HFBoards

Go Back   HFBoards > NHL Western Conference > Pacific Division > Edmonton Oilers
Notices

Draft analysis, total GP, and outlier results?

Reply
 
Thread Tools
Old
01-25-2007, 01:09 PM
  #1
Replacement
Now with 9% more zen
 
Replacement's Avatar
 
Join Date: Apr 2005
Location: Hockey Hell
Country: Canada
Posts: 36,134
vCash: 500
Draft analysis, total GP, and outlier results?

Hockeyaddicts's excellent draft threads spawned this but it got buried in one of those threads.

This thread is basically referring to analyzing draft pick relative success on the basis of counting total NHL games played by draft picks. Trouble is IMO the Hemsky type stats tend to dominate. 1, or 2 players in each year virtually make the NHL games played stat for a club.

I wonder what those more knowledgeable than myself on the subject think. Or anybody

Inherent draft analysis difficulty.

--------------------------------------------------------------------------------

Biggest difficulty IMO in draft analysis using NHL games played is that the dynamic drives results where statistically, the outliers rule the roost. The outliers in this case being players that make it to the show vs the overwhelming mode, or primary area of the sample that doesn't.

Statistically, outliers are often excluded from data sets due to distortion they create in the stats. But here they are the stats.

A related problem is that a few cases rather than the whole sample drive the stat.

But how to get around that? Are there ways to reduce the noise?

Not currently with the avaiable stats as far as I can see.

Solution?

I think whats necessary to improve drafting, improve accountability, provide accurate feedback loop, is for pro teams to start investing in BEFORE and AFTER draft analysis. performed by an independant, objective 3rd party(am I asking a lot?

This is necessary for accurate scientific measurement of drafting efficacy.

How it would work would be the strict draft numbers are compared at point of draft, i.e.# selected and the player would be subsequently evaluated respective to their draft class cohort in say 3, or 5, or even 10 year post evaluations. The results could be scored.

eg. If the player who was drafted at 118, 5 years later is ranked 323 out of that cohort this would be a -205 score on that teams ledger for that score. Conversely a player drafted 212 that rises to 111 would be +101score.

Results for each team for each year would be tabulated. One resultant +/- # would be derived for each/team/yr

Note that this system would not specifically be dependant whether the drafted player was in the NHL or not, or for how long(as this is a noted team by team distortion of results) but simply how they comparatively subsequently rank.

But the STRENGTH of this method of analysis is that it includes data on the entire draft sample rather than outliers and also controls for draft order or #pick selection.

The nightmare is in the detail of how to subsequently rank players playing in different leagues around the world but not much different than the pre haystack searching that currently goes on. The trouble though is whether the post score is appreciably more objective than the pre(draft) score.

Longitudinal study is a time and cost commitment. But one that usually pays dividends in the real world.
I think it could be $ well spent in terms of NHL optimal drafting.


Last edited by Replacement: 01-25-2007 at 01:27 PM.
Replacement is offline   Reply With Quote
Old
01-25-2007, 02:51 PM
  #2
t0mf00lery
Registered User
 
t0mf00lery's Avatar
 
Join Date: Sep 2003
Location: Barrie, Ontario
Country: Canada
Posts: 2,857
vCash: 500
Quote:
Originally Posted by Replacement View Post
This thread is basically referring to analyzing draft pick relative success on the basis of counting total NHL games played by draft picks. Trouble is IMO the Hemsky type stats tend to dominate. 1, or 2 players in each year virtually make the NHL games played stat for a club.
I understand what you are getting at and it seems like there is some real merit to this type of thinking. However, one thing I fail to see is how can knowing where each player sits on this hypothetical ledger in 5 or 10 years positively impact future draft results? Do you develop certain types of players to avoid based on this analysis? If so, what determines the parameters of the different types and groups of players?

t0mf00lery is offline   Reply With Quote
Old
01-25-2007, 03:38 PM
  #3
oil slick
HFBoards Sponsor
 
Join Date: Feb 2004
Posts: 7,359
vCash: 500
I see what you're saying, but I still think it would be useful to see. I think if you just took the average games played for each draft pick, at each draft position over a 20 year period, you'd get something resembling an exponential probability distribution. With the scale and location parameters, you can get an expected number of games played for every draft position. You're right that almost every player will be an outlier from this expected value (either above or below), but if you look at the total number of games played from all picks, over a long period and compare it to the expected number of games for those picks, I think this would still give an indication of good drafting from a team.

oil slick is offline   Reply With Quote
Old
01-25-2007, 03:50 PM
  #4
Replacement
Now with 9% more zen
 
Replacement's Avatar
 
Join Date: Apr 2005
Location: Hockey Hell
Country: Canada
Posts: 36,134
vCash: 500
Quote:
Originally Posted by Gretzky View Post
I understand what you are getting at and it seems like there is some real merit to this type of thinking. However, one thing I fail to see is how can knowing where each player sits on this hypothetical ledger in 5 or 10 years positively impact future draft results? Do you develop certain types of players to avoid based on this analysis? If so, what determines the parameters of the different types and groups of players?
All interesting questions. What well designed longitudinal study does tend to demonstrate after time is a clearer examination of the operating systems , variables, results.

What I'd expect is that some drafting teams would consistently rank higher than others and interestingly, possibly with central scouting which should also be evaluated. Other drafting teams would have scores they'd wish weren't exposed.

