Why are Anthony Edwards impact metrics so poor relative to his hype?

[Fencer reregistered]

you don’t explain why something isn’t true

good bye forever

I'm not going to teach you basic math, friend. Perhaps take your own advice? The burden here is on you to explain why 16 games is a meaningful sample, especially in the context of visibly-unsustainable shooting percentages with the weight of nearly a century of basketball behind them (even factoring in modern trends).

16 games is an inherently noisy sample. It's the same line you hear every year at the beginning of the season when someone is championing something happening over the first 10 or 20 games of the season, before that eventually evens out. Same same with a single playoff run. It's something to be viewed with at least SOME skepticism until and unless it is repeated. And especially so when there are some eye-popping numbers so discordant with RS production (and league/personal averages).

In any case, take care.

it appears that you did not see the article i shared with you in the post you quoted, therefore let me give it to you again https://www.statisticshowto.com/large-enough-sample-condition/

That article is about chi-squared tests, which examine correlation to determine whether one can regard two variables as independent. Which two variables do you think we are or should be testing the independence of?

[Godymas]

I'm not going to teach you basic math, friend. Perhaps take your own advice? The burden here is on you to explain why 16 games is a meaningful sample, especially in the context of visibly-unsustainable shooting percentages with the weight of nearly a century of basketball behind them (even factoring in modern trends).

16 games is an inherently noisy sample. It's the same line you hear every year at the beginning of the season when someone is championing something happening over the first 10 or 20 games of the season, before that eventually evens out. Same same with a single playoff run. It's something to be viewed with at least SOME skepticism until and unless it is repeated. And especially so when there are some eye-popping numbers so discordant with RS production (and league/personal averages).

In any case, take care.

it appears that you did not see the article i shared with you in the post you quoted, therefore let me give it to you again https://www.statisticshowto.com/large-enough-sample-condition/

That article is about chi-squared tests, which examine correlation to determine whether one can regard two variables as independent. Which two variables do you think we are or should be testing the independence of?

the article is titled "Large Enough Sample Condition" and is about the "Large Enough Sample Condition"

the chi square is not what is being used here, just the large enough sample condition. the article details how you can apply the large enough sample condition in the event of wanting to use a chi square. that is why the subsection where the chi-square test is mentioned (which for some reason you skimmed to and believe is the main point of the article) is titled: "Chi Square and the Large Enough Sample Condition"

[Fencer reregistered]

it appears that you did not see the article i shared with you in the post you quoted, therefore let me give it to you again https://www.statisticshowto.com/large-enough-sample-condition/

That article is about chi-squared tests, which examine correlation to determine whether one can regard two variables as independent. Which two variables do you think we are or should be testing the independence of?

the article is titled "Large Enough Sample Condition" and is about the "Large Enough Sample Condition"

the chi square is not what is being used here, just the large enough sample condition. the article details how you can apply the large enough sample condition in the event of wanting to use a chi square. that is why the subsection where the chi-square test is mentioned (which for some reason you skimmed to and believe is the main point of the article) is titled: "Chi Square and the Large Enough Sample Condition"

Fair enough. I haven't taught statistics for decades, and it only was at the college level when I did. That particular phrase is more recent, perhaps something made up when statistics was introduced the high school curriculum.

But come on. What in basketball is this supposed to be relevant to? We know the outcome of every play Anthony Edwards has ever been involved in. Why would we be able to learn something special about that data set by sampling 16 plays or 16 games or whatever you have in mind?

[tsherkin]

Is a player going from fringe top 20 player in the regular season to literal BITW candidate in the playoffs a believable story?

Yeah. 16 games. Outlier shooting that he doesn't sustain at all in the RS. It's not really a surprise. It's probably an indicator that he's better than he has looked in the RS to some degree, and of course I have repetitiously noted that he has been trending upward even there. The point is patience, because we've seen this song and dance many times and it does not always end with the player realizing that PS/short sample performance over those larger samples. Like, he's never been better than 36.9% from 3 in the regular season over a very significant sample size (2,232 3PA), right? But he has shot 40%+ in two different postseasons (one ongoing), and on 9.5 and 7.8 3PA/g. Does that sound to you like revealed ability or a shooting streak? He's also been about 5.5% better from the foul line in the playoffs, and +4.5% from inside the arc. There's just stuff happening that makes you wonder a little about sustainability over a more stable sample size.

Even this year, he has two games of 50% shooting from 3pt range over a 5-game sample. He's shooting nearly 44% on 7.8 3PA/g. Do you expect that to maintain, or do you think it's heavily influenced by the the 7/13 and 4/8 games and that a different percentage will bear out as time progresses?


On a note for a broader audience, what mystifies me at times is why some are so dead-set against any kind of skepticism or pushback, even when it comes in a respectful tone which doesn't rule out his potential ability. That sort of mindset is for the team boards, not the GB. Like, great, defend your guy. But the amount of energy devoted to the barest hint of skepticism when there is much going into why that might be is a little frustrating to see.

