<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Alon Daks on Medium]]></title>
        <description><![CDATA[Stories by Alon Daks on Medium]]></description>
        <link>https://medium.com/@alondaks?source=rss-4ab76ad6f44c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*4sFeeQFf3Q1NyuSZVs4kdA.jpeg</url>
            <title>Stories by Alon Daks on Medium</title>
            <link>https://medium.com/@alondaks?source=rss-4ab76ad6f44c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 20:37:40 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@alondaks/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Predicting the 2017 NBA Finals]]></title>
            <link>https://medium.com/nba-data-science/predicting-the-2017-nba-finals-60d99bb86dff?source=rss-4ab76ad6f44c------2</link>
            <guid isPermaLink="false">https://medium.com/p/60d99bb86dff</guid>
            <category><![CDATA[2017-nba-playoffs]]></category>
            <category><![CDATA[nba-finals]]></category>
            <category><![CDATA[sports-analytics]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[nba]]></category>
            <dc:creator><![CDATA[Alon Daks]]></dc:creator>
            <pubDate>Thu, 01 Jun 2017 21:51:28 GMT</pubDate>
            <atom:updated>2017-06-22T06:09:23.866Z</atom:updated>
            <content:encoded><![CDATA[<p>By <a href="https://medium.com/@alondaks">Alon Daks</a> and <a href="https://medium.com/@nishantdesai">Nishant Desai</a></p><p>We estimate that the Golden State Warriors have an 84% chance of winning the 2017 NBA finals. This conclusion is drawn from extensive quantitative research into modeling how NBA teams and players have interacted in all 15,519 games since the beginning of the 2005–2006 season leading up to tonight. Our method consists of a principled combination of bayesian inference and deep learning: augmenting two powerful paradigms of machine learning that are driving advances in computer vision, natural language processing, and robotics among many other fields.</p><p>Although the Warriors have an 84% chance of winning the title, and therefore the Cavaliers have only a 16% chance, there are many ways the series could unfold. A best-of-seven series could occur in one of 70 possible ways. If the series is won in 4 games, there are two ways that could happen: WWWW and CCCC, where W indicates a Warriors win and C indicates a Cavaliers win. There are eight ways the series could end in five games, 20 ways it could end in six games, and 40 ways it could end in seven games. Estimating the Warriors or Cavaliers’s chances of winning the title requires first estimating the probability for each of the 70 possible series outcomes. Since enumerating all outcomes would be too lengthy for this post, we are publishing these probabilities in <a href="https://docs.google.com/spreadsheets/d/1A86-YvFnfigp-LYdJafmMgUioeUWqbdTQI0v4oLyEzI/edit?usp=sharing">this google sheet</a>. According to our model, the Warriors have a 29.3% chance of sweeping whereas the Cavaliers have only a 1.7% chance of doing the same.</p><p>As a consequent of computing the probabilities for each possible series outcome, we can calculate the probability that a team clinches the title in a given number of games. This distribution is displayed below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*_q0_aeoSJ9ttyxmAEv4KoA.png" /></figure><p>In order for the model to predict game outcomes, it first builds an understanding for how players will perform, and then learns how those players interact with one another. Having a player-centric method for prediction gives us the ability to quantitively assess any player’s contribution to his team’s success.</p><p>We wanted to see how much lift LeBron brings to the Cavaliers, so we ran estimates for the series supposing LeBron didn’t play at all. Unsurprisingly, the Cavaliers shot at the title is substantially worse without him. Their chances of winning drop from 16% to 9%. In this scenario the Warriors chance at sweeping go up to 39%. Full estimates for a world where he does not play is found on the <a href="https://docs.google.com/spreadsheets/d/1A86-YvFnfigp-LYdJafmMgUioeUWqbdTQI0v4oLyEzI/edit#gid=640346917">second tab of the google sheet</a>. Below displays the new distribution for each team clinching the title in a given number of games.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*RXbI1_tP56bxYk45Gg2CCg.png" /></figure><p>Next, we analyze the lift Stephen Curry brings to the Warriors by estimating the finals in the case where he does not play. In this scenario the Warriors chances drop from 84% to 77%, still leaving them in a safe position to avenge last year’s defeat. These results are summarized in the <a href="https://docs.google.com/spreadsheets/d/1A86-YvFnfigp-LYdJafmMgUioeUWqbdTQI0v4oLyEzI/edit#gid=1734881176">third tab</a> as well as the following distribution.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*trZsz6zKP5P-MdkBgggg9w.png" /></figure><p>N.B. Our model was trained on the 11 seasons ending in years 2006 through 2016. It’s test accuracy on the 1304 games that already occurred in the 2016–2017 season is 64.57%, with a <a href="https://en.wikipedia.org/wiki/Likelihood_function">log-likelihood</a> of -824.70.</p><p><strong>Edit 6/2:</strong> Conditioning on the game 1 win, we give Golden State a 90% chance of winning. This <a href="https://en.wikipedia.org/wiki/Conditional_probability">conditional probability</a> is computed by adding up all the outcomes where the Warriors win both the series and game 1 and dividing by the chance that they win game 1 (which we had at 78%). This is the updated distribution:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*sexN2p6bRAzRtaJZ1-ExwA.png" /></figure><p><strong>Edit 6/5:</strong> Following Game 2, the Warriors chance of winning the title stands at 95%.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/555/1*M0xCLyaX9HcjRbIHxF8emQ.png" /></figure><p><strong>Edit 6/8:</strong> Following Game 3, the Warriors chance of winning the title is 99%.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/555/1*DI1MtMv4DF2wtj1tbzfr3g.png" /></figure><p>Alon Daks (alon.daks@berkeley.edu) and Nishant Desai (nishantdesai@berkeley.edu)</p><p>UC Berkeley, EECS Department</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=60d99bb86dff" width="1" height="1" alt=""><hr><p><a href="https://medium.com/nba-data-science/predicting-the-2017-nba-finals-60d99bb86dff">Predicting the 2017 NBA Finals</a> was originally published in <a href="https://medