Alts to draft models predicting performance in year 3-5

Home for all your discussion of basketball statistical analysis.
Post Reply
Crow
Posts: 6206
Joined: Thu Apr 14, 2011 11:10 pm

Alts to draft models predicting performance in year 3-5

Post by Crow » Thu Jun 29, 2017 8:44 pm

Most draft model tends to be designed to predict player performance, box score performance, in years 3-5. I recently saw Dean Demakis project total career player value. Other options?

1. Immediate performance
2. Rookie contract performance / cost ratio
3. Second contract performance / cost ratio
4. Career performance / cost ratio
5-8. All these but based on RPM.
9. Trade level after year 3 or year 5-6. "Worth" and market expectations considering other factors and historical patterns. Players are production pieces AND "Assets". The two are related but may not be identical based on position, age, excitement / revenue perspective, etc.

What other options have a rationale?

Any draft model gurus produce or want to
produce any of these?

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

Re: Alts to draft models predicting performance in year 3-5

Post by Statman » Thu Jul 13, 2017 4:40 pm

Crow wrote:Most draft model tends to be designed to predict player performance, box score performance, in years 3-5. I recently saw Dean Demakis project total career player value. Other options?

1. Immediate performance
2. Rookie contract performance / cost ratio
3. Second contract performance / cost ratio
4. Career performance / cost ratio
5-8. All these but based on RPM.
9. Trade level after year 3 or year 5-6. "Worth" and market expectations considering other factors and historical patterns. Players are production pieces AND "Assets". The two are related but may not be identical based on position, age, excitement / revenue perspective, etc.

What other options have a rationale?

Any draft model gurus produce or want to
produce any of these?
I do college to NBA rating projections year by year, compile for career (assumed average team & pace, based on latest NBA season rating breakdowns). I have integrated separate offensive & defensive "potential" factors that appear to mimic RPM a bit more (not by design, just how it worked), & it gave me the best predictive correlations I've ever hit in the 20 year retrodiction by a decent margin. Quite a bit better predictive than actual draft position, WAY better past lottery & 1st round. Model seems to adjust to the current game (stretch bigs, versatile wings, 3 & d, etc.) really well now, something that was a bit of an issue to me a couple years ago.

Cost ratios & the like are all doable with a model like mine (& I assume some others) that project year by year career overall &
type of impact (does a guy project to a max player & by what year. Etc.). Now, of course, player projections are fluid, so as soon as summer league, exhibition, & obviously season data (NBA & G-L) starts getting included in the projections of players (especially rookies) - things change somewhat.

As for this type of modeling, it is MUCH more work (which is why I don't think it's done much, at least not publicly) & testing (although, that's more my choice), but I just feel the scope of results offer much more flexibility of analysis.

Post Reply