS anticipates that the Coyotes will get roughly two full time NHLers out of this draft. Barring a monumental bust or injury, you’d have to think that both Keller and Chychrun are sure things. Dineen may be a toss up, but he’s a player to watch as well, meaning that my Expected NHLers metric could be underselling this stellar class.

These reports don’t even include the fact that Arizona also picked up defenceman Anthony DeAngelo via trade at the draft, and power forward Lawson Crowse in August. Hats off to new Arizona GM John Chayka, who is three days younger than me. How depressing is that?

Calgary Flames

Draft # PLAYER POS LEAGUE Rank Δ Pick GP Pts pGPS Exp. pGPS Δ pGPS pGPS R
6 Matthew Tkachuk LW OHL 5 +1 57 107 80% 57% +23% 58.4
54 Tyler Parsons G OHL 55 -1
56 Dillon Dubé C WHL 52 +4 65 66 49% 23% +26% 27.5
66 Adam Fox D USHL 68 -2 25 22 31% 20% +11% 11.9
96 Linus Lindström C SuperElit 58 +38 40 44 5% 14% -9% 3.7
126 Mitchell Mattson C USHL 83 +43 21 2 2% 10% -8% 0.3
156 Eetu Tuulola RW Liiga Jr.A 97 +59 29 14 0% 7% -7% 0.0
166 Matthew Phillips C WHL 114 +52 72 76 22% 6% +16% 6.1
186 Stepan Falkovsky D OHL 128 +58 58 32 27% 4% +23% 6.6

Picking in the top ten is frequently going to land you a pretty impressive player, and this year was no exception for the Calgary Flames, who nabbed disruptive goal scorer Matthew Tkachuk at sixth overall. His comparables include truly elite talents like Steven Stamkos and John Tavares, as well as less esteemed but still talented players like Bobby Ryan and Bryan Little.

The Flames drafted London Knights goaltender Tyler Parsons at the end of the second round. pGPS isn’t not currently built to evaluate goalies. pGPS runs away. pGPS got away safely!

Dillon Dube of the Kamloops Blazers looks like a strong pick at 56th. His comparables that made it to the NHL range from Jordan Eberle to Adam Deadmarsh to Kris Versteeg. Jarret Stoll and Peter Schaefer are a couple of successful comparables that were closer to Dube’s draft position.

A defenceman out of the U.S. Development Program, Adam Fox‘s comparables that include Keith Ballard and Matt Carle. Frans Nielson and Gustav Nyquist met the similarity threshold for Linus Lindstrom, as did a number of players that never found success in the NHL. With just two points in 21 USHL games, Mitchell Mattson‘s lone successful match was Jared Boll.

As is often the case, pGPS was more impressed with the players taken from the Canadian Hockey Leagues, as opposed to the European junior leagues, which is sensible given that the CHL reliably provides more NHLers that lower European leagues. That said, both Lindstrom and Eetu Tuulola spent limited time in European elite leagues (four SHL games for Lindstrom, 10 Liiga games for Tuulola), which bodes well for their development. Tuulola’s pGPS percentage from his Liiga season is actually 17 percent, which demonstrates good value at 156th overall.

Matthew Phillips was one of the highest scoring first-time eligible players in the WHL this year, but at just 5-foot-7 and 170 pounds, there are plenty of questions about his ability to continue producing at higher levels. With his high production, he still managed a pGPS score of 22 percent, and his successful comparables (Nigel Dawes, Jordin Tootoo, and Scott Nichol, all of whom are in the 5-foot-8 to 5-foot-9 range) demonstrate that smaller players can still make it to the big leagues.

Stepan Falkovsky, the Flames final pick, wields an impressive pGPS score of 27 percent, and has Jassen Cullimore and Kurtis Foster among his comparables.

CGY Scouting

The Flames took advantage of the fact that other teams were consistently reaching into pools of lower ranked players, and came away with a draft class that is almost entirely composed of players taken later than they were ranked. Calgary’s latest pick, Stepan Falkovsky, picked 168th overall, was ranked 128th by the CSS hybrid ranking.

