使用高级统计和分析细分NBA获胜情况(第1部分) (Using Advanced Stats & Analytics to Breakdown Winning in the NBA (Part 1))

If you’re anything like me or the millions of avid NBA fans around the world, watching the NBA restart has been a little different, to say the least. The games feel more like a televised scrimmage with performances reminiscent of a Drew League game. There is no doubt the competition is there, but given the times, it’s hard to determine how high the stakes are for each player and team. Needless to say, one part of the NBA Restart that has been (thankfullly) unapologetically familiar is the banter between long-time Inside the NBA frenemies, Shaq and Chuck.

至少可以说，如果您是像我这样的人，还是全世界数百万狂热的NBA球迷，那么观看NBA重新开始的感觉就有些不同了。 游戏的表现更像是电视转播，其表现让人想起了德鲁联赛的游戏。 毫无疑问，比赛在那儿进行，但是考虑到时间，很难确定每个球员和球队的赌注有多高。 毋庸置疑，NBA Restart一直(非常)自然而然地熟悉的一个部分就是长期在NBA内部的狂热分子Shaq和Chuck之间的戏ban。

The other night during the halftime show between the Portland Trailblazers and the Brooklyn Nets, the guys got into a heated debate over Portland’s defense, or lack-there-of. Despite putting on an offensive clinic in the first half, scoring a whopping 73 points, Charles and Kenny both emphasize the amount of points Portland has allowed Brooklyn to score — an equally alarming 67 pts. Kenny and Chuck both challenge Portland to step up on defense in the second half. Shaq doesn’t necessarily agree, noting Portland’s role players needing to step up on offense, which leads to another classic argument between the guys.

在波特兰开拓者队和布鲁克林篮网队之间的半场表演的前一天晚上，这些家伙就波特兰的防守或缺乏防守展开了激烈的辩论。 尽管上半年开设了进攻性诊所，获得了惊人的73分，但查尔斯和肯尼都强调了波特兰让布鲁克林得分的得分-同样令人震惊，为67分。 肯尼和查克都向波特兰发起挑战，希望在下半场加强防守 。 Shaq不一定同意，他指出波特兰的角色扮演者需要加强进攻 ，这导致了球员之间的另一个经典争论。

演示地址

争论 (The Argument)

Alright, there’s a lot here. On one hand we have a Hall-of-Famer breaking down the box score and simply saying, Portland needs to lock it down defensively. End of story. On the other hand, we have a different Hall-of-Famer telling us the other players need to step up and capitalize on their opportunities to score, specifically because Damian Lillard is being double teamed. Both arguments are fair, but who’s right in this situation? Rather, who is more right?

好吧，这里有很多东西。 一方面，我们有一个“名人堂”来打破比分，并简单地说，波特兰需要防守将其锁定。 故事结局。 另一方面，我们有一个不同的名人堂，告诉我们其他球员需要加强并利用自己的得分机会，特别是因为达米安·利拉德(Damian Lillard)正处于双队状态。 两种说法都是公平的，但在这种情况下谁是对的？ 相反，谁更合适？

If you take Chuck and Kenny’s argument at face value, you’d think they’re absolutely correct. The Nets averaged 111.8 points per game this season (119.8 ppg during restart) and were 22nd out of 30 in Offensive Rating. Furthermore, the Nets are fielding a lineup without Kevin Durant, Kyrie Irving, DeAndre Jordan, or Spencer Dinwiddie. So naturally, Terry Stotts should march into the locker-room preaching defense. However, with some analytical methods and the power of Python, we can begin to understand why this argument is more complex than it seems.

如果您以查克和肯尼的论点为准，那么您会认为它们是绝对正确的。 本赛季网队平均每场比赛得到111.8分(重新开始时为119.8分)，在进攻得分30分中排名第22。 此外，篮网在没有凯文·杜randint，凯里·欧文，迪安德烈·乔丹或斯宾塞·丁威迪的情况下也有阵容。 因此，特里·斯托特斯自然应该进军更衣室的宣讲防御。 但是，借助一些分析方法和Python的强大功能，我们可以开始理解为什么此参数比看起来更复杂。

相关性 (Correlation)

In order to breakdown this argument analytically, we’ll use correlation. In short, correlation is a statistical technique that can tell us whether pairs of variables are related as well as the strength of the relationship. That relationship can quantified by calculating the correlation factor between the two variables.

为了从分析上分解这个论点，我们将使用correlation 。 简而言之，相关性是一种统计技术，可以告诉我们变量对是否相关以及关系的强度。 这种关系可以通过计算两个变量之间的相关因子来量化。

Correlation factor ranges from -1 to +1:

相关因子的范围是-1至+1：

• factors closer to +1 denote a strong positive relationship,接近+1的因子表示很强的正向关系，
• factors closer to -1 denote a strong negative relationship,接近-1的因子表示强烈的负相关关系，
• while factors close to 0 denote a weak to no relationship.而接近0的系数表示弱关系或无关系。

It’s important to remember that just because a correlation score is negative, doesn’t mean there is no correlation. Generally speaking, correlation scores over 0.5 and under -0.5 indicate strong correlation. So when comparing the correlation between metrics, we’ll look at the magnitudes of their correlation scores.

