The Math Behind Basketball's Wildest Moves | Rajiv Maheswaran | TED Talks

翻译人员: Li Li 校对人员: Jing Peng

My colleagues and I are fascinated by the science of moving dots.

我和我的同事对移动圆点 背后的科学非常着迷。

So what are these dots?

那么这些小圆点是什么呢？

Well, it's all of us.

就是我们自己。

And we're moving in our homes, in our offices, as we shop and travel

我们在家里，办公室里来回走动，

throughout our cities and around the world.

也在世界各地旅行和购物。

And wouldn't it be great if we could understand all this movement?

如果我们能弄清这些移动，

If we could find patterns and meaning and insight in it.

并从中发现规律，意义并提出见解， 不是一件很棒的事吗？

And luckily for us, we live in a time

很幸运的是， 我们生活在这么一个时代，

where we're incredibly good at capturing information about ourselves.

我们非常擅长捕捉关于自身的信息。

So whether it's through sensors or videos, or apps,

不管是通过传感器，视频，或软件应用，

we can track our movement with incredibly fine detail.

我们都能详尽地追踪到个人移动的轨迹。

So it turns out one of the places where we have the best data about movement

这就让我们发现， 最佳的数据来源之一

is sports.

就是体育运动。

So whether it's basketball or baseball, or football or the other football,

因此无论是篮球、棒球、橄榄球或足球，

我们都可以在场馆内， 甚至运动员身上装上设备来追踪

we're instrumenting our stadiums and our players to track their movements

every fraction of a second.

他们每个时刻的运动数据。

So what we're doing is turning our athletes into --

所以我们要做的 ——你们大概已经猜到了——

you probably guessed it --

就是把运动员的移动

moving dots.

转化成圆点的移动。

So we've got mountains of moving dots and like most raw data,

所以我们收集了不计其数的移动小圆点，

it's hard to deal with and not that interesting.

就像多数原始数据一样， 难以处理，也没什么趣味。

But there are things that, for example, basketball coaches want to know.

但数据里面蕴藏着， 比如篮球教练想知道的事情。

And the problem is they can't know them because they'd have to watch every second

但问题是，除非教练们把每场比赛里 每一秒数据都记下来再去思考，

of every game, remember it and process it.

否则他们没法从中得到想要的信息。

And a person can't do that,

人类大脑无法做到这件事，

but a machine can.

但机器没问题。

The problem is a machine can't see the game with the eye of a coach.

然而，机器没办法自己 以教练的视角去看一场比赛。

At least they couldn't until now.

直到现在，我们做到了。

So what have we taught the machine to see?

那么， 我们让机器去观察些什么呢？

So, we started simply.

先从简单的开始。

We taught it things like passes, shots and rebounds.

我们先教会它传球、投篮和篮板球，

Things that most casual fans would know.

这类普通球迷也知道的事。

And then we moved on to things slightly more complicated.

然后我们开始教它一些 稍复杂点的事情，

Events like post-ups, and pick-and-rolls, and isolations.

比如落位背打、挡拆和拉开单打。

And if you don't know them, that's okay. Most casual players probably do.

你们如果不了解这些名词， 没关系。打球的人大都了如指掌。

Now, we've gotten to a point where today, the machine understands complex events

迄今为止，我们已经能够让机器理解

like down screens and wide pins.

下掩护和无球掩护这类复杂的，

Basically things only professionals know.

只有专业人士才懂的战术。

So we have taught a machine to see with the eyes of a coach.

于是我们已经教会电脑用 教练的视角去观察数据了。

So how have we been able to do this?

我们是怎么做到的呢？

If I asked a coach to describe something like a pick-and-roll,

如果我让一个教练讲解挡拆，

they would give me a description,

我会得到一个定义，

and if I encoded that as an algorithm, it would be terrible.

如果我把这个定义编码成一个算法 估计会惨不忍睹。

The pick-and-roll happens to be this dance in basketball between four players,

挡拆就是四个球员之间的舞蹈，

two on offense and two on defense.