CGY Delta pGPS

The Flames gained value on each of their CHL selections, as well as on their USDP selection. Mitchell Mattson (126), who had strong high school numbers but abysmal USHL stats, drags the average value down, as do the Flames’ two European picks.

CGY xPts

The Flames walked away with plenty of firepower, led by their top pick, Matthew Tkachuk (6). Dube’s (56) comparables also show offensive potential, and while Lindstrom (96) had a poor percentage, his expected points score is quite impressive.

TEAM OVERVIEW

TEAM Selections Exp. NHLers (pGPS) Exp NHLers

(Pick pos.)

Δ pGPS Δ pGPS / Selection Δ Picks Overall Rating
Calgary Flames 8 2.17 1.42 +76% 9.5% +253 114.5

The Flames had a heck of a draft, managing to impress both in terms of value against scouting rankings and by the determination of pGPS. Additionally, with eight selections, the Flames not only managed to extract great per-player value, but great overall value, with a whopping 76 percent Delta pGPS across their draft class, which suggests that the got nearly a full extra player over expected based on their draft positions by virtue of strong picks.

Edmonton Oilers

Draft # PLAYER POS LEAGUE Rank Δ Pick GP Pts pGPS Exp. pGPS Δ pGPS pGPS R
4 Jesse Puljujärvi RW Liiga 3 +1 50 28 94% 63% +31% 60.1
32 Tyler Benson LW WHL 29 +3 30 28 33% 31% +2% 18.6
63 Markus Niemeläinen D OHL 71 -8 65 27 24% 21% +3% 7.5
84 Matthew Cairns D OJHL 120 -36 46 33 5% 16% -11% 2.0
91 Filip Berglund+ D SuperElit 168 -77 43 41 7% 15% -8% 3.6
123 Dylan Wells G OHL 171 -48
149 Graham McPhee LW USHL 157 -8 20 5 1% 8% -7% 0.5
153 Aapeli Räsänen C Liiga Jr.A 108 +45 50 38 7% 7% 0% 4.0
183 Vincent Desharnais++ D NCAA N/A 19 2 12% 5% +7% 2.4

It really wouldn’t be an NHL draft without an elite player falling into the Oilers’ lap, and that’s exactly what happened once again when Columbus took Pierre-Luc Dubois at third overall and allowed Jesse Puljujarvi to slip to the Oilers at four. Puljujarvi’s closest comparable to his draft season in Liiga is Olli Jokinen – of course, you don’t need cohort models to know how special Puljujarvi is, especially if you got the opportunity to see him dismantle his competition and entirely dominate the Young Stars Classic tournament in Penticton recently.

The second round wasn’t much different, as the Vancouver Giants’ Tyler Benson fell to the Oilers at 32. Once a projected top ten pick, Benson has been severely hampered by injuries – and they seem to keep on coming – he missed the prospects tournament in Penticton due to a shoulder ailment. This was of course the risk in selecting the oft-injured winger, but if he manages to stay healthy, the payoff will be well worth it. Top pGPS matches include Jarome Iginla, Jordan Eberle, and Adam Deadmarsh.

The Oilers grabbed towering Finnish defenceman Markus Niemelainen at 63rd overall. While our CSS hybrid ranking had him at 71, the Nation Network rankings pegged him 44th. Niemelainen has Brent Burns and Derian Hatcher among his pGPS comparables, despite relatively unimpressive point totals. There was much talk among scouts, however, that his numbers could – or should – have been better than they were, with many suggesting that he was mishandled in Saginaw.

Playing in a Junior A league in your draft year is a good way to lower your projections, but it is necessary to retain college eligibility. Matthew Cairns will head to the USHL this season, and is committed to Cornell University the following year. Kevin Bieksa is among his few comparables that stuck in the NHL. Fifth round pick Graham McPhee, son of Las Vegas general manager George McPhee, is in a similar situation: he played in the USHL last season, and is joining Boston College this year. His lone successful comparable is Nashville’s Craig Smith. Vincent Desharnais, another towering defenceman, may have been in his third year of eligibility, but he’s already played in the NCAA, and thus garnered a slightly higher projection. His comparables include Hal Gill and Eric Gryba – defenders with impressive height but less than impressive scoring.

