2009 NHL Draft: Revisiting 2001 With Math
by Jonathan Willis on Jun 24, 2009 8:00 AM PDT
10 comments
The 2001 NHL Draft saw a much better job done by scouts than the previous three years; at least, when it came to making North American selections. Here's the draft order:
- Jason Spezza
- Stephen Weiss
- Ales Hemsky
- Colby Armstrong
- Adrian Foster
And here they are again, this time ranked by percentage of total team offense:
- Jason Spezza (63.8%)
- Stephen Weiss (37.7%)
- Adrian Foster (37.3%)
- Ales Hemsky (36.8%)
- Colby Armstrong (25.7%)
This is a decidedly worse ranking, although it should be noted that Foster played in only five games; in other words, anyone with half a clue about statistics would know not to trust that sample. Spezza's number is phenomenal, although he had a hot stretch for a miserable Missisauga team (before getting moved to Windsor) that probably inflated his totals too much.
Regardless, the NHL scouts are well ahead of the undiscriminating math. Let's se if that changes with late round gems. Here's my list of steals from 2001:
- Derek Roy (32nd overall)
- Jason Pominville (55th overall)
- Kyle Wellwood (134th overall)
- Ryan Clowe (175th overall)
- Brooks Laich (193rd overall)
- Marek Svatos (227th overall)
Here they are again, this time ranked by percentage of team offense:
- Derek Roy (46.3%)
- Kyle Wellwood (42.9%)
- Jason Pominville (30.6%)
- Marek Svatos (26.5%)
- Ryan Clowe (21.0%)
- Brooks Laich (10.6%)
I'd give this one ot the NHL scouts, and this well illustrates the need to use discernment with statistics. Kyle Wellwood's a uniquely talented player, but he's got all sorts of warts and there are good reasons for him bouncing around the waiver wire the last few seasons. The math has an edge with Derek Roy, but the scouts were better this year - which they obviously should be.
0 recs |
10 comments
|
Email
Print
Comments
Percentage of team offence
Jonathan, a question not just related to this post, but to your blog lately in general. Why the sudden heavy use of the percentage of team offence to rank prospects? Isn’t Desjardins’ NHLE a much better tool for ranking a prospect’s offensive abilities?
I see the percentages at best as a secondary tool to check against player’s NHLE numbers to confirm any assumptions about a player’s team situation, not a means of ranking players. Just curious why you seem to be doing so much work with them right now?
by Boondock Saint on Jun 24, 2009 8:31 AM PDT reply actions 0 recs
One big reason
Desjardins does great work, but it doesn’t allow for differences in team strength. So Player X scores 100 points on a team that scores 300 goals, and Player Y scores 90 points on a team that scores 200 goals. Player Y is probably better, but using Desjardins’ number straight won’t show that, so I prefer percentage of offense.
But Desjardins does a fantastic job of showing what junior numbers mean at the NHL level; what I really ought to do is normalize every player to a team that scores a median number of goals (say 250 or so) and then use Desjardins’ equivalencies to translate it to NHL offense.
If I ever get the time ;)
It's only my opinion, but it's right.
Writer for The Copper & Blue, OilersNation, and CanucksArmy.
by Jonathan Willis on Jun 24, 2009 10:57 AM PDT up reply actions 0 recs
80s offense
Have you ever ran up the NHLE numbers for players on teams that scored a ridiculous amount of goals, such as the 80s Oilers and Islanders? I was always curious what they would look like.
RT40 writes with Strange Deadfellows Oilerblog and is an avid hockey fan.
by raventalon40 on Jun 24, 2009 11:20 AM PDT up reply actions 0 recs
You’re missing the main point, Jonathan. Everything you do with percentages is based on the underlying assumption that draft-eligible players on teams that score more goals will have higher individual point totals. And that’s not necessarily true.
Team strength doesn’t always work in a player’s benefit. A draft eligible player on a strong team might get fewer PP time and less EV ice-time, whereas a draft eligible player on a weak team will probably play on the 1st PP and probably play more minutes.
Case in point, Ryan Howse vs. Burke Gallimore. They both scored at a similar point per game clip, Ryan on the last place Bruins, Burke on the first place Blades.
I was a Bruin season ticket holder the last two seasons, so I saw Howse a lot. The Bruins were horrible. Ryan lead the team in scoring while starting almost all PP’s on the top unit, sometimes playing the whole two mintues. At EV, he was thrown out for tons of offensive zone faceoffs. Basically, he was given the most offensive oppurtunities possible. I believe his goal totals are probably inflated based on the type of minutes he received. So not only are his offensive totals skewed from playing so many cherry minutes, his percentage of total offence will be skewed even higher because the rest of his team scored so few goals. Another way of looking at it, if the Bruins scored 308 goals last year, instead of 154, do I think Ryan’s offensive numbers would have doubled? Not a chance.
Second case, Burke Gallimore. 5th highest scoring forward on a really good team. He wouldn’t have been getting to play the whole PP, or even on the 1st PP unit all the time. And not only that, his percentage of total offence is driven way down because he team scores twice as many goals as the Bruins. If his team loses their older players and scores a lot fewer goals, do Burke’s numbers go down too? Perhaps, but they just as likely might go up as he would be one of their better players and get more offensive chances.
