10-26-2018, 01:42 PM(This post was last modified: 10-26-2018, 01:44 PM by Coops.)
Quote:2830 words, double SMJHL draft media pay, and yeah lotta research
With the SMJHL draft last night and dozens of mock drafts, scouting reports and other assorted media on the topic (and drama, of course), I figured it would be a good time to release the SMJHL version of the win shares above replacement I've been writing about in the SHL media. You can read the first post here, and the second one here.
Backing up a little, this is going to use a fun little statistic called point shares to evaluate SMJHL players and prospects. Point shares are, in concise terms, the team points on the standings attributed to individual skaters on the ice. Really, they are a "single stat," something all the rage in real hockey statistic circles: something that tries to value a player's whole game. Now obviously we don't have the metrics that real hockey does (thank you STHS), so we're starting the race with a missing leg. But point shares are a fun way to hobble to "advanced" (LOL) metrics in the context of STHS.
Basically, point shares are determined from two subcategories: offensive point shares and defensive point shares. Putting aside the fact that this obviously doesn't include transition (which works since STHS doesn't either), it really focuses on outcomes and not method. Offensive point production is goals and assists (the latter weighted heavier). They are factored not just individual, but through a lens of the player's position, team and league. The defensive point share is calculated with more statistics that we track, like time on ice, goals against, shots against, etc. Once again through the filters of position, team and league. In the end, we get something called point shares.
I've tweaked it slightly from there in a desperate attempt to look cool in statistics. Basically I've converted to "win shares" through the ground-breaking formula of dividing point shares by 2. I've also followed the rest of real hockey statistics class and caved to peer pressure and weighted each player's win shares "against replacement." In our leagues, I've used the average rookie win shares as the replacement threshold. Why? They are the easy "fill-in" for a player in each level (in the SHL from calling up in the SMJHL, and in the SMJHL from new creates).
What do we end up with? Some fun statistic I've called "win shares above replacement" or wSAR. How useful is it? Don't ask questions. (AKA probably not at all, but have fun looking at it anyway).
Now - for the hard part (not that this wasn't hard). But in order to properly analyze where prospects ended up, we need to consider how good the prospects are. The catch? They are all starting out and have roughly the same statistics. So, I decided to investigate how to go about predicting a player's wSAR. If you're still reading (trying to follow that is), then you've realized we are waayyy down the rabbit hole. Fake statistic accumulating largely random statistics from a junior league? And we are now going to try and predict it? Crazy right.
How to go about it though? Well, I looked at all the S44 rookies last season and how their attributes correlated to their wSAR score. The result? Not well. Not well at all. Just take a look at this graph.
Fun first graph right? Basically there is little correlation between any of the attributes and a player's win shares (aka accumulated on-ice statistics). Of course this is mitigated by it being the SMJHL and not the SHL, so its hard to say with the sample size as well. But this definitely won't impact my dedication to making increasingly outlandish predictions for the S44 prospects.
Attributes are all well and good, but we know that TPE is the real king of the jungle in our sim leagues. So, thanks to @"luketd" I used his open S44 TPE tracking to compare the S43 SMJHL rookie class (S44 SHL) to their wSAR scores in S43 to get a look at how useful TPE might be for predicting scores.
In short, nope. Not really. There is basically no correlation in Week 1 TPE to a rookie's final wSAR score in the SMJHL, and there is a parabolic relationship to final season TPE and wSAR scores (isn't that fun?). It really goes to show how much of a crapshoot the SMJHL draft can be, especially when you consider how some of those low TPE earners as of the draft ended up being the highest in the class (my boy @StamkosFan was at 166 in Week 1 and went #2 OA in the S44 SHL draft with 291).
So what to do next? How about using all of these fun, poor correlation findings to predict rookie performance anyway? Sure. Let's look into it.
And here we go: predictions for rookies in terms of wSAR for S44.