Defenceman Filip Berglund spent most of the 2015-16 season in SuperElit, the top Swedish U20 league, though he did have a five game cameo in the SHL. Alexander Edler is his most successful comparable. Centre Aapeli Rasanen‘s comparables include Jussi Jokinen, Valteri Filpula, and recent Oiler Lauri Korpikoski.

The Oilers also took a goaltender, Dylan Wells of the Peterborough Petes, at 123rd overall. He seems to have been taken earlier than he was projected to go. Learn about him and other goalies from the experts at inGoal magazine instead.

EDM Scouting

The Oilers followed convention for the first few picks, before twisting all over the place in their remaining selections. Aapeli Rasanen (153) is considered the biggest steal here, while Filip Berglund (91) is the biggest reach. However, the Oilers walked away from this draft with Puljujarvi and Tyler Benson, so no one really cares what they did with their remaining picks.

EDM Delta pGPS

Even at fourth overall, Jesse Pulujarvi shows tremendous value over expected. The rest of the OIlers’ pick hover within about plus-or-minus ten percent or breaking even.

EDM xPts

No surprises here, but pGPS sees good offensive upside in Puljujarvi and Benson.Berglund (91), McPhee (149), and Aapeli Rasanen (153) all demonstrate high offensive prowess despite having unimpressive projection scores.

TEAM OVERVIEW

TEAM Selections Exp. NHLers (pGPS) Exp NHLers

(Pick pos.)

Δ pGPS Δ pGPS / Selection Δ Picks Overall Rating
Edmonton Oilers 8 1.83 1.66 17% 2.1% -80 98.7

The Oilers did a fair bit of reaching, at least according to the standard rankings. On the other hand, pGPS found them to be in the value in terms of projected value over expected value. The model expects them to get roughly two full-time NHLers out of this class. While Puljujarvi and Benson seem like the natural locks, Benson’s injury history is still a risk factor not captured by the model.

Los Angeles Kings

Draft # PLAYER POS LEAGUE Rank Δ Pick GP Pts pGPS Exp. pGPS Δ pGPS pGPS R
51 Kale Clague D WHL 32 +19 71 43 20% 24% -4% 6.6
112 Jacob Moverare D SuperElit 64 +48 41 21 6% 12% -6% 1.5
142 Michael Eyssimont+ C NCAA N/A 40 33 22% 8% +14% 11.1
202 Jacob Friend+ D OHL N/A 54 21 3% 3% 0% 0.8

Kale Clague, a defenceman with the 2016 WHL champion Brandon Wheat Kings, was projected to go higher than his eventual 51st overall, according to the CSS scouts. Some of his successful matches are Mike Green, Josh Gorges, Tyson Barrie and David Schlemko.

Jacob Moverare, a Swedish defenceman out of the NV71 U20 program in the SuperElit league, had Niklas Hjarmalsson as his lone match that hit 200 NHL games, though Hjarmalsson’s former teammate David Rundblad was also on the list.

Michael Eyssimont, a centre for St. Cloud State University, was in his second year of eligibility, but given his early September birthday, he’s almost the same age as some of the first-time eligible players this year. He put up impressive numbers as an NCAA freshman, and pGPS commends him accordingly. Comparables include Rich Peverley, Matt Cullen, and Dominic Moore.

Jacob Friend was also in his second year of eligibility, but comes with a considerably lower projection, owing mostly to lackluster production. His comparables include Jamie Allison, Sean O’Donnell, and Dan Girardi.

LAK Scouting

The Kings managed to pick up Clague (51) and Moverare (112) later than they were projected to go, but their other two draft picks – Eyssimont (142) and Friend (202) weren’t even ranked by Central Scouting.

LAK Delta pGPS

Los Angeles just just slightly below expected value on their first two picks, and broke even on their final selection. Eyssimont makes up for the rest of the class by being a solid value selection at 142nd overall.

LAK Delta xPts

There isn’t a ton of offensive upside to go around here, but Eyssimont represents the highest potential, with a projected points per 82 games of 40. It’s worth noting of course that the other three picks were defencemen.