[og15]

Regardless of the player, really high shooting is hard to maintain over longer samples, so when the sample size is small, we simply don't know yet, and that's in no way a discredit to the player, it's just life. Devin Booker shot 59/51 in the playoffs over 11 games last season for example, but we wouldn't expect him to continue that.

[dakomish23]

Because basketball is played on the court and not on an Excel spreadsheet.

Shout it from the **** rooftops

[Capn'O]

Sample sizes are weird to apply to young players. Sometimes you get a young guy who puts it together in the playoffs in a way that they hadn't before and that defines their career moving forward. SGA, for example, had a few games in the 2019 playoffs where he flashed a sort of dominance that became something he could do more consistently the very next season. Or Wade's first playoffs (rookie year at 22) where he showed out and then just exploded the next two seasons. Whereas, if an older player like Paul George goes nuts for a few games we still pretty much already know what we have with him. That's easier to assign to distribution models. There isn't such a benchmark to assign to Edwards as his career trajectory has thus far been a moving target with improvement each season both for production and efficiency. So can he shoot 50% from 3 forever? Probably not. Can he score 29+ ppg on 60%+ TS next season? Wouldn't be surprised at all.

Then again, sometimes a young guy just goes on a heater. But with Edwards you do see a skillset that could be functional at a superstar level. His athleticism but also functional athleticism (body control, reads, pacing, timing, etc) are just bananas. We'll see what we have here shortly.

[tsherkin]

Sample sizes are weird to apply to young players. Sometimes you get a young guy who puts it together in the playoffs in a way that they hadn't before and that defines their career moving forward.

It can certainly happen. And he's trended well in the playoffs. But it's a lot about hot shooting specifically, so it's at least worth examining and considering. Sample size burns people every year, season after season, decade after decade, so it's a relevant consideration even if some buck the trend.

So can he shoot 50% from 3 forever? Probably not. Can he score 29+ ppg on 60%+ TS next season? Wouldn't be surprised at all.

Sure, absolutely. He's pretty close to that now, that's "only" +2% TS. The odds of him shooting 44% from 3 on 7 3PA/g are lower, though, and that's what he's doing in this particular postseason so far, you know? It's basically what he did on almost 10 per game in 2022. It's elements like that which create the space for SOME degree of skepticism, like the 69% he's shooting from 16-23 feet in this postseason, you know what I mean?

[Klomp]

Face of the league. Superstar. Best SG in the league. Things I’m seeing a lot right now. He’s basically the league darling right now. People talk about him like he’s the next Jordan. You even have insane people trying to partake in revisionist history and saying he’s in the same tier this season as Luka and SGA.

But something I find odd is he’s not well liked by impact metrics. SGA and Luka for example are .250-.270 WS/48 players which is HOF/MVP tier. Ant is at .130, which is literally not even all star tier basically. Normally players at that level are high end starters, to low end all stars. .100 is a league average player, .150 area is typically an all star, .200 is a superstar, .250-.300 is usually a first ballot HOF/Multiple time MVP (guys like MJ, Jokic, Wilt, LeBron, KG etc).

EPM has SGA and Luka at #2 and #3 for example with 8.8 and 7.9. Edwards is at 4.2 which is ranked 22nd in the league after guys like FVV. Why is his hype/reputation so much better than his impact metrics?

Using the .200 WS as your barometer...Kobe Bryant hit that benchmark just once in his first 6 regular seasons and just once in his first 11 postseasons. Jordan hit .200 in the postseason for the first time in his fourth playoffs and jumped up to a .300 postseason in his 7th postseason.

Luka hit it just once in his first 4 years. And hasn't in the postseason yet.
SGA didn't hit it until his 5th season and this is his first time hitting it in the postseason at age 25 (almost 26).

Last postseason, Ant was sitting on the doorstep at .190. This postseason he has busted the door down at .309. At 22.

[Papi_swav]

Because analytics is not the end all be all, that's what you stat nerds fail to realize . The eye test holds more power and wins, always will .

[KembaWalker]

RealGM when you say something good about a player who isn’t over 60% TS in the regular season

[OkcSinceSGA]

Face of the league. Superstar. Best SG in the league. Things I’m seeing a lot right now. He’s basically the league darling right now. People talk about him like he’s the next Jordan. You even have insane people trying to partake in revisionist history and saying he’s in the same tier this season as Luka and SGA.

But something I find odd is he’s not well liked by impact metrics. SGA and Luka for example are .250-.270 WS/48 players which is HOF/MVP tier. Ant is at .130, which is literally not even all star tier basically. Normally players at that level are high end starters, to low end all stars. .100 is a league average player, .150 area is typically an all star, .200 is a superstar, .250-.300 is usually a first ballot HOF/Multiple time MVP (guys like MJ, Jokic, Wilt, LeBron, KG etc).

EPM has SGA and Luka at #2 and #3 for example with 8.8 and 7.9. Edwards is at 4.2 which is ranked 22nd in the league after guys like FVV. Why is his hype/reputation so much better than his impact metrics?