重要的是要记住，仅仅因为相关分数为负，并不意味着就没有相关性。 一般而言，相关得分超过0.5且低于-0.5表示强相关。 因此，在比较指标之间的相关性时，我们将查看它们的相关性得分的大小。

数据分析 (Data & Analysis)

Using the NBA Season Summary stat tables on Basketball-Reference, I aggregated a dataset that included Team Misc. stats for the 2015–16 through 2019–20 seasons (up to March 11, 2020). What is nice about the Team Misc. stats section is that it already includes advanced metrics — like Offensive/Defensive Rating (Points Scored/Allowed per 100 Possessions), effective FG % (weighted adjustment for two pointers vs three pointers), and Pace (possessions per 48 minutes). These advanced metrics are normalized and will give us the best indication on what correlates to winning.

使用Basketball-Reference上的NBA Season Summary统计表，我汇总了包含Team Misc的数据集。 2015-16到2019-20赛季(截至2020年3月11日)的统计信息。 杂项团队有什么好处。 stats部分是它已经包含高级指标-进攻/防御等级(每100得分/允许的得分)，有效FG％(两个指针对三个指针的加权调整)和Pace(每48分钟的占有率)。 这些高级指标已标准化，将为我们提供与获胜相关的最佳指示。

First, we must make sure the dataset is normally distributed and continuous.

首先，我们必须确保数据集是正态分布且连续的。

I created a histogram on number of teams and W/L and as you can see above it follows a normal distribution pattern, meaning we can gain significant insights from this dataset. Because of this, we’ll use the Pearson correlation method to evaluate the linear relationship between. When we run this correlation on the W/L column and sort from highest correlation to lowest, we get this:

我创建了关于团队数量和W / L的直方图，如您在上方看到的那样，它遵循正态分布模式，这意味着我们可以从该数据集中获得重要的见解。 因此，我们将使用Pearson相关方法评估之间的线性关系。 当我们在W / L列上运行此相关性并从最高相关性到最低相关性进行排序时，我们得到：

I’ve made boxes around offensive and defensive metrics that have significant correlation factors. When comparing magnitudes, we see Offensive Rating (73%) is more correlated to W/L than Defensive Rating (67%). Furthermore, offensive effective FG (Off eFG%) correlates to W/L about 3.5% more than defensive effective FG (Def eFG%).

我在具有重要相关因素的进攻和防御指标周围打了几格。 在比较幅度时，我们发现进攻等级(73％)与W / L的相关性比防守等级(67％)高。 此外，进攻有效FG(Off eFG％)与防御有效FG(Def eFG％)的W / L相关性约为3.5％。

演示地址

So there we have it. The NBA is an offensive league, defense isn’t as important as offense, players don’t get paid to play defense, the game is all about getting buckets…Shaq is right. Right?

因此，我们有它。 NBA是一个进攻性联盟，防守并不像进攻那么重要，球员没有得到报酬参加防守 ，比赛就是为了赚钱……沙克是对的。 对？

Let’s dig a little deeper to get the full picture.

让我们更深入地了解整个情况。

深入了解 (Deeper Look)

Anyone who’s played organized basketball at any level has heard this before: Offense wins games, defense wins championships. NBA fans also know that the playoffs are a whole different game than regular season. So I went ahead and ran the same analysis for the 2015–19 Post-season playoffs.

曾经在任何级别打过有组织篮球的人都曾听说过：进攻赢得比赛，防守赢得冠军。 NBA球迷也知道季后赛与常规赛完全不同。 因此，我继续对2015-19赛季后季后赛进行了同样的分析。

Interestingly enough, all correlation factors drastically fall. However notably, Defensive Rating and Defensive effective FG% have stronger correlation factors than their offensive counterparts. So in a sense offense wins [regular season] games, but defense does win championships.

有趣的是，所有相关因素都急剧下降。 但是，值得注意的是，防御等级和防御有效FG％的相关因素比攻击性因素要强。 因此，从某种意义上说，进攻赢得了(常规赛)比赛，但是防守却赢得了冠军。

To be fair, there are several factors that we are over looking: there are less teams in the playoffs, teams are inherently “better” so metrics are skewed, there are less games, teams get eliminated, etc…But from this we can see winning in the playoffs looks different than winning in the regular season.