TEAM OVERVIEW

TEAM Selections Exp. NHLers (pGPS) Exp NHLers

(Pick pos.)

Δ pGPS Δ pGPS / Selection Δ Picks Overall Rating
Los Angeles Kings 4 0.52 0.48 +4% 1.0% +67 20.0

The Kings didn’t do themselves any favors by working their way down to just four picks. They got approximately the same projected value as was expected based on their draft positions, but that isn’t a particularly good one – about half an NHLer. Kale Clague is their best shot, but it wouldn’t be all that surprising to see the Kings with nothing to show for this year’s haul.

San Jose Sharks

Draft # PLAYER POS LEAGUE Rank Δ Pick GP Pts pGPS Exp. pGPS Δ pGPS pGPS R
60 Dylan Gambrell++ C NCAA 92 -32 41 47 33% 22% +12% 23.6
111 Noah Gregor C WHL 61 +50 72 73 31% 12% +19% 18.4
150 Manuel Wiederer+ C/RW QMJHL 313 -163 54 64 6% 8% -1% 3.1
180 Mark Shoemaker D OHL N/A 67 13 4% 5% -1% 0.8
210 Joachim Blichfeld RW/LW SuperElit N/A 45 28 4% 2% +2% 2.5

The Sharks’ first pick wasn’t until 60th overall, and they use it on an NCAA centre in his third year of eligibility. Dylan Gambrell had a breakout year with the University of Denver Pioneers, scoring 47 points in 41 games and earning him a 33 percent pGPS score (with comparables that include Derek Stepan and Tony Amonte), but his previous seasons in the USHL were uninspiring, contributing to his availability at age 20.

The Sharks got Moose Jaw Warriors centre Noah Gregor at pick 111, far after he was projected to go. Our hybrid Central Scouting ranking had him at 61, but our Nation Network countdown held him in even higher esteem, at 41. His comparables include Ray Whitney, Scott Gomez, and Clarke MacArthur.

As a QMJHL forward in his second year of eligibility, it isn’t surprising that Manuel Wiederer‘s pGPS percentage is quite low. Not much is expected of a 150th overall pick, so not much value was lost here. Scanning his QMJHL cohort, we see the David Krejci has slipped in there, as has Michael Frolik – but the vast majority saw little or no time in the NHL.

Mark Shoemaker‘s few successful comparables include Jake Muzzin and Jakub Kindl. Their final selection, Danish winger Joachim Blichfeld, has Kristian Huselius, Jakob Silfverberg, and Marcus Kruger in his statistical cohort.

SJS Scouting

Central Scouting viewed Gambrell (60) as a reach, while Gregor (111) can be viewed as a steal. The three remaining picks weren’t anywhere close to where they were ranked by the scouts.

SJS Delta pGPS

Despite the protests of scouts, pGPS has found the Sharks’ picks to be somewhat agreeable. They gained plenty of value on their first two picks, while the last three hovered around what was expected for their respective draft positions.

SJS xPts

While San Jose’s draft class is only a little above average in terms of graduation percentages, there’s plenty of offensive upside to be had here. The Sharks picked up four players with Expected Points values above 40, which is pretty spectacular. Gambrell (60) leads the pack, with a pGPS Points per 82 of 58.

TEAM OVERVIEW

TEAM Selections Exp. NHLers (pGPS) Exp NHLers

(Pick pos.)

Δ pGPS Δ pGPS / Selection Δ Picks Overall Rating
San Jose Sharks 5 0.79 0.49 +30% 6.1% -145 48.5

For a team that only had five selections and didn’t pick their first player until 60th overall, the Sharks acquitted themselves quite well, gaining about 30 percent value over expected. That said, the pGPS model doesn’t expect them to walk away with more than a single full-time NHLer here. Additionally, the view from the traditional scouting perspective indicates that they did a lot of reaching.