Using the .200 WS as your barometer...Kobe Bryant hit that benchmark just once in his first 6 regular seasons and just once in his first 11 postseasons. Jordan hit .200 in the postseason for the first time in his fourth playoffs and jumped up to a .300 postseason in his 7th postseason.

Luka hit it just once in his first 4 years. And hasn't in the postseason yet.
SGA didn't hit it until his 5th season and this is his first time hitting it in the postseason at age 25 (almost 26).

Last postseason, Ant was sitting on the doorstep at .190. This postseason he has busted the door down at .309. At 22.

If Ant’s 16 game postseason sample is his new normal, he’s a future MVP probably

[MrGoat]

I was curious about that myself. Why is his PER so terrible for a purported superstar?

Not currently in the playoffs where it’s amazing and superstar level, but regular season.

PER always discriminates against shooting guards who don't put up the rebounding numbers of a center and/or the assist numbers of a point guard especially if they're the type that sometimes just has to take bad shots. PER is largely just a glorified counting stats measure and I've always considered it a flawed metric, especially when evaluating 2 guards

D Wade had regular season PER stats of over 25, six times, and had one season over 30. Ant hasn't broke 20 yet in a single season.

Still largely comes down to counting stats, and Wade's PER will be inflated because those stats are compared to total team stats.

Ant's PER averaging 25, 5, and 5 on a team that averages 113 points a game was destined to be well lower than Wade's 30, 5, and 7 on a team that averaged 98 points per game even though Ant actually shot the ball more efficiently

[CDM_Stats]

Because analytics is not the end all be all, that's what you stat nerds fail to realize . The eye test holds more power and wins, always will .

Analytics is the combination of metrics and analysis

And if we were to compare metrics to most fans' eye tests, metrics would probably crush them. Analytics would definitely crush them. And almost all that are against metrics, and I'd even suggest that a lot who use them incorrectly in arguments, have no idea how they are formed other than a glossary definition provided by the site they are ripping them from

This is the equivalent of a bad driver crashing a car and bemoaning about how cars arent safe. Maybe you just arent a good driver?

BTW - metrics, and only metrics, drop the ball on scoring wings all the time. Because they are based on numbers they hardly have any context to them. Even the metrics at NBA.com are mostly trash (dont get me started on possession types).. but the public facing ones for wings have always been punitive. In Ant's case, he is the engine of their offense, on a team that skews heavily towards defense. Gobert, McDaniels, Conley, NAW, SloMo.. so he's putting in yeoman's work on offense. If you compare that to other teams that are more balanced across the board, or who skew offensive, then sure, the percentages may look similar

[KembaWalker]

The thing that I hate most about “he’s gonna regress to the mean just wait till the sample size increases” especially in the playoffs is like..no ****. And it’s barely even relevant. This isn’t flipping a coin 1000 times it’s two teams going at it. If Edwards is burning them from mid range they will have to make an adjustment with either more help or a different defender which has cascading effects all over the games. If they assign more resources to Edwards and his efficiency goes down , that’s not “regression to the mean” it’s pulling those resources away from his teammates and making the games easier for them (in theory) whether the coach and teammates can take advantage of this is not really in Edwards’ hands. You won’t get that from a spreadsheet though you have to watch the games and see what’s actually happening

[OkcSinceSGA]

I regret this thread man. He’s growing on me fast lol. He’s just a legit dawg. I can’t even be mad.

Read on Twitter

[KrAzY3]

Individual stats don't mean everything, it's as simple as that.

The most important statistic, the only that really matters, the reason you play the game is to win. Everything else can be manipulated or misleading.

Just look at a guy like Robert Horry. He was a winner. That's all there is to it, but his statistics were not nearly as good as a lot of players who never had the sort of positive influence he had on winning teams. In fact, there has been a lot of players who had far better statistics and were arguably a negative influence.

So, next time you want to look up statistics, start with the most meaningful one and then go from there...

[An Unbiased Fan]

Spreadsheets in shambles

[theonlyclutch]

PER always discriminates against shooting guards who don't put up the rebounding numbers of a center and/or the assist numbers of a point guard especially if they're the type that sometimes just has to take bad shots. PER is largely just a glorified counting stats measure and I've always considered it a flawed metric, especially when evaluating 2 guards

D Wade had regular season PER stats of over 25, six times, and had one season over 30. Ant hasn't broke 20 yet in a single season.

Still largely comes down to counting stats, and Wade's PER will be inflated because those stats are compared to total team stats.

Ant's PER averaging 25, 5, and 5 on a team that averages 113 points a game was destined to be well lower than Wade's 30, 5, and 7 on a team that averaged 98 points per game even though Ant actually shot the ball more efficiently

Wade scored more points in a slower paced team where the league scoring environment is significantly more adverse, of course his PER is higher, primarily because (from formula perspective), what Wade did is significantly more outlier than Edwards.

[NO-KG-AI]

Because analytics is not the end all be all, that's what you stat nerds fail to realize . The eye test holds more power and wins, always will .

But his playoff analytics are amazing lol. If he played this way in the regular season too, his analytics would be better.

If anything, it reinforces that these numbers are doing a good job of telling the story of how the guy is doing.