两人进攻，两人防守。

And here's kind of how it goes.

大概是这么个过程：

So there's the guy on offense without the ball

一个没有带球的进攻球员

the ball and he goes next to the guy guarding the guy with the ball,

跑向持球的防守队员，

and he kind of stays there

站在那里待一会儿，

and they both move and stuff happens, and ta-da, it's a pick-and-roll.

然后他们一起移动（制造机会）， 嗒哒，这就是挡拆。

(Laughter)

（笑声）

So that is also an example of a terrible algorithm.

这也是个糟糕的算法实例。

So, if the player who's the interferer -- he's called the screener --

如果那个干扰的球员—— 或者叫掩护者——

goes close by, but he doesn't stop,

只是跑过来干扰一下而不停下，

it's probably not a pick-and-roll.

这可能就不是挡拆了。

Or if he does stop, but he doesn't stop close enough,

就算他停下来， 但停的位置不够接近，

it's probably not a pick-and-roll.

那也不算是挡拆。

Or, if he does go close by and he does stop

或者，就算他足够近，而且停下来，

but they do it under the basket, it's probably not a pick-and-roll.

但他是在篮下完成的 那也不算挡拆。

Or I could be wrong, they could all be pick-and-rolls.

或者我可能错了， 这些都是挡拆。

是否是挡拆要根据发生的时间、 球员间距、位置而定，

It really depends on the exact timing, the distances, the locations,

and that's what makes it hard.

这些都很难去界定。

So, luckily, with machine learning, we can go beyond our own ability

幸运的是，有了机器学习技术， 我们就能超越自身的能力

to describe the things we know.

来描述我们已知的事物。

So how does this work? Well, it's by example.

这个技术要如何实现呢？ 举个例子：

So we go to the machine and say, "Good morning, machine.

我们对机器说， “早上好，机器。

Here are some pick-and-rolls, and here are some things that are not.

这儿有些挡拆例子，还有一些不是。

Please find a way to tell the difference."

你来找出不同点吧。”

And the key to all of this is to find features that enable it to separate.

这其中的关键是电脑能找出 区别两者的特征来。

So if I was going to teach it the difference

所以如果我要教会机器

between an apple and orange,

辨别苹果和橘子，

我可能会说： “不妨用颜色和形状来区分吧？”

I might say, "Why don't you use color or shape?"

And the problem that we're solving is, what are those things?

而目前要解决的问题就是， 要区分事物的特征是什么？

What are the key features

电脑需要掌握的整个

that let a computer navigate the world of moving dots?

移动圆点世界的关键特征是什么？

So figuring out all these relationships with relative and absolute location,

搞清楚所有这些相对位置、 绝对位置、距离、时机、

distance, timing, velocities --

速率之间的关系——

that's really the key to the science of moving dots, or as we like to call it,

就是移动圆点科学的真正关键所在， 换成专业术语，

spatiotemporal pattern recognition, in academic vernacular.

我们喜欢称之为：时空模式识别。

Because the first thing is, you have to make it sound hard --

因为首先，你要让它听起来 很难懂，很专业——

because it is.

因为事实的确如此。

The key thing is, for NBA coaches, it's not that they want to know

对于NBA教练们来说，判断是否是

whether a pick-and-roll happened or not.

挡拆并不是关键，

It's that they want to know how it happened.

而这个挡拆是怎么发生的 才是他们关注的。

And why is it so important to them? So here's a little insight.

为何教练们如此关心这一点？ 这儿我要解释一下。

It turns out in modern basketball,

在现代的篮球比赛中，

this pick-and-roll is perhaps the most important play.

挡拆几乎是最重要的战术。

And knowing how to run it, and knowing how to defend it,

了解如何使用以及怎样防守挡拆，

is basically a key to winning and losing most games.