Based on your rankings, Howse has a 33.7%, while Gallimore has a 19.0%. If they switch spots, does Howse score 33.7% of 280 goals, which would be 94 points. Does Gallimore score 19.0% of 154 goals, which would be 29 points? I think Desjardins’ equivalency that they would put up similar numbers at the next level is much more realistic than yours based on percentages which gives Howse such a big advantage.
I know I’m kind of cherry-picking here, but I really believe the underlying assumption of players scoring a percentage of their teams offence is really flawed.
by Boondock Saint on Jun 24, 2009 12:01 PM PDT reply actions 0 recs
Thanks for the reply
And that’s a fair argument. I’ve been playing around with % of offense for a relatively short time, but I’ve been fairly excited by the initial results, and I think it does some things well, although I haven’t made a firm opinion on where its boundaries lie.
First off, I think there probably needs to be a middle range. Players at either end of the spectrum (awesome teams, horrible teams) are probably going to get unfair treatment using this measure. On the other hand, it does help compensate for a player on a poor team (like Scott Gomez) putting up poor scoring totals in his draft year.
Ultimately, I suspect that I’ll eventually end up softening the percentage using a modifier so that the curve is less radical on the ends.
Right now, you’re more or less watching me experiment with the statistic; I’m running it against these older drafts to figure out where its limits are. I have to admit too that part of the reason I’m so excited by this method (like the AHL QualComp) is that it’s a little off the beaten path – it borrows from what others are doing, but it’s also original enough that I don’t feel I’m copying the work of others.
In short: this is a measure with kinks in it, and I’m trying to work some of them out and see if it’s useful. As a stand alone measure it’s almost certainly flawed in a fatal way, but as a piece of the puzzle I think it could be very useful.
It's only my opinion, but it's right.
Writer for The Copper & Blue, OilersNation, and CanucksArmy.
by Jonathan Willis on Jun 24, 2009 12:36 PM PDT up reply actions 0 recs
Closer than you think
I think, for the most part, your use of % of team offence is closer than you think. This is obviously not a formula that is the be all and end all of draft selection, and I am impressed with the results you have shown so far. I think that if you make small tweaks, it will be spooky at how generally accurate the results will be. Radical adjustments or the inclusion of too many variables takes away from the purity of the math to strictly measure offence.
Can’t wait to see what your future test subjects are, keep up the great work.
by toprightcorner on Jun 24, 2009 10:38 PM PDT up reply actions 0 recs
Does this apply to Marc-Antoine Pouliot as well?
RT40 writes with Strange Deadfellows Oilerblog and is an avid hockey fan.
by raventalon40 on Jun 24, 2009 4:01 PM PDT up reply actions 0 recs
Without looking at the actual numbers, based on what I remember about his situation in his draft year, I’d expect him to high on a list for his draft year ranking by percentage of team points.
by Boondock Saint on Jun 24, 2009 4:18 PM PDT up reply actions 0 recs
Too literal
There are always going to be oddities in any mathimatical formula, but it is the general accuracy that needs to be looked at. This is a great tool to look at the probability of success in the NHL base purely on offensive statistics.
Who is to say that Howse could not handle the pressure of being the only go to guy on the team and buckle at the knees? There is now way to accurately measure that so we can’t assume that a strictly statisical analysis will always be accurate. There are always anomolies so all you can do is look for probabilities.
It is extremely difficult to use math to include not statistical variables such as work ethic, heart, attitude and natural hockey sense. So to expect all the variable to be considered, IMO is way to much to ask. Maybe Charlie Eppes could do it, but don’t expect Jonathan to create the “perfect” formula. If he could, he would end up being the star of a TV show and we would never have the privilege to see how his analytical brain works when he thinks about hockey!!
by toprightcorner on Jun 24, 2009 10:49 PM PDT up reply actions 0 recs
I think the first step would be to check if goals for has an effect on percentage of team offence for a player.
Do players coming from low scoring teams have a higher average percentage? Do players from high scoring teams have a lower average percentage?
by Boondock Saint on Jun 24, 2009 2:50 PM PDT reply actions 0 recs
Comments For This Post Are Closed
Secondary Sidebar
FanPosts
Community blog posts and discussion.
Recent FanPosts
Northwest Standings
Oilers Schedule
Next 5 Games
@
Toronto
|
Sat 03/13 | 4:00 PM PST |
|
@
Columbus
|
Mon 03/15 | 4:00 PM PDT |
|
@
Minnesota
|
Tue 03/16 | 5:00 PM PDT |
|
Detroit
|
Fri 03/19 | 6:30 PM PDT |
|
San Jose
|
Sun 03/21 | 5:00 PM PDT |
Last 5 Games
|
@
Montreal
|
Thu 03/11 | 4:00 PM PST |
|
Ottawa
|
Tue 03/09 | L 1 - 4 |
|
New Jersey
|
Sun 03/07 | W 2 - 0 |
|
Minnesota
|
Fri 03/05 | SO - W 2 - 1 |
|
@
Chicago
|
Wed 03/03 | L 2 - 5 |
Latest stories and headlines from NHL.com
- Franzen has two goals, Zetterberg adds goal and two assists lead Red Wings
- Mayer and Wallen score in shootout, Carrozzi stops 29 as Majors edge IceDogs 2-1
- Hutchings has four points as Barrie defeats Battalion 7-3
- Theberge scores twice, Missiaen stops 35 as Petes down Wolves 4-1
- Sisca's two goals, three assists lead Storm to 7-4 win over Sting
Certain photos copyright © 2010 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