Vancouver Canucks

Draft # PLAYER POS LEAGUE Rank Δ Pick Pts GP pGPS Exp. pGPS Δ pGPS pGPS R
5 Olli Juolevi D OHL 8 -3 42 57 41% 60% -19% 15.9
64 William Lockwood RW USDP 154 -90 33 59 6% 21% -15% 1.6
140 Cole Candella D OHL 119 +21 20 37 14% 9% +5% 4.6
154 Jakob Stukel+ LW WHL 115 +39 60 69 7% 7% 0% 3.5
184 Rodrigo Abols++ C WHL 307 -123 49 62 9% 4% +5% 2.9
194 Brett McKenzie+ C/LW OHL N/A 53 66 15% 4% +11% 5.0

Elite defenders like P.K. Subban, Drew Doughty, and Alex Pietrangelo pop up in Olli Juolevi‘s comparables, but his pGPS percentage is below what was expected at fifth overall. I’ve spoken of this numerous times before. His projection is vastly improved when accounting for pedigree by using draft position as a proxy), improving to 70 percent when comparing against first rounders and 100 percent when comparing against other top ten picks.

William Lockwood played in several different leagues in 2015-16, with varying levels of success. Former top ten pick Jack Skille is Lockwood’s only successful comparable for his time with the USDP. Poor scoring totals in the USHL netted him no successful comparables at all. He was at his most productive in international play – he had seven points in seven games at the IIHF under-18 tournament.

Cole Candella‘s comparables include Dennis Wideman, T.J. Brodie, and Carlo Colaiacovo, with the majority of his matches being second pair defenders. Jakob Stukel‘s matches include Martin Erat, Kris Versteeg, Blake Comeau, and Kyle Brodziak.

Rodrigo Abols, a centre in his third year of eligibility, had comparables that included Paul Gaustad, Dave Scatchard, and Lance Bouma, after a pretty unimpressive 20-year old season in the WHL. He spent part of the 2014-15 season in the KHL with Dynamo Riga, where his pGPS numbers were substantially better (38 percent) with comparables that included Alexei Ponikarovsky (one of my favourite names to say) and Artem Anisimov. He’ll need to recapture some of that KHL success if he wants to improve as a prospect.

Despite being the Canucks’ final selection, Brett McKenzie has one of the better projections among their class, particularly in relevance to where he was taken. His cohort is populated by NHL fourth liners like Stephane Yelle, Chris Neil, and Tom Kostopolous, though Logan Couture and James Neal snuck in there as well.

VAN Scouting

Juolevi (5) was grabbed a couple of spots before his consensus ranking, but it was certainly a defensible selection. Candella (140) and Stukel (154) were grabbed well after they were projected to go, but Lockwood (64) was grabbed a couple of rounds earlier than expected. Abols (184) wasn’t ranked anywhere near the top 211, and McKenzie (194) wasn’t ranked at all.

VAN Delta pGPS

The Canucks gained value on most of their later round picks, despite a large focus on overagers and general dissatisfaction from the fanbase. Brett McKenzie (194) represents the greatest value over expected value among the Vancouver class. Juolevi’s apparent loss of value is due to lower point totals that what would be expected from a pick that high. His international play and Memorial Cup domination suggests that he is capable of better totals than he provided in the 2015-16 regular season.

VAN xPts

With no players having an Expected Points value (pGPS NHL points per 82 games) over 40 (Stukel comes closest at 39), Vancouver came away with a rather uninspiring class in terms of offensive upside.

TEAM OVERVIEW

TEAM Selections Exp. NHLers

(pGPS)

Exp NHLers

(Pick pos.)

Δ pGPS Δ pGPS / Selection Δ Picks Overall Rating
Vancouver Canucks 6 0.92 1.04 -12% -2.0% -156 33.6

The Canucks lost a small amount of value based on the expected pGPS for their draft positions, but with a top five selection in Olli Juolevi, the Canucks should at least be all but guaranteed one NHLer out of this draft class. With an Expected NHLers number of 0.92 (calculated by total pGPS), they should consider themselves pretty lucky if they manage to get a second full time player out of this group. The next best candidate is probably Lockwood, despite his low projection, with the hope being that he can produce more like he did in international play going forward.


This concludes our pGPS review of the Pacific division teams’ draft classes. Early analysis at this point has both Alberta teams looking strong, and Arizona leading in terms of Overall Rating (total of pGPS R), indicating strong likelihoods of success and enticing offensive upside. We’ll continue the series by moving east, starting with the West’s Central division.

