[外絮] NBA全新推出的灌籃分數(其實有點不新了)

https://www.nba.com/news/what-is-the-dunk-score
https://www.nba.com/news/a-deep-dive-into-how-nba-dunk-score-is-calculated

https://reurl.cc/zpgq1p
https://reurl.cc/A6nMrE

NBA.com在10月底推出了全新的追蹤數據--Dunk Score

這個數據綜合了25個球員追蹤的原始數據來統計出灌籃的難易度

而官網也會在數據頁面的右上角每天更新灌籃分數最高的幾球

https://reurl.cc/b34Wvr

這個是今天最高的Jaden Hardy--101.7分

而上次Jalen Green灌在Cade Cunningham頭上的那球則是本季第二高的111.3分

至於NBA用了哪些數據來當作灌籃的標準呢?

以下開始翻全文:

Dunks provide some of the most electrifying moments in basketball. Whether it’s a gravity-defying leap, an explosive finish through contact, or a
windmill slam that ignites the crowd, there’s nothing quite like a dunk to
get fans out of their seats. But what if there was a way to quantify the
magic behind each slam? With the help of advanced optical tracking systems
and cutting-edge AI, that’s exactly what the NBA Dunk Score sets out to do.

The NBA Dunk Score breaks down every dunk using data-driven metrics resultingin an objective score based on technical excellence, athleticism and
difficulty — without any bias derived from game context or the players
involved.

Ready to dive into how this works? Let’s explore the NBA Dunk Score.

灌籃提供了籃球最具動能的一些時刻。無論是無視重力的飛、呼嘯過對抗的終結還是點燃群眾的大車輪灌籃，沒有任何東西可以比灌籃能讓粉絲們離開他們的座位歡呼。但我們要怎麼數據化灌籃背後的魔法呢?在一些進階追蹤系統的幫助與劃時代的AI，這就是NBA灌籃分數準備要做到的事情。

NBA灌籃分數會利用以數據為基底的標準來分析每一個灌籃。而這結果將會是一個基於技術高度、體能與難度所總結出來的客觀分數--而這些不會因為灌籃參與的球員或是比賽的情況而有所偏差。

準備好要深入探討這如何運作了嗎?讓我們一起探索NBA灌籃分數。

Data Overview
The foundation of any model is the data that goes into it. That’s where the
NBA’s optical player tracking systems come in. Using advanced 3D pose
detection models, these systems track 29 points on each player’s body with
sub-centimeter accuracy, generating 3D coordinates for every movement. All ofthis is captured 60 times per second.

所有模型的基礎都是他被餵進去的資料。這也是NBA的球員光學追蹤系統參與的地方。(還記得NBA前幾年新增了鷹眼系統? #1a2VrOV6 (NBA) 這個可能就是應用之一)運用3D姿勢偵測的模型，這些系統追蹤每一位球員身上29個點直至公分以下的準確度並藉此產生每一個動作的3D座標。而這些東西每秒鐘會捕捉60次。

Dunk Score Overview
The NBA Dunk Score is a live model that analyzes over 25 different attributesof a dunk—each calculated directly from the player tracking data—to assign
an objective score. The system is entirely data-driven, meaning the score is
independent of external factors like the dunker’s identity, the game’s
score, or whether it’s a preseason or playoff game. This makes the Dunk
Score a true “technical score,” focused solely on the execution of the dunk
itself.

NBA灌籃分數是個分析灌籃現時模型。他分析一個灌籃超過25個不同的元素，而這些元素都是完全自球員追蹤數據產生，並最終給出一個客觀的分數。這個系統完全是數據驅動的，這也意味著這些分數獨立於灌籃者的身分、分數或者是比賽性質--無論這場比賽是個熱身賽還是季後賽。這使得灌籃分數完全是個"技術分數"，只專注於灌籃的執行本身。

The Dunk Score is broken down into four distinct subscores: Jump, Power,
Style, and Defensive Contest —each highlighting a different aspect of the
dunk. The Jump subscore is all about the athlete’s verticality and
explosiveness in the air. Power captures the raw force of the dunk. Style
measures dunk flashiness, quantifying how much creativity the player adds to
their dunk. Finally, the Defensive Contest subscore adds context by
evaluating how much defensive pressure the dunker faced. Together, these foursubscores provide a deeper, more nuanced breakdown of every dunk. Rather thanjust focusing on the overall score, they give us insight into the different
dimensions of the play.