基本上是比赛输赢的关键。

So it turns out that this dance has a great many variations

因此挡拆的步伐多种多样，

and identifying the variations is really the thing that matters,

能够识别这些不同的形式 是非常重要的，

and that's why we need this to be really, really good.

这就是为什么我们对 机器的智能性要求相当高。

举个例子。

So, here's an example.

There are two offensive and two defensive players,

这儿有两个进攻队员和 两个防守队员，

他们准备开始实施挡拆。

getting ready to do the pick-and-roll dance.

那么持球人既可以选择利用挡拆， 也可以放弃挡拆，

So the guy with ball can either take, or he can reject.

His teammate can either roll or pop.

他的队友可以拆向篮下， 或撤到一个无人盯防的空位。

The guy guarding the ball can either go over or under.

防守持球者的人可以上前绕过掩护， 或者从后方绕过掩护。

His teammate can either show or play up to touch, or play soft

而他的队友则可以探出补防，或保持 近距离防守，亦或者向后消极防守。

and together they can either switch or blitz

两个防守球员也可以换防，或者包夹。

and I didn't know most of these things when I started

一开始的时候我也不是很懂这些，

and it would be lovely if everybody moved according to those arrows.

如果每个人都能沿着箭头方向移动， 事情就好办多了。

It would make our lives a lot easier, but it turns out movement is very messy.

这会让我们的工作简单很多， 但往往这些移动非常杂乱。

球场上会发生很多突然的变动， 要在查准率和查全率方面

People wiggle a lot and getting these variations identified

准确识别这些变化

with very high accuracy,

both in precision and recall, is tough

是相当困难的，

because that's what it takes to get a professional coach to believe in you.

但只有这样， 才能让专业教练相信你的技术。

And despite all the difficulties with the right spatiotemporal features

尽管在准确的时空特性识别上 困难重重，

we have been able to do that.

我们还是成功地做到了。

Coaches trust our ability of our machine to identify these variations.

教练相信我们的机器 能够识别这些变化。

We're at the point where almost every single contender

目前，我们已经推出了 相关的识别软件，几乎每个

觊觎今年NBA总冠军的球队，

for an NBA championship this year

is using our software, which is built on a machine that understands

都在使用我们的这款软件， 其功能就是通过机器

the moving dots of basketball.

识别篮球领域的移动。

So not only that, we have given advice that has changed strategies

不仅如此， 我们还对如何改善战术提供建议，

并帮助球队赢得过重要的比赛。

that have helped teams win very important games,

and it's very exciting because you have coaches who've been in the league

能够让联盟中执教30年的 老教练愿意听取

机器提供的意见，这太让人激动了。

for 30 years that are willing to take advice from a machine.

And it's very exciting, it's much more than the pick-and-roll.

不仅仅局限于挡拆， 更让我们兴奋的是

Our computer started out with simple things

我们让电脑从简单的事情着手，

and learned more and more complex things

逐渐学会了更复杂的事物，

and now it knows so many things.

如今它已经掌握了丰富的知识。

Frankly, I don't understand much of what it does,

老实说，我不大明白它是怎么做到的，

and while it's not that special to be smarter than me,

不过就算比我聪明也没什么特别的，

we were wondering, can a machine know more than a coach?

但我们在想， 机器能否比教练懂得更多呢？

Can it know more than person could know?

它能比人类懂得更多吗？

And it turns out the answer is yes.

事实上，答案是肯定的。

The coaches want players to take good shots.

教练想让球员投出好球。

So if I'm standing near the basket

所以如果我站在篮筐旁边，

and there's nobody near me, it's a good shot.

周围没人，这就是好的投篮时机。

如果我站得远，而且被对方包围住， 通常来讲这球投不进。

If I'm standing far away surrounded by defenders, that's generally a bad shot.

But we never knew how good "good" was, or how bad "bad" was quantitatively.

但我们无法定量衡量这个“好”有多好， “差”有多差，

Until now.