  • Bob Long

    Here’s a pGPS exercise request. How do true #1 d-men score with pGPS. e.g. Scott Neidermayer, Pronger, Ohlund, Lidstrom. Especially those drafted top 10.

    • Because you asked nicely:

      Scott Niedermayer: 86%
      Chris Pronger: 81%
      Matthias Ohlund: 53%
      Nicklas Lidstrom: 14% (?!?!?!)

      Some notes: all but Ohlund were run in the year they were drafted. Ohlund played his draft season in a league that is not part of my database. I used his draft+1 season instead.

      Lidstrom’s number is on the surface very surprising, as one of the greatest defencemen ever. But bear in mind, he scored just 2 points, both assists, in 20 games in his draft year, and was taken in the mid-3rd round.

      Taking this exercise further would be entertaining and perhaps a good idea for a later date.

  • Bob Long

    Jeremy, I’ve mentioned this to you before, as a matter of methodology, but it bears repeating: manipulating the model by correcting for “pedigree”, in the case of Juolevi, represents bad data management. The purpose of this tool, from my understanding, is to provide an unbiased predictive metric to evaluate players independently of their scouting ranking, as a way to compare statistically predicted success rates with those subject to the view of scouts and GMs. By baking those biases into your score, you invalidate the findings of the model. If the model is inadequate as a predictive tool, that’s fine, but you can’t selectively tinker with it to get the results you want or else you have taken away the objectivity of the results.

    I happen to think the pGPS model has some kind of deficiency when it comes to accounting for the results of international play (itself “noisy” data, but still important to consider as it tends to inform subjective scouting consensuses), but that has to be corrected for within the model, not by ex post facto data manipulation.

    • Graphic Comments

      This would be a valid critique if the actual pGPS numbers were adjusted, but they’re not. He’s still shown at 41%, which is where the base model puts him. The adjustment is strictly in the discussion to provide some additional context and recognizing the fact that he was considered a top 10 prospect in this draft class.

      If anything, this additional context adds some nuance to the purely numbers-driven analysis, which is what pGPS produces. At the end of the day, marrying together a data-driven approach with traditional scouting rankings is going to give you better results and a better evaluation of the probability of success imo.

      pb

      • OK, but he hasn’t done so universally. If he wants to give context to the Canucks’ draft picks in an article where comparisons are made, the same must be done for others’ as well. The tone of that part of the article was that “well, Juolevi looks like a reach at #5 based on the numbers, but when you consider other things like scouting reports and draft rankings, it looks better”. Fine, but only making that statement for the one pick means that the qualitative assessments that follow don’t cover an even playing field. Specifically, most readers of this article would want an idea of whether passing on Tkachuk to take Juolevi was statistically justified. However, Tkachuk’s raw numbers are not similarly adjusted, so we don’t have a way of comparing the “nuanced” part of the assessment between them.

        If you want to change the methodology used to produce your findings, that’s your business. I’m just asking for proper rigour in using the data (whose generation we can’t duplicate because of the proprietary model, to verify your findings or extend upon them, meaning we have to trust you) to make unbiased comparisons. I know this site isn’t a peer-reviewed scientific journal or anything, but in these data rich posts, it’d be nice not to see obvious no-nos like selective interpretation of the data.

    • I use proxies for pedigree only for context, and not to change the actual reported numbers. That being said, suggesting that it’s “bad data management” is not necessarily fair or true. Previous research by Cam Lawrence and Josh Weissbock has suggested that incorporated scouting rankings into statistical models leads to better draft results, not more volatile ones.

      It is simply a way of including information that basic statistics (and thus this model) cannot account for. The only reason I don’t use it more often is because it reduces sample size to unusable proportions in many instances.

      I don’t typically use international play to create pGPS scores for a number of reasons, and if I do, it is again only for context.

    • Bob Long

      This.

      Working the model to fit your preconceptions is exactly whats wrong with most of the so-called advanced stats in hockey. It would be laughed off any scientific review board. Is that setting the standard too high, maybe, but its the only thing thats ever going to give this any credibility.