灌籃分數可以在往下細分成4個分開的小分:跳躍、力量、風格與防守者挑戰--這其中每一項分別會關注灌籃中不同的面相。跳躍分數全是關於運動員跳得多高以及他在空中的爆發力。力量捕捉了灌籃的原始力量。風格評估了灌籃的華麗程度並數據化球員在灌籃中增添的創意。最後，防守者挑戰則提供灌籃的背景，評估灌籃者所面對道的防守壓力。而這4個全部加起來共同為每次灌籃提供了更深入且細膩的解析。不僅僅著眼於總分，更讓我們深入了解這項動作的不同面相。

Base Features Overview
Let’s take a closer look at the key attributes that factor into the Dunk
Score. These metrics, pulled directly from the 3D pose data via a real-time
processing pipeline, are the building blocks of the model. Some features
carry more weight than others, but each one helps break down the technical
aspects of a dunk:

以下是一些灌籃分數的主要屬性詳細解析。這些都是直接提取自透過實時處理管道蒐集的3D姿勢數據並構成了模型的基礎。有些特徵的權重比其他人更高，但每一個都幫助我們分析灌籃的技術面:

Player Vertical: Based on the mid-hip height when the player’s jumping foot
leaves the ground versus their peak mid-hip height during the jump.

球員彈跳高度:根據球員起跳時單腳支撐腳離地的髖部中點高度，與起跳過程中達到的髖部中點最高高度差值計算。

Takeoff Distance: The distance from the hoop at the point of takeoff. If the
player jumps from one foot, it’s measured from the jumping foot’s big toe.
If the player jumps from two feet, it’s measured from the midpoint between
the big toes.

起跳距離:從起跳點到籃筐的距離。如果球員單腳起跳，測量自起跳腳的大拇趾；如果雙腳起跳，測量腳大拇趾中點。

Hang Time: How long the dunker is airborne.

滯空時間:灌籃者在空中的總時間。

Maximum Ball Height: The peak height of the center of the ball during the
dunk.

最高球高度:灌籃過程中籃球中心的最高高度。

Reach Back Distance: The furthest distance between the player’s head and the
ball as it is pulled back from the hoop.

回拉距離:球員頭部與籃球在向後拉時的最大距離。

Ball Speed Through the Rim: How fast the ball is moving as it passes through
the rim.

球速:籃球穿過籃筐時的移動速度。

Total Ball Acceleration: The total acceleration of the ball toward the hoop,
accounting for the force the player applies during the dunk.

球的總加速度:籃球向籃筐移動的總加速度，包含球員灌籃過程中施加的力量。

Ball Movement: The distance the ball travels over the course of the dunk,
minus the takeoff distance. This captures many flashy motions often performedduring dunks: Windmill, Double Clutch, Behind the back, etc.

球移動距離:籃球在灌籃過程中的總移動距離，減去起跳距離。此特徵捕捉了灌籃中的花式動作，如大車輪、拉桿、背後換手等。

Not all features are continuous — some are boolean (true or false) features
that focus on specific stylistic elements, including:

而不是所有特徵都是連續性的--有一些是布林(T&F)特徵，而這些特徵專注於一些特定的風格元素，而這其中包含了:

Reverse Dunk 倒灌

360 Dunk 360度灌籃

Through the Legs 胯下換手

Alley-oop and Self-Oop: The score is adjusted based on the length of the
pass, where the ball is caught, and whether the player catches the ball with
one or two hands.

