View Single Post
Old
02-18-2013, 12:41 PM
  #43
J17 Vs Proclamation
Registered User
 
Join Date: Oct 2004
Location: Reading.
Country: South Korea
Posts: 7,986
vCash: 500
Send a message via MSN to J17 Vs Proclamation
Quote:
Originally Posted by Mathletic View Post
I started posting rankings based stats on this board for the 09 draft. That said, I was only in my beginning back then. My list missed a ton of players and shouldn't be taken that seriously. Same for the 2010 draft, I had barely done any work during that year. I started re-working on my model for the 2011 draft. Back then I think I only posted mock drafts. I could look back at previous posts though to see if I actually posted rankings on HF. Last year I posted rankings every other months. I was still working quite a bit on my rankings then, so they could change quite a bit from month to month. The changes weren't only based on players performance but mostly based on the changes made to the model itself. That said, you could still see the model taking form and getting some good hits at the draft with guys like Hudon, Bozon, Severson and so on.
hfboards.hockeysfuture.com/showthread.php?t=1161587&highlight=
You'll need 3-5 years before we can devise a relatively accurate conclusion on these players, so i wouldn't count your chickens on Hudon, Bozon etc yet.

Quote:
Originally Posted by Mathletic View Post
II was still struggling comparing positions and leagues at that time.

This year would be the first year that I'm much more confident about my model. Much less changes from one edition to another. The whole rankings is objective also for this year.

That said, I could post my rankings for previous drafts. That is, applying the model to previous years. Note, I'm not using any data or stats that happened beyond that draft year. For example, I'm not simply doing a regression analysis of players out of the OHL from 1998-2008 and see how they did in the NHL to do my rankings, and then use that info to rank players for the 2004 draft.

On the first page, I posted a what if scenario. That is, I applied the model to previous drafts. I gave myself the 20th overall pick and the equivalent pick of subsequent rounds to see what would have happened.

On average, since 2000, my model predicts a "better player" (for a player at the same position (F, D, G). ex. say the St. Louis Blues selected Shawn Belle n'30 overall in 2003. The top rated d-man on my list at that spot would be Shea Weber. In comparison, Weber has played 495 games, scored 266 points and recorded 55ish point shares pending on the model compared to Belle's 20 games, 1 point and -1 point share) On average the player on my list compared to the player selected by the NHL team would be "better" roughly 60% of the time pending on the variable you want to use to compare those players.
Interesting. I would assume the lists take many people by suprise simply because they deviate so drastically from the percieved normality. Not neccessairly a bad or good thing. I disagree on some of the rankings of the few players i've seen. Nonetheless interesting. If nothing else, it creates conversation. Hopefully in a few years you will provide detailed reports on each year of measurements, so we can truely understand the merit and credibility of the model.

I look forward to the point in time where we can be reflective and know the models capabilities/limitations a little better. As a side note, how far in advance do you use these models for drafts? I am aware you have a provisional 2014 ranking. How limited are those rankings in work and projectability compared to draft year rankings? I'd be interested in seeing the effectiveness of the model a year or two out.

J17 Vs Proclamation is offline   Reply With Quote