但现在不同了。

So what we can do, again, using spatiotemporal features,

同样，我们能做的就是利用时空特性

we looked at every shot.

来分析每次投篮。

We can see: Where is the shot? What's the angle to the basket?

我们可以看到：在哪里投篮？ 投篮的角度是多少？

Where are the defenders standing? What are their distances?

防守方的站位？ 他们间的距离，

以及角度如何？

What are their angles?

For multiple defenders, we can look at how the player's moving

防守球员不止一名的情况下， 我们能够通过观察球员的移动

and predict the shot type.

来预测投篮类型。

We can look at all their velocities and we can build a model that predicts

我们可以根据他们的速度 建立一个模型，

what is the likelihood that this shot would go in under these circumstances?

预测在这些情况下，进球的可能性。

So why is this important?

为什么这一点很重要？

We can take something that was shooting,

因为我们可以通过分析投篮 这一单一行为得到

which was one thing before, and turn it into two things:

不同以往的两种信息：

the quality of the shot and the quality of the shooter.

投篮的质量，以及投手的质量。

So here's a bubble chart, because what's TED without a bubble chart?

我们可以看一下这个气泡图， 没有气泡图，还算什么TED呢？

(Laughter)

（笑声）

Those are NBA players.

这些气泡都是NBA球员。

The size is the size of the player and the color is the position.

大小代表球员的体型， 颜色代表他们的位置。

On the x-axis, we have the shot probability.

x轴代表投篮的命中率。

People on the left take difficult shots,

靠左的球员偏向勉强投篮，

on the right, they take easy shots.

靠右的球员会在有空当时才出手。

On the [y-axis] is their shooting ability.

Y轴代表的是投篮质量。

People who are good are at the top, bad at the bottom.

好投手在上面，较差的在下面。

So for example, if there was a player

举个例子，有一个球员的

who generally made 47 percent of their shots,

投篮命中率是47%，

that's all you knew before.

以前你只能知道这么多。

But today, I can tell you that player takes shots that an average NBA player

但如今，我能告诉你NBA球员投篮的

would make 49 percent of the time,

平均命中率是49%，

and they are two percent worse.

他还低了两个百分点。

And the reason that's important is that there are lots of 47s out there.

因为我们要在众多47%的 球员中选择一个。

And so it's really important to know

那么重点就在于要搞清楚

if the 47 that you're considering giving 100 million dollars to

让你支付了一大笔美金的人

is a good shooter who takes bad shots

到底是个经常勉强投篮的神投手，

or a bad shooter who takes good shots.

还是一个愿意空位出手的差投手。

Machine understanding doesn't just change how we look at players,

机器分析不只改变了 我们对球员的看法，

it changes how we look at the game.

也改变了我们看待比赛的方式。

So there was this very exciting game a couple of years ago, in the NBA finals.

几年前有一场很激烈的NBA总决赛，

Miami was down by three, there was 20 seconds left.

迈阿密落后三分，只剩20秒了。

They were about to lose the championship.

他们将要失去总冠军了。

A gentleman named LeBron James came up and he took a three to tie.

一位叫勒布朗詹姆斯的年轻人 上去想投个三分追平。

He missed.

但他没投中。

他的队友克里斯波什拿到篮板，

His teammate Chris Bosh got a rebound,

passed it to another teammate named Ray Allen.

传给另一个队友雷阿伦。

他投中了个三分，比赛进入加时。

He sank a three. It went into overtime.

They won the game. They won the championship.

最后他们赢了比赛，得了总冠军。

It was one of the most exciting games in basketball.

这是篮球比赛中 最激动人心的时刻之一。

And our ability to know the shot probability for every player

而我们能知道每个球员在每一刻的

at every second,

投篮命中率

以及抢到篮板的可能性，

and the likelihood of them getting a rebound at every second

这种能力是前所未有的。

can illuminate this moment in a way that we never could before.

Now unfortunately, I can't show you that video.

