Dean Oliver at ESPN introduces Net Points

[Mike G]

Using a metric that is not 'points above or below NBA avg' -- along the lines of WS/48, in which zero=zero -- would you say a .075 player has Half the positive impact that a .150 player has?

[Crow]

Half the rating implies half the measured / estimated impact. "Actual" impact can vary from measured and from subjective impressions.

[Mike G]

If a 24-hour period has 11 hours of daylight and 13 hours of darkness, is that a net -2 hours of light?

[Crow]

Use the reference point and adjective you want and not what you don't want... and others may vary.

[Crow]

Not seeing factor detail by player tonight. Did that ever show? I don't recall. Team level there.

I'd prefer to see factor detail, though it could be considered "too much".

[DSMok1]

Using a metric that is not 'points above or below NBA avg' -- along the lines of WS/48, in which zero=zero -- would you say a .075 player has Half the positive impact that a .150 player has?

There is no baseline that is non-negotiable besides NBA average. Point differential does translate to wins sort of linearly near league average, but not at the extremes, and many players are have too large an impact for the nonlinearity to be ignored.

[Mike G]

OK so using my example the .075 WS/48 players -- Nembhard, Markkanen, Beal -- are #135-137 of 182 qualifying.
In BPM the #136 guy is -1.6.

Does it seem counterintuitive to say those players are 75% as productive/efficient (proficient?) as NBA average?
Or does it feel right to say they do more negative than positive inputs when they play?

Tangentially, I was wondering how the 2020 bubble season BPM are worked out. The better teams played more games than the 8 weakest. The range was 64 games to 75.

Bubble teams played 8 games against mostly playoff-bound teams. So 'NBA avg' may be the season avg 100 possessions or avg team, and either one is actually negotiable?

[DSMok1]

OK so using my example the .075 WS/48 players -- Nembhard, Markkanen, Beal -- are #135-137 of 182 qualifying.
In BPM the #136 guy is -1.6.

Does it seem counterintuitive to say those players are 75% as productive/efficient (proficient?) as NBA average?
Or does it feel right to say they do more negative than positive inputs when they play?

Tangentially, I was wondering how the 2020 bubble season BPM are worked out. The better teams played more games than the 8 weakest. The range was 64 games to 75.

Bubble teams played 8 games against mostly playoff-bound teams. So 'NBA avg' may be the season avg 100 possessions or avg team, and either one is actually negotiable?

I don't think % as productive makes sense to discuss. It's just a continuum of a player being generally better than another player (or at least better in their current role/context). I don't even think it makes sense to say that a -1.0 player is a negative... just that the minutes-weighted average NBA player is a bit better than they are. Generally a player is still better than the player behind them in the rotation (if we had a better player for that role, we'd play them!), unless weaker players that are young are being played for development purposes.

[Mike G]

So I've been tempted to say something along the lines of: I doubt that anyone making a system in which zero = average would claim that such a player is net zero in influence, and anyone less is a net negative.

eWins uses 1.0 to mean average, and 2.0 means 2x avg, etc. It feels a lot more intuitive to casually say one player is twice as good (impactful, proficient) as another and be close to agreeable to most fans, players, coaches, GM's, and owners.
I've seen casual comments that indicate a 1.0 BPM player is only half as good as when he was 2.0; this makes sense like saying 20*F is twice as cold as 40* maybe?

Now VORP declares another level to be zero, and it's negotiable; but there's some motivation to establish this baseline. And we can say player A had twice as much as player B. Even if there are no units to VORP (are there?), it's conversational.

I see 149 players with decent minutes at or above -2.0 BPM, just 5 per team. Add some available players with fewer minutes, and I'm guessing an avg of ~2 are unavailable on an average night, and the avg team still only has about 5 above-zero VORP generators per game.

[DSMok1]

....
I've seen casual comments that indicate a 1.0 BPM player is only half as good as when he was 2.0; this makes sense like saying 20*F is twice as cold as 40* maybe?

Now VORP declares another level to be zero, and it's negotiable; but there's some motivation to establish this baseline. And we can say player A had twice as much as player B. Even if there are no units to VORP (are there?), it's conversational.

I see 149 players with decent minutes at or above -2.0 BPM, just 5 per team. Add some available players with fewer minutes, and I'm guessing an avg of ~2 are unavailable on an average night, and the avg team still only has about 5 above-zero VORP generators per game.

The temperature analogy is a good one. 40° is obviously not twice as hot as 20°. It's an arbitrary scale.

VORP has the units of point differential per 100 TEAM possessions above the theoretical -2 replacement player.

More recently I've recognized that replacement level is actually lower, more like -3 or -3.5.

[Crow]

Sent on X

@DeanO_Lytics

Are there or will there be pages that give team level "net points" by factor for season on a continuously updated basis? And splits by month or custom timeperiod searches?

Player by factor for season?


Not seeing either at this time. The factor level detail is the main draw for me.


Recent reply is:

"Appreciate the suggestions."

[Crow]

70 players rated as worth 5+ WAR. 13 of them over 10.

393 with positive WAR. 172 with negative WAR.

Only 107 positive on Net Points per 100p.

There is estimated positive and estimated positive compared to replacement players. The former group is only about 1/4th the size of the latter.

4 guys per team on average you want to play. 9 more to fill things out, over the 5 guys worse than them.

[Crow]

https://espnanalytics.com/nba-daily-summary

Yesterday's winners won an average of 2.75 of the 4 net factors listed (FTs apparently blended into 2pt and 3pt activity).

Each winner had a +10 or bigger factor win, Knicks had two. 2 winners survived a pretty big factor loss, Knicks had two.

5 of 6 best individual ratings yesterday for players 6-9 and taller. Each of the winners had 1, TWolves had 2. No loser had any.

4 guys who scored 30+pts had: 1 good but not top net points rating, 2 mild, 1 negative. 1 winner, 3 losers. The winner, Brunson, was the negative and was carried by KAT and Cam Payne. Not a particularly good day for 30+ scorers.


ESPN not currently set up to show individuals by factor in regular season or any cumulative playoff data (though we aren't into multiple games yet).

LeBron James, 4th worst on net points yesterday. Exceeded by 2 teammates. Doncic a very mild positive.

Kuzma, by far the worst.

3 of the 4 net points wins by TWolves were expected by regular season factor data, though 2 of the 4 were close. They didn't need the huge 3pt factor win to win yesterday but maybe later. Was biggest factor win of day.

Nuggets were expected to win 3 factors by regular season data. They won 2 and it was barely enough.

The 5 biggest factor wins were 2 from net 3pt shooting, 2 from net turnovers and 1 from net 2pt shooting.

None from rebounding, which broke 2W-2L for the winners. But "Denver had NINE OREB in the last 9 mins."