This might shed more light on operational drafting styles, select efficacy in different rounds, regional success i.e. domestic, Euro, when to gamble, when to choose the safer established pick, etc. some of this would seem to be common sense but...

All kinds of approaches seem to currently operate in the system and everybody thinks they're playing their hands the best on draft day.

I think my system inevitably differentiates the cardsharks from the crapshooters and would chase the posers from the highstakes table.

Nowhere to run and hide with these types of stats.

Trouble is the 3 to 5 year latency results in extended futility before draft team correction is recommended while 10 year doesn't bite any draft team butts but offers evidence on which draft teams did a consistently good job in the past.

The blue chip winners would hopefully share their secrets, likely at a much later date

In any case if these 3, 5 10 draft cohort stats were all available it would make junkies out of some of us


Last edited by Replacement: 01-25-2007 at 09:14 PM.
Replacement is offline   Reply With Quote
Old
01-25-2007, 05:07 PM
  #5
21Gator*
 
Join Date: Aug 2005
Posts: 1,123
vCash: 500
This is a case where I think statistics goes too far. I think the Oilers would be more successful in coming up with a better interview process then analyzing past drafts.

Drafting is about finding a kid who will become a man with the right amount of perseverance, hard work, talent, vision, etc.

Everyone is a gamble.

The fact is that you could probably get solid NHLers in the first three rounds if you knew how things were to turn out, but you can't know that, no matter how far you chase the numbers. Because it quite simply can't be known how the future is going to turn out.

21Gator* is offline   Reply With Quote
Old
01-25-2007, 09:09 PM
  #6
Replacement
Now with 9% more zen
 
Replacement's Avatar
 
Join Date: Apr 2005
Location: Hockey Hell
Country: Canada
Posts: 36,134
vCash: 500
Quote:
Originally Posted by oil slick View Post
I see what you're saying, but I still think it would be useful to see. I think if you just took the average games played for each draft pick, at each draft position over a 20 year period, you'd get something resembling an exponential probability distribution. With the scale and location parameters, you can get an expected number of games played for every draft position. You're right that almost every player will be an outlier from this expected value (either above or below), but if you look at the total number of games played from all picks, over a long period and compare it to the expected number of games for those picks, I think this would still give an indication of good drafting from a team.
Thanks for confirming that this could be considered an outlier. It would be interesting if a team on draft day had to proof each pick in the form of a detailed essay detailing the reasons for pick, time frame closely watched, games researched, sources spoken too, interviews, testing done, optimal expectation, games played prediction. Well, especially if they imo blow first round picks

I want someone writing essays... Maybe writing lines too

I will NOT pick another Jesse Ninimaaki in the first round

I will NOT pick another Alexei Mikhnov in the first round

I will NOT pick another Jesse Ninimaaki in the first round

I will NOT pick another Alexei Mikhnov in the first round

I will NOT pick another Jesse Ninimaaki in the first round

I will NOT pick another Alexei Mikhnov in the first round

I will NOT pick another Jesse Ninimaaki in the first round

I will NOT pick another Alexei Mikhnov in the first round

I will NOT pick another Jesse Ninimaaki in the first round

I will NOT pick another Alexei Mikhnov in the first round

I will NOT pick another Jesse Ninimaaki in the first round

Replacement is offline   Reply With Quote
Old
01-26-2007, 12:26 AM
  #7
hockeyaddict101
Registered User
 
hockeyaddict101's Avatar
 
Join Date: Jul 2002
Posts: 19,903
vCash: 500
I hate most draft threads

Look at the present one on Oilfans for example.

Its end with the post that look who the Oilers missed and could have pick at 228, etc, etc, etc. That would be great reasoning if the Oilers had all 227 picks before that player. And if that player was PREDICTED to be so great why did he slip to 228? It was because all 30 teams believed he was a long shot to every play in the NHL?

They list the busts the Oilers had, ok then list the busts for all the other teams. Do the Oilers have more? Do they have less? I can honestly list busts for EVERY NHL team in every year.

You would also have to list the successes from the same period and then list the successes for each team? Do the Oilers have more? Do they have less. ?

Very spurious reasoning and obviously not very informative.

Obviously it is IMO too early to be able to rate the draft (even from 5 years ago) and Lowe does not have enough of a draft history to rate him fairly but indications are that he has done decently when compared with other GM's so far. While Games played admittedly is not perfect it is better than having to read the rubbish that is posted in most draft threads.

As we have more history and more seasons GP's will be a fairly accurate barometer as it was for Glen Sather when we found out that in GP's he was middle of the pack. Sather was average (probably below average after the first few years which were obviously incredible), I don't think any one would argue that fact and the GP's stat painted a fairly accurate picture over the long haul.

Though it would obviously be interesting to rate by different criterias and then lists how the Oilers finished as compared to other teams in various categories.

1. Games played
2. Goals scored
3. Points scored.
4. GAA, GP for goaltenders drafted only.
5. GP, points for dman only.

IMO 2 and 3 and 5 over a short term will be more skewed than games played because the top 5 drafting teams should have a huge advantage but it would be interesting nonetheles.

4 would also be skewed because obviously over a short term few goaltenders are selected and the team that picked a goalie in the top 10 is going to win the category (not a worthwhile endeavour IMO over a short term)

hockeyaddict101 is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 12:14 PM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. 2014 All Rights Reserved.