有点可惜， 我无法给大家展示这个精彩片段。

But for you, we recreated that moment

但为了在座的各位，我们在三周前的

at our weekly basketball game about 3 weeks ago.

篮球周赛上重塑了那经典一刻。

(Laughter)

（笑声）

And we recreated the tracking that led to the insights.

我们也重新加入了 电脑追踪数据的演示。

So, here is us. This is Chinatown in Los Angeles,

这就是我和同事们， 在洛杉矶的唐人街，

a park we play at every week,

我们每周都会去打球的公园，

and that's us recreating the Ray Allen moment

我们在重塑雷阿伦时刻，

and all the tracking that's associated with it.

所有的轨迹都与之相关。

So, here's the shot.

就是这个投篮。

I'm going to show you that moment

你们会看到这一经典时刻，

and all the insights of that moment.

以及这一刻背后都发生了什么。

The only difference is, instead of the professional players, it's us,

唯一的不同就是 我们取代了专业球员，

and instead of a professional announcer, it's me.

而我取代了专业讲解员。

So, bear with me.

大家请见谅。

Miami.

迈阿密。

Down three.

落后三分。

Twenty seconds left.

还有20秒。

Jeff brings up the ball.

杰夫带球。

Josh catches, puts up a three!

约什接球，三分出手！

[Calculating shot probability]

[计算命中率]

[Shot quality]

[投篮质量]

[Rebound probability]

[篮板球概率]

Won't go!

进不了！

[Rebound probability]

[篮板球概率]

Rebound, Noel.

诺尔的篮板。

传回给达丽亚。

Back to Daria.

[Shot quality]

[投篮质量]

Her three-pointer -- bang!

球进了——三分！

Tie game with five seconds left.

打平了，还剩5秒。

The crowd goes wild.

观众们沸腾了！

(Laughter)

（笑声）

That's roughly how it happened.

真实情况大概就是这样。

(Applause)

（掌声）

Roughly.

差不多。

（掌声）

(Applause)

That moment had about a nine percent chance of happening in the NBA

在NBA有9%的概率 会发生这样的时刻，

and we know that and a great many other things.

我们知道的还有很多。

I'm not going to tell you how many times it took us to make that happen.

我是不会告诉你们 我们尝试了多少次才成功的。

(Laughter)

（笑声）

Okay, I will! It was four.

好吧，我还是说吧，四次。

(Laughter)

（笑声）

Way to go, Daria.

达丽亚，三分球还得努力啊。

But the important thing about that video

但那段视频以及我们对

and the insights we have for every second of every NBA game -- it's not that.

每场NBA比赛的细微观察 并不是重点。

It's the fact you don't have to be a professional team to track movement.

事实上，你无需组建 一个专业团队才能追踪移动。

You do not have to be a professional player to get insights about movement.

你也无需成为专业运动员 去理解那些移动。

In fact, it doesn't even have to be about sports because we're moving everywhere.

而且，这不仅限于运动， 因为我们无时不刻不在移动。

We're moving in our homes,

我们在家里，

in our offices,

在办公室里来回走动，

as we shop and we travel

我们也在世界各地

throughout our cities

各个城市

and around our world.

购物旅行。

What will we know? What will we learn?

我们能发现什么？ 我们能学到什么？

Perhaps, instead of identifying pick-and-rolls,

或许，除了识别挡拆，

a machine can identify the moment and let me know

机器还能识别某些时刻，

when my daughter takes her first steps.

让我知道我女儿何时 迈出她的第一步。

Which could literally be happening any second now.

她现在随时都有可能学会走路。

Perhaps we can learn to better use our buildings, better plan our cities.

或许我们能合理地利用我们的建筑物， 更加好地规划我们的城市。

I believe that with the development of the science of moving dots,

我相信随着移动圆点这一科学的发展，

we will move better, we will move smarter, we will move forward.

我们能更好地移动， 更智能地移动，一路向前。

Thank you very much.

谢谢大家。

(Applause)

（掌声）