A full listing here of predicted wSAR from high to low:
Sanyi Kocsis - 0.451
Alexander Selich - 0.447
Dominic Montgomery - 0.416
Hiro Fujikawa - 0.408
Konstantin Voloshin - 0.391
Leopold Lockhart - 0.386
Cassius Darrow - 0.37
Griffith Cadwalader - 0.36
Mika Mayfield - 0.357
Nolan Sawchuk - 0.325
Samuel Jalopski - 0.319
Andreas Kvalheim - 0.318
Jakub Novak - 0.313
Arsène Arsenich - 0.296
Donnie Dicks - 0.281
Patrick Stevely - 0.251
Cash Considerations - 0.229
Kenji Yoshimura - 0.226
Nat Emerson - 0.209
Evan Dale - 0.209
Vegeta Muerto - 0.202
Shooter McGavin - 0.198
Slip McScruff - 0.188
Elias Hughes - 0.186
Jimmy Cahill - 0.181
Ivars Ozols - 0.181
Andrew Joycon - 0.174
Chuck Freedom - 0.169
Cam Takinson - 0.163
Maxime Bouchard - 0.157
Marshal Ray - 0.149
Ben McBen - 0.129
Bobrice Gainergeron - 0.128
Big Troy - 0.124
Yannis Kanter - 0.113
Guy McCool - 0.096
Gareth Rush - 0.064
Tomy LeeNing - 0.021
Marc-Antoine Pepin - 0.019
Alex Henry - 0.018
Donk Lordington - 0.012
Tate Treehouse - 0.006
Misha Paketovich - -0.023
Jeffery Murphy - -0.057
Tim Phieffer - -0.082
Scott Stelling - -0.351
A couple things to note: the predictions seem to heavily weigh in favour of defensemen. Take a look at how they're predicted to do compared to the rest of the draft class. The 5th best defensemen prediction (Patrick Stevely at 0.251 pwSAR) would be the 2nd best projected center. The reason? Well defensemen tend to put lots into defense (yeah, look at that analytical insight) and defense trends well for all positions in terms of how they perform on ice, and therefore in wSAR. Centers tend to lean leftward in the distribution since they put TPE into things like faceoffs and puck handling which, for some reason, have even poorer correlations than the already poor correlations I used. Puck handling had negative correlations.
This seems as good a time as any to point out that no, this is not an attempt to figure out where you should develop. This is a fun, meaningless formula that aims to predict (very inaccurately I'm willing to guess) how a player will perform on ice. In the SMJHL. Using rookies from last season. Talk about a limited sample size and also flawed approaches. /Disclaimer over.
But, we are this deep, so leggo as they say. Let's look at how the draft went down last night.
From a draft perspective, some great picks made by teams last night, as well as some interesting ones from an on-ice perspective. Why I feel the need to disclaim, again, I don't know. But basically remember this is from an on-ice in your first season perspective. @"luketd" does a much better job at looking at TPE and activity - which is a much better barometer for future success and the sim league in general. But, since we're here, lets consider the draft through the lens of player's predicted on-ice performance in S44.
First off, great work in the first round by teams picking (by and large) the best predicted players. My own Detroit Falcons have the lowest predicted wSAR, but even that at 0.281 isn't a significant drop from the average score around 0.350. I'd argue a safe, and perfectly reasonable first round.
In the second round, things start to separate. Kelowna gets a great selection at 10th with Sanyi Kocsis (the highest predicted wSAR player this season). Colorado makes things interesting with Garreth Rush 13th - a decision that started all the drama we've seen today in the SMJHL. It looks like the formula is piling on him, but as a center the predictions aren't kind to him (as discussed above). There's also a couple players who joined later or I missed, who don't have scores - sorry boys.
Love the Hiro Fujikawa pick in the third round to Montreal - he's projected to be great this season with 80 scoring, 70 defense and 195 TPE at the time of the predictions. Also love the Nolan Sawchuk pick in the third. With Griffith Cadwalader in the 4th round, those three could end up being the steals of the SMJHL draft in terms of S44 production on-ice.
Since we're on a roll, let's look at how each team did in terms of predicted players (sorry again guys I missed).
ANCHORAGE ARMADA D - Leopold Lockhart - 0.386 LW - Mika Mayfield - 0.357 RW - Arsène Arsenich - 0.296 C - Kenji Yoshimura - 0.226 D - Vegeta Muerto - 0.202 RW - Maxime Bouchard - 0.157
Solid draft by the Armada, in terms of wSAR, where they added solid contributors in almost all positions. D gets two +0.2 wSAR infusions, and Mika Mayfield leafs the way on forwards, although Arsenich and Yoshimura are both better than average for their position.
COLORADO RAPTORS RW - Alexander Selich - 0.447 C - Konstantin Voloshin - 0.391 D - Cassius Darrow - 0.37 LW - Jakub Novak - 0.313 LW - Guy McCool - 0.096 C - Gareth Rush - 0.064
A top heavy draft for the Raptors, who added top-end talents like Alexander Selich, Konstantin Voloshin and Cassius Darrow (not to mention the above average predicted Jakub Novak). Interesting selections of Guy McCool and Gareth Rush, who is now on St. Louis.
DETROIT FALCONS LW - Donnie Dicks - 0.281
D - Jimmy Cahill - 0.181
A limited draft for my Falcons, but they added a solid left winger in a class with weaker options on the left than the right. For what its worth, Donnie Dicks is the 4th ranked LW in the class in terms of predicted on-ice performance. With a large rookie class in S43, it makes sense that Detroit reverted to the norm a bit with a smaller cohort entering S44.
I know that some of teams drafted more than I have listed, because I only listed the guys who I had predictions in wSAR for. So Halifax might have more than these two guys, but I'm too lazy at this point to go check. Regardless, damn I loved the Dominic Montgomery pick. I was thinking of calling it out above in the draft recap section, but I'll do it here. Great first round stud selection by the Raiders.
Great draft for the Knights who got two high-end players in Sanyi Kocsis (a fantastic second round pick) and Nolan Sawchuk at RW. Jeffery Murphy was picked later in the draft and could turn out to be a great value if he can work on earning some TPE and getting his defense a little higher for positioning on the ice.
MONTREAL MILITIA RW - Hiro Fujikawa - 0.408 RW - Griffith Cadwalader - 0.36 LW - Andreas Kvalheim - 0.318 D - Elias Hughes - 0.186 C - Marshal Ray - 0.149
The wingers lead the way for the Militia who absolutely stole Hiro Fukikawa in the third round, as well as Griffith Cadwalader in the fourth. Some savvy management here might've helped find some high-end contributors for rookies in their first season for the Militia. Elias Hughes also has a high prediction build, just needs to get on the TPE train now.
ST. LOUIS SCARECROWS D - Samuel Jalopski - 0.319
D - Slip McScruff - 0.188
The Scarecrows added two very solid defensemen prospects in the draft, and later centre Gareth Rush in a trade from Colorado. It'll be interesting to see how the rookie cohort plays out in St. Louis. McScruff's low offensive-oriented build tends to lean him lower on the wSAR predictions.
VANCOUVER WHALERS D - Cash Considerations - 0.229
The Whalers definitely drafted more than just Cash, but he's the only one I had any predictions in wSAR for. So - nice D pick Vancouver. Let's see how it turns out. Cash has a very solid all-around build, and this prediction is based of 155 TPE. If he gets on development high track this could be a great pick for the Whalers.
JUST FOR FUN - THE UNDRAFTED D - Patrick Stevely - 0.251 D - Nat Emerson - 0.209 LW - Evan Dale - 0.209 LW - Shooter McGavin - 0.198 LW - Andrew Joycon - 0.174 D - Chuck Freedom - 0.169 RW - Cam Takinson - 0.163 LW - Ben McBen - 0.129 C - Bobrice Gainergeron - 0.128 RW - Big Troy - 0.124 RW - Yannis Kanter - 0.113 RW - Tomy LeeNing - 0.021 D - Marc-Antoine Pepin - 0.019 RW - Alex Henry - 0.018 C - Donk Lordington - 0.012 C - Tate Treehouse - 0.006 LW - Misha Paketovich - -0.023 RW - Tim Phieffer - -0.082 LW - Scott Stelling - -0.351
Alright, folks, that's it for the SMJHL Draft S44: Prospect Predictions and Results piece. Keep an eye out for season previews for S44 in both the SHL and SMJHL on this exact topic, because why not? I have all the leg work done and its fun.
I love this Coops. I love that someone can use what I collected to actually do some deeper analysis and shit. Continue with it man, if you need help with any stats and stuff let me know I should be collecting data and stuff for a long long time, if you need any help or if you think I can help in any way let me know. I love these kinds of articles
10-26-2018, 02:38 PMFerdy Wrote: Very well written, good job!
Do you know how many TPE my player Cadwalader was at for the calculations? Out of curiosity, really.
No worries, 169 TPE was what I used from your update page - not sure if you've gotten more