That is a good point. Maybe because Sexton has guys miss more shots when he sets him up than anyone else and Lauri (who has certainly improved on his passing) doesn’t have as much?
Let’s talk about NBA stats
- Thread starter NAOS_pt2
- Start date
It could be. Those are paywall stats afaik (I think at least Second Spectrum has them), but I think you could use them for all-around metrics if you didnt share the source data.
This is from the last 15 games. Sexton leads the team in assists and potential assists by a wide margin. Plus the difference between his potential assists and assists is highest on the team as well.
For every 11 shots a player takes from a direct pass from Sexton about only 6.3 go in.
It’s all shots that occur, so yeah those are included but not necessarily the bulk of the "luck".
It does explain it, those things that are considered "luck" have favored Lauri when he's on the court. How much it is actually luck versus player performance....that's a judgement call each individual will have to make.
This is an extreme example, but let's say when Sexton is on the court his teammates suddenly shoot 0% from the FT line when he's on the court and 100% from the line when he's off the court. That is something that will hurt his raw RAPM, but the LA-RAPM will adjust for that.
A normal play by play data point will look like this:
5man offensive lineup - 5man defensive lineup - result
The result will be the result of the play: 1 point scored, 2 points scored, 3 points scored etc.
For Luck Adjusted, they will essentially replace the result with an expected value.
5man offensive lineup - 5man defensive lineup - result
The result will be the result of the play: 1 point scored, 2 points scored, 3 points scored etc.
For Luck Adjusted, they will essentially replace the result with an expected value.
Yeah but over time that should adjust in relation to your own performance. If Lauri shoots 40% from 3 with the highest volume, and Sexton shoots 40% with the second highest volume, then the shooting of "others" is more favourable to bump Lauri's numbers as when his sample gets removed from team sample it lowers the team value more. Same goes for turnovers. Remove Lauris 1.2 and team averages more than if you remove Sextons 1.9.
That is actually what explains the Sexton bump vs Lauri bump, not your previous post.
So its essentially neglecting things like off ball impact and lineup compatability as well and categorizes that under luck. This explains it perfectily, since without Lauri we have trouble putting out working lineups, and those minutes are "bad rating minutes" that Sexton gets a lot more of. And its not Sextons fault, even though Lauri does deserve credit for making the team work better.
Uh…..not exactly. Btw, both posts say the same thing lol. In theory the luck adjusted and pure ramp would converge over time and get closer as the sample gets larger. What it really adjust for (or attempts to do) is take out the noise from make or miss.
That first part, no relevance. Second part….maybe when it comes to teammate compatibility…just not necessarily in the way you stated. Compatibility also affects raw rapm. Same goes for off ball impact.
For simplicity sake, let’s just ignore the rebounding and turnovers. A 40% shooter hoists up a 3 pointer. That means 40% of the time that will count as a 3 in the +\- column and 60% of the time it will be a zero. LA would remove the actual result and replace it with 1.2 for each of those possessions.
What do you mean by ignoring volume? It simply replaces the actual result of a possession with the expected value of that possession.
Can’t say for certain what luck adjustment formula Krishna used. But you can google a lot of the popular methods. Nathan Walker is the first guy who started doing luck adjusted stuff. He works for a team now though, so his stuff might have been taken down from the web. Here’s an article explaining LA methods in long form:
Nylon Calculus: Defining and calculating luck-adjusted ratings for the NBA
On-off data is the most basic data available for tracking how a team performs when a player is on or off, but it can be improved with luck-adjusted ratings.
fansided.com
Also btw you didnt understand my latter point.
If you replace the actual result with the average result, you are completely ignoring factors like "the quality of look" which is know to have massive impact on shooting success. So functioning well as a team and the classic "making others better" factor is balanced out, which essentially means off ball play and compatability.
Quality of the look is often considered in this LA adjustment methods. Like I said before, I can’t speak to what Krishna did specifically. The expected value has to be created somehow.
But regardless, the reason why LA became in a thing in the first place is because the things they adjust for, there is a statistical argument to show that it is actually random. I guess each person is free to determine if it’s actually random based on feel….but a lot of work has been done to investigate what is or isn’t random.
The thing about functioning better as team…that affects both the raw and the luck adjusted. Like if you put 5 non-shooting C’s on the court. Their plus minus would be cooked and not be representative of their true ability. Doesn’t matter if it’s raw or luck adjusted. I suppose lineup compatibility could be a potential explanation as to why certain times are lucky are unlucky….but there is also potential for a team to be lucky with bad lineup compatibility and unlucky with good lineup compatibility. There is some degree of randomness in every situation.
Just a question. I'm not well acquainted with most of the all around stats and asking silly questions to get grasp of what the number is about.What do you mean by ignoring volume? It simply replaces the actual result of a possession with the expected value of that possession.
Can’t say for certain what luck adjustment formula Krishna used. But you can google a lot of the popular methods. Nathan Walker is the first guy who started doing luck adjusted stuff. He works for a team now though, so his stuff might have been taken down from the web. Here’s an article explaining LA methods in long form:
Nylon Calculus: Defining and calculating luck-adjusted ratings for the NBA
On-off data is the most basic data available for tracking how a team performs when a player is on or off, but it can be improved with luck-adjusted ratings.
I thought one of the biggest critique towards APM was based on the fact that things like coaching and roster compatability arent really accounted for. Maybe the LA-RAPM actually sort of "tries" to address this as it doesnt punish player for the bad fits (which of course could be caused by the player in some instances). But if you balance out the impact of incompatability, you also stop rewarding compatability.
I think OG who is very impactful off the ball falling from 1 to 282 and Chet who is crucial to OKC lineup ifts falling from 18 to 307 sort of support that the LA-version can overcorrect pretty massively.
It is still a chief criticism, but that's not necessarily what LA tries to address. There is some degree of randomness in the make or miss game and that's what LA tries to address and you've got to start there. Having said that, I would agree that LA tends to overcorrect and there are valid reasons to think that there is more than "luck" that caused the actual values. For example, lets just say reality shows that Sexton's teammates shot 30% when LA (using whatever method) expected them to shoot 40%. I'm definitely open to the idea that LA overcorrects because the lineups Sexton plays in have bad synergy. But you also have to accept that no matter what the lineup, randomness is still present. A player can be in perfect lineups 100% of the time, but there's still a possibility his teammates miss more shots than they should due to chance.
I think you could argue LA does do a better job of addressing compatibility than raw, but it would be an indirect consequence that follows from the primary goal. This is why I said earlier that it's totally reasonable to have different feelings on LA and it's usefulness. I wouldn't hassle with anyone who felt strongly that one is better than the other, but personally I think it's good to consider both.