Do I have to buy the 2005-06 Pro Basketball Forecast?!

[Phil Plantier]

I'd like to learn more about PER.

Pro Basketball Reference has the raw formula. A couple folks try to approximate it. I'd like to dig into it and read the justification. But it seems like the only printed version of the rationale is one of the old Pro Basketball Forecasts. Is there some other way to find out? I'd like to support Hollinger's work, but not buy something that's 85% out-of-date.

 

[Lazy vs Crazy]

Hollinger explains it a bit here:
https://www.espn.com/nba/columns/story?columnist=hollinger_john&id=2850240

PER is pretty much regarded as an outdated stat that overemphasizes the value of bigs these days.

edit: the Wiki article is also a good overview, if you haven't read it yet https://en.wikipedia.org/wiki/Player_efficiency_rating

 

[Phil Plantier]

Hollinger explains it a bit here:
https://www.espn.com/nba/columns/story?columnist=hollinger_john&id=2850240

PER is pretty much regarded as an outdated stat that overemphasizes the value of bigs these days.

edit: the Wiki article is also a good overview, if you haven't read it yet https://en.wikipedia.org/wiki/Player_efficiency_rating

Thanks! The wiki is the best so far, for sure. It's got the best explanation of the formula I've seen.

I'd like to dig into Hollinger's reasoning for each component. That's the part I'd still like to examine.

Edit: If it's obsolete, and there's something better generally used, would love to hear more about it.

 

[RorschachsMask]

Thanks! The wiki is the best so far, for sure. It's got the best explanation of the formula I've seen.

I'd like to dig into Hollinger's reasoning for each component. That's the part I'd still like to examine.

Edit: If it's obsolete, and there's something better generally used, would love to hear more about it.

https://hoopshype.com/lists/advanced-stats-nba-real-plus-minus-rapm-win-shares-analytics/

One of our own posters is the mind behind DPM/DARKO.

 

[mcpickl]

I'd like to learn more about PER.

Pro Basketball Reference has the raw formula. A couple folks try to approximate it. I'd like to dig into it and read the justification. But it seems like the only printed version of the rationale is one of the old Pro Basketball Forecasts. Is there some other way to find out? I'd like to support Hollinger's work, but not buy something that's 85% out-of-date.

Definitely don't spend any money on this.

IMO it's a worthless stat.

I always use this example, Marcus Smart has a lower PER in every season of his career than Enes Freedom had in his worst season.

Even last season when they were on the same team, Marcus Smart as the Defensive Player of the Year posted a 13.67 PER. Bench warmer, before being traded to and waived by Houston, Enes Freedom posted a 15.03 PER.

There's just no value in it.

 

[Jimbodandy]

Agreed with the above. PER really has no value. If you wanted to study it from an academic POV, that might be interesting. But it's probably not a good use of your cash. It's 85% out of date for sure and really doesn't capture significant parts of the game. mcpickl's examples are perfect.

 

[InstaFace]

Definitely don't spend any money on this.

IMO it's a worthless stat.

I always use this example, Marcus Smart has a lower PER in every season of his career than Enes Freedom had in his worst season.

Even last season when they were on the same team, Marcus Smart as the Defensive Player of the Year posted a 13.67 PER. Bench warmer, before being traded to and waived by Houston, Enes Freedom posted a 15.03 PER.

There's just no value in it.

Isn't it quite upfront about saying that it only accounts for offensive efficiency as a rate stat? That defense is not included? A good job of doing that, when reasoned from first principles, seems like it could still be useful as long as it's on a mutually comparable scale (i.e. that a player with an 8 and a player with a 16 are going to be as effective as two players with 12s).

Either way, I think understanding the reasoning for how it was built would be helpful to building an understanding of how these metrics are designed - and Hollinger is one of the better thinkers in this department, like a poor man's Bill James for basketball, is he not?

 

[Eddie Jurak]

Isn't it quite upfront about saying that it only accounts for offensive efficiency as a rate stat? That defense is not included? A good job of doing that, when reasoned from first principles, seems like it could still be useful as long as it's on a mutually comparable scale (i.e. that a player with an 8 and a player with a 16 are going to be as effective as two players with 12s).

Either way, I think understanding the reasoning for how it was built would be helpful to building an understanding of how these metrics are designed - and Hollinger is one of the better thinkers in this department, like a poor man's Bill James for basketball, is he not?

Don’t some aspects that aren’t offense factor in? Rebounds, steals, blocks.

I think PER is a useful answer to the question “How efficiently does this player compile box score stats?” Obviously, compilation of box score stats is not the be all and end all of player value, and, thus, PER is not the be all and end all of statistical measures - there are much better ones.

 

[InstaFace]

Fair enough.

I piped up because I find the design of metrics to be very interesting, and could understand why Phil wanted to follow that thinking from beginning to end regardless of whether it was the best single expression of a player's value or not. Perhaps the question is, who has explicated their metric design the best? rather than who has made the best metric.

 

[wade boggs chicken dinner]

Here's the thread where Bowiac discussed what he was doing when he started DARKO: https://sonsofsamhorn.net/index.php?threads/introducing-name-tbd-nba-box-score-projection-project.28622/

 

[mcpickl]

Isn't it quite upfront about saying that it only accounts for offensive efficiency as a rate stat? That defense is not included? A good job of doing that, when reasoned from first principles, seems like it could still be useful as long as it's on a mutually comparable scale (i.e. that a player with an 8 and a player with a 16 are going to be as effective as two players with 12s).

Either way, I think understanding the reasoning for how it was built would be helpful to building an understanding of how these metrics are designed - and Hollinger is one of the better thinkers in this department, like a poor man's Bill James for basketball, is he not?

It is not, no.

 

[Jimbodandy]

Fair enough.

I piped up because I find the design of metrics to be very interesting, and could understand why Phil wanted to follow that thinking from beginning to end regardless of whether it was the best single expression of a player's value or not. Perhaps the question is, who has explicated their metric design the best? rather than who has made the best metric.

Agreed completely. Understanding the how and understanding the good and bads of his approach and its shortcomings--could be fun and maybe even useful academically. But I wouldn't pay for it, largely because we have so many others that are so much better. Imo, it's a deep dive on the usefulness of the abacus or leech therapy.

I have nothing against Hollinger. But based on what we know about the OP just from even considering an approach like this and diving into analysis, there's a non zero chance that PHIL, his own homegrown metric, turned out to be more useful in six months than PER.

 

[moondog80]

PER is more or less as good as you can do using only information that come from the box score. And it's actually pretty good. Outlier examples like the Marcus Smart one from above aside, it tracks pretty well with the generally accepted rankings of players -- in most cases, the guys at the top of the PER rankings are guys everyone agrees are stars. And "breakout" guys are often identifiable in PER before they start to get widespread recognition. It's also been surpassed by far better metrics. But, just because nobody needs a VCR anymore, that doesn't mean we still can't appreciate the VCR for its role in the progress chain.

 

[Devizier]

PER is more or less as good as you can do using only information that come from the box score.

I disagree; the Dean Oliver metric (ORtg) is probably a better “box score only” metric and it’s even older than PER. It’s also easy to see the logic in how he developed it. DRtg is a different story, but PER is no good on defense either.

Regardless, either is better than the old NBA “efficiency” metric.