Sketching a multi-year player value model

Home for all your discussion of basketball statistical analysis.
Post Reply
Crow
Posts: 6188
Joined: Thu Apr 14, 2011 11:10 pm

Sketching a multi-year player value model

Post by Crow » Thu Jun 08, 2017 6:29 pm

A player has projected impact value of x1 in year 1, x2, x3, x4 or x4 and beyond per game / year considering expected minutes played. Actual or projected salary of y1 to y4. A retention probability of z1 to z4 (only worries about involuntary loss). Find x, y, z values for expected median replacement if lost, given draft picks, money, trade assets, etc.

This is just a simple model. What else would you need to calculate expected 4 year (or 4 plus) value? Weights for the importance of the years 1 to 4 for your team. What else?

"Trade value" is a more complicated exercise but probably could use that as a base main component.

Statman
Posts: 548
Joined: Fri Apr 15, 2011 5:29 pm
Location: Arlington, Texas
Contact:

Re: Sketching a multi-year player value model

Post by Statman » Sat Jun 10, 2017 2:55 am

Crow wrote:A player has projected impact value of x1 in year 1, x2, x3, x4 or x4 and beyond per game / year considering expected minutes played. Actual or projected salary of y1 to y4. A retention probability of z1 to z4 (only worries about involuntary loss). Find x, y, z values for expected median replacement if lost, given draft picks, money, trade assets, etc.

This is just a simple model. What else would you need to calculate expected 4 year (or 4 plus) value? Weights for the importance of the years 1 to 4 for your team. What else?

"Trade value" is a more complicated exercise but probably could use that as a base main component.
My player projection model projects out any/all players throughout a career (NBA season & playoffs, G League, D1, Summer League, & Exhibition league normalized & weighted data used from past 2 seasons) - with varying career curves based on exact age, skill sets, etc. Applying a theoretical projected salary "value" for future seasons (1 thru 4 or beyond) wouldn't be all that hard if the future salary structure of the NBA could be somewhat accurately surmised for those seasons.

I would assume at least some teams have a model that does at least something "similar" to this. I don't know why they wouldn't, it seems like a fairly common sense approach. I don't think anyone publicly produces such a model, although I believe I may start posting my results shortly after the draft (draft model will be the focus before the draft) or after summer league.

Crow
Posts: 6188
Joined: Thu Apr 14, 2011 11:10 pm

Re: Sketching a multi-year player value model

Post by Crow » Sat Jun 10, 2017 2:59 am

I was focused on player already in the league; but, sure, it could / should be done for rookies too.

Statman
Posts: 548
Joined: Fri Apr 15, 2011 5:29 pm
Location: Arlington, Texas
Contact:

Re: Sketching a multi-year player value model

Post by Statman » Sat Jun 10, 2017 3:17 am

Crow wrote:I was focused on player already in the league; but, sure, it could / should be done for rookies too.
The majority of the players in my projection model are NBA players. NBA data (regular season w/ it's 82 games & playoffs) is the majority of the data - G League, D1, & usually summer league data is obviously non existent for NBA vets. G, D1, summer, & exhibition are included to help find the hidden treasures - say a Robert Covington, Jonathon Simmons, Hassan Whiteside, etc. - and to quickly get more data blended into the draft model player projections beyond college production.

When I use the term "player projections" - that just means projecting future performance & expected rate of ascension or decline of that player year to year, whether it be a 19 year old rookie or a 32 year old vet. But I can completely understand why one would assume I'm talking about my draft model - which are rookies & still just uses (for now) D1 college data.

Post Reply