Been tinkering with an NBA Predictive Model

[Aike]

Similar methodology to my NCAA model.

Collected data from 2012 - 2022 to predict the 2023 playoffs. Did pretty well. Had Boston and Denver 1/2 going into the playoffs last year.

Lakers and Miami both outperformed expectations.

 

[StubbornPenny]

I could see this working better than an NCAA metric with an extremely random tournament. Usage rates and minutes are generally very stable as there aren’t near as many mismatches in talent or size.

 

[Aike]

I could see this working better than an NCAA metric with an extremely random tournament. Usage rates and minutes are generally very stable as there aren’t near as many mismatches in talent or size.

I think the toughest part might be dealing with veteran teams who get healthy right before playoff time.

 

[StubbornPenny]

I think the toughest part might be dealing with veteran teams who get healthy right before playoff time.

Yeah specific lineups will have to accounted for. Hell the lakers can go through 3 distinct teams in one year lol

 

[ukypug]

Aike, slightly changing subjects, but how was FAU ranked in the model before their tournament run last year? If you still have the info.

 

[Aike]

Aike, slightly changing subjects, but how was FAU ranked in the model before their tournament run last year? If you still have the info.

I’d have to go back and look. As I recall, they were somewhere around top 15 over the last month of the season.

 

[Aike]

Yeah specific lineups will have to accounted for. Hell the lakers can go through 3 distinct teams in one year lol

I think it’s a case of understanding what the tool can and can’t do. It can tell you who is most likely to advance mathematically, but you also need to use some common sense around injuries, etc.

 

[YardCat]

I think it’s a case of understanding what the tool can and can’t do. It can tell you who is most likely to advance mathematically, but you also need to use some common sense around injuries, etc.

Can’t predict Willis Reed moments

 

[kybassfan]

What I would be most interested in reading is your compare and contrast of the models with the college game, My own impression is that the games are vastly different. I’d be curious to see if that folds into the predictive models.

 

[Aike]

What I would be most interested in reading is your compare and contrast of the models with the college game, My own impression is that the games are vastly different. I’d be curious to see if that folds into the predictive models.

I’ll give you one tidbit. Things like blocked shots and steals don’t matter as much to NBA playoff success as they do to NCAA Tournament success.

My theory is that it has to do with the one and done nature vs. the “best of” format.

I think being disruptive in a one game setting can be useful - you get a team off balance and elimination pressure gets the best of them.

In the NBA, teams are better coached and better prepared to begin with. Then game to game adjustments are made. It’s difficult to maintain any kind of “gimmick” without being figured out.

 

[Aike]

Can’t predict Willis Reed moments

Can’t at any level.

 

[kybassfan]

I’ll give you one tidbit. Things like blocked shots and steals don’t matter as much to NBA playoff success as they do to NCAA Tournament success.

My theory is that it has to do with the one and done nature vs. the “best of” format.

I think being disruptive in a one game setting can be useful - you get a team off balance and elimination pressure gets the best of them.

In the NBA, teams are better coached and better prepared to begin with. Then game to game adjustments are made. It’s difficult to maintain any kind of “gimmick” without being figured out.

I think you are on to something in that line of reasoning. On a best of series, you can suck up an off night. In the one and out tournament, over a stretch of 6 games, your offense will be disrupted at least once. Without a formidable defense, you go home. Another aspect of this is the NBA, you have one thing, Basketball. In college you have classes and other obligations, plus your practice is limited. This reduces IQ for the game and skills development.

 

[BBUK]

Similar methodology to my NCAA model.

Collected data from 2012 - 2022 to predict the 2023 playoffs. Did pretty well. Had Boston and Denver 1/2 going into the playoffs last year.

Lakers and Miami both outperformed expectations.

Don't let the knuckleheads mess with you Aike. (You do some skilled work.)

Not meant for posts in this thread so far but I've seen some knucklehead posts in the past that were down right ignorant and cruel. (Ignorant because the knuckleheads had no idea of the reasoning behind predictive modeling.)

 

[KAB Cats]

I think it’s a case of understanding what the tool can and can’t do. It can tell you who is most likely to advance mathematically, but you also need to use some common sense around injuries, etc.

As a data nerd, I find this interesting.

Have played around mentally with this, not as far as implementation like you. But would be difficult on a macro level it seems for that reason. Would almost need to calculate game by game basis, aggregating stats on a player by player basis based on time in game.

 

[Aike]

As a data nerd, I find this interesting.

Have played around mentally with this, not as far as implementation like you. But would be difficult on a macro level it seems for that reason. Would almost need to calculate game by game basis, aggregating stats on a player by player basis based on time in game.

You might be able to improve a model that way, or you might end up boiling the ocean and not improving much.

Like, what happens when a player is out for 2 weeks and returns, vs. when the player stays healthy continually? Is he the same player before and after, and how do you account for the difference?

Generally, I’ve found that what a team does over the course of an entire season is the best predictor of how they’ll do in the postseason.

Part of that calculus is how healthy and consistent they were with lineups. And if they are consistent with lineups, it’s much less important to boil down to individual stats, etc.

 

[Aike]

Don't let the knuckleheads mess with you Aike. (You do some skilled work.)

Not meant for posts in this thread so far but I've seen some knucklehead posts in the past that were down right ignorant and cruel. (Ignorant because the knuckleheads had no idea of the reasoning behind predictive modeling.)

Appreciate it. Can be a little tiring at times for sure.

 

[CincyUKFan]

Similar methodology to my NCAA model.

Collected data from 2012 - 2022 to predict the 2023 playoffs. Did pretty well. Had Boston and Denver 1/2 going into the playoffs last year.

Lakers and Miami both outperformed expectations.

A bit strange and disconcerting that Miami ran through the playoffs last year without Herro, and now have performed better without him this season.