፨ Team Prospects
Provides predicted performance (NHLdx - draft/development expectations for NHL PTS/82) for each prospect.

፨ Team Depth Ratings
A simple view of each team's prospects laid out in table form for easy viewing as a unit. Also includes an overall rating for the club's forward depth, based on the quality and quantity of players expected to make the NHL.

፨ 2020 NHL Draft
Early predictions for the top players in the upcoming NHL Entry Draft based on statistical information and consensus rankings.

፨ Past NHL Drafts
A simple view of each team's prospects laid out in table form for easy viewing as a unit. Also includes an overall rating for the club's forward depth, based on the quality and quantity of players expected to make the NHL.



The Basics 

Welcome to #prospectpred.

Looking at prospect reports (and for one player specifically), I wanted to know what similar players had done at the NHL level - historical precedents so we could predict future performance.

#prospectpred is the result - a classification and regression model based on age, league, draft position, and age-adjusted points. NHLdx is the measure (NHL Draft/Development Expectation - NHL points at the player's peak; measured in points/82 games).

Read more about the model here.



Results 

For each player, presentation of the data includes:

  + Predicted points range (their NHLdx with an expected high and low based on average error for that points range)



  + Where the player's NHLdx would fit on an NHL depth chart (e.g., first/second/etc. line)
  + Their likelihood of making it to the NHL (Make It%)

Updates are provided for each of the players' seasons following their draft (so in total we have predicted production for ages 18, 19, 20, 21, 22, and 23). To help address the randomness of small samples, for all seasons past age 18, the NHLdx number weights performance, which means that an age 19 season's NHLdx will include information from the age 18 season; age 20's NHLdx will include information from the age 19 season, etc.

For line breakdowns, a basic rule of thumb here is 55-40-25-15 for minimum 82-game point totals for first-second-third-fourth lines (the exact numbers may differ by a few points in any given year).



As one would expect, the accuracy of the model increases each year that the player ages, and so we see an increase in correlation and decline in average error.  With the weighting of results, stability between seasons is fairly high.



Notes/Disclaimer 

Please don't yell at me if it hates your favorite player. I just figured that something was better than nothing, and so I decided to share something that developed out of sheer curiosity for me. I'm not a data scientist. There may be better ways of doing this, but this is what I've got.

Data was gathered for this project from Elite Prospects, which is an invaluable resource for hockey statistics and I highly recommend subscribing to their premium version if you are interested in doing your own work.

At the moment, the model only includes forwards. Expanding this to defense is certainly on the to-do list (I don't remember why I separated F and D), as is accounting for changes in scoring rates by era.

Read more about the model here.