空中接力與自拋自灌:根據傳球距離、接球位置，以及球員是單手還是雙手接球調整分數。

Tip Dunk: Scaled higher based on how aggressively the player contests for therebound versus simply being in the right place for an easy tip-in.

補灌:根據球員爭搶籃板的激烈程度提高分數，而非僅僅因站在合適位置完成補灌。

The Impact of Defense
Defensive factors are treated as unweighted bonus points beyond the base
score. This way, the lack of a defender doesn’t penalize a dunk, but great
defense can elevate its difficulty and the overall score. Multiple defenders'contributions are additive, but each defender’s influence diminishes with
every additional body in the mix. Key defensive features include:

防守因子則被作為基礎分數以外(灌籃本身)額外的加分。這樣的話，即使前方沒有防守者，灌籃者也不會因此被扣分，但好的防守可以因為增加其難度而使其獲得總分的增加。多名防守者的貢獻是可以疊加的，但每多一位防守者，個別防守者的貢獻將減少。以下是主要的防守特徵：

Defensive Contest Level: A combined measure of (1) how close the defender’s
body is to the dunker and how directly they are positioned between the dunkerand the hoop, and (2) how close the defender’s hand is to the ball, with
slight weighting towards the ball-hand-rim angle. If the defender is under orbeyond the hoop, they are penalized.

防守干擾等級:綜合評估(1)防守者與灌籃者身體的距離及其在灌籃者和籃筐之間的直接位置，以及(2)防守者的手與籃球的距離，並在手-球-籃筐角度上增加權重。如果防守者站在籃筐下方或更遠的位置，他們將被扣分。

Alignment Score: Measures how directly the dunker and defender are facing
each other, using dot products of their direction vectors and scaled by
distance. A chest-to-chest contest would result in a perfect 1.0 alignment
score.

對位分數:測量灌籃者和防守者的直接對位程度，通過他們方向向量的點積計算，並按距離縮放。面對面的防守會產生完美的1.0對位得分。

Collision Score: Measures the intensity of contact between the dunker and thedefender. It evaluates the sum of their velocity components toward each
other, scaled inversely by the distance between them. Two players in very
close proximity moving directly at each other at high speeds would have a
high collision score.

碰撞分數:衡量灌籃者與防守者之間接觸的強度。此分數通過兩人相向移動速度分量的總和計算，並按距離的反比縮放。兩位球員非常接近且高速向對方移動時，將得到高碰撞分數。

The model also detects situations where a defender attempts to take a charge
or when the dunker jumps over the defender. In these cases, special logic is
used to focus on 2D body-based features rather than hand-based features.

模型同時還可以偵測防守者是試圖製造進攻犯規還是灌籃者真的飛過了防守者。在這些情況會採用特殊的邏輯更加重視基於身體的2D特徵，而非手部特徵。

Overall Scoring Distributions
Below is a histogram representing the overall distribution of the Dunk Score
for all dunks from the 2023-24 NBA season. Most dunks land somewhere in the
20-50 range. It’s a right-skewed distribution, meaning only a select few
make it into the upper echelon.

底下是2023-24 NBA 球季所有灌籃的灌籃得分分布直方圖。大多數灌籃的得分範圍介於
20 到 50 分之間。這是一個右偏分布，意味著只有少數灌籃進入了頂尖行列。

https://i.imgur.com/tL4Re45.jpeg

NBA灌籃分數系統將於2024-25賽季Beta版正式推出。這是一種突破性的分析方式，專注於籃球運動中最令人興奮的表現之一。通過分析球員動作、防守壓力以及飛行物理，灌籃現在可以被以客觀的精確性進行比較。

--

歐肥的分數在這個評分機制下還真的不高

我猜會 但是評審作為噱頭之一應該還是評分的主要

他其他Metric應該不會太醜 但肯定是很難勝過AE那種體能怪物

今天最接近O'Neal大部分型態的可能就這球 只有50分左右

現在還在測試階段 可能暫時只會有當天的吧 但之後應該有機會可以看本季以後的