公平地说，我们要考虑的因素有很多：季后赛的球队减少，球队本来就“更好”，因此指标存在偏差，比赛减少，球队被淘汰等等，但是……由此可见季后赛的胜利看起来与常规赛的胜利不同。

结论 (Conclusions)

So who was right? In the clip above Ernie asks Shaq, “You’re not saying defense isn’t important to winning championships?” To which Shaq replies, “It is, it is. But Portland is not known for defense.”If you take this with his initial statement on role players making shots, you can argue that Shaq, the Big Aristotle, is in fact right. If Portland wants to win, that numbers tell them to focus on true shooting and effective FG %.

那谁是对的？ 在上面的片段中，厄尼问沙克：“您不是说防守对赢得冠军并不重要吗？” Shaq回答说：“是的，是的。 如果波特兰最初发表关于角色球员投篮的陈述，那么你可以说， 大亚里士多德沙克实际上是正确的。 如果波特兰想获胜，那么这个数字告诉他们要专注于真实的投篮和有效的FG％。

Fans of the show know Shaq will often justify his arguments with the fact that he has 4 rings and Chuck has none. Even though this is a low-blow — and frankly a cheap shot in most of these arguments — I think it’s interesting that most of the time there is statistical relevance to his points. He didn’t flat out deny the importance of defense, he emphasized the importance of not letting the double teams hinder the teams scoring — a take that requires one watch the game and not just the box score.

节目的粉丝知道，沙克经常会以他有4个戒指而查克没有戒指的事实来证明自己的论点是正确的。 即使这是一场低潮-坦率地说，在大多数这些论据中都是便宜的镜头-我认为很有趣的是，大多数时候他的观点与统计相关。 他并没有完全否认防守的重要性，他强调了不要让双支球队阻碍球队得分的重要性，这需要一个人观看比赛，而不仅仅是盒子得分。

Portland would go on to win the game, despite being outscored by the Nets in the second half. Down the stretch Portland would be lifted by a Gary Trent Jr. three-pointer (assisted by a double-teamed Dame Lillard), a Jusuf Nurkic layup (assisted by a double teamed Dame Lillard), another Jusuf Nurkic layup (assisted by a double-teamed CJ McCollum) and a CJ McCollum pull-up (no FGs for Dame Lillard final 7 minutes of game). Caris Levert’s missed 22-footer should not overshadow the fact Portland allowed the Nets to shoot 55.2% from the field (42.2% from 3P) while the Blazers were 48.1% from the field (33.3% from 3P).

尽管下半场被篮网队击败 ，波特兰仍将继续赢得比赛。 向下伸展波特兰将由加里遄小三指针(由包夹圣母院利拉德辅助)被提升，一朱素福·纳基奇上篮(由双辅助联手圣母院利拉德)，另一个朱素福·纳基奇叠层(由双辅助(由 CJ McCollum队)和CJ McCollum上拉(比赛最后7分钟没有FG进入Dame Lillard)。 卡里斯·莱弗特(Caris Levert)失手的22英尺长，不应该掩盖波特兰允许篮网的投篮命中率达55.2％(3P占42.2％)，而开拓者则占48.1％(3P占33.3％)。

下一步 (Next Steps)

In short, I think stats and analytics are a great way to breakdown the game. We were able to take a seemingly meaningless argument from a halftime segment, and find tangible insights from the data. It’s cool that the numbers can confirm feelings and thoughts players and coaches have about the game as well as uncover dynamics players and coaches would not be able to see having only watched the game and box score.

简而言之，我认为统计数据和分析是解决游戏问题的好方法。 我们能够从半场休息中获得看似毫无意义的论据，并从数据中找到切实的见解。 这些数字可以确认球员和教练对比赛的感受和想法，以及发现动态的球员和教练仅看比赛和得分都看不到动态，这很酷。

From here, we can go from a lot of different places. We can pull more metrics and make a correlation matrix to see how those metrics relate to each other.

从这里，我们可以从许多不同的地方出发。 我们可以提取更多指标，并建立相关矩阵以查看这些指标之间的关系。

Or we can pull in shooting stats see what they tell us about effective offensive FG%

或者我们可以获取投篮数据，看看他们告诉我们有效进攻FG％的情况

However, I think it’s important to understand that stats don’t tell us the full story. For all we know, Terry Stotts could have only talked about defense at halftime and that’s what made them buckle down at the end of the fourth quarter. Regardless, analytics helps us see the game in a new perspective and I’m excited to see how that continues to move the game further in the future.

但是，我认为了解统计数据不能告诉我们完整的故事很重要。 就我们所知，特里·斯托特斯可能只在中场休息时谈论防守，这就是他们在第四节末失败的原因。 无论如何，分析技术可以帮助我们以新的视角看待游戏，而我很高兴看到它在未来如何继续推动游戏的发展。

Enjoy the NBA Playoffs!

享受NBA季后赛！

Source Code: GitHub

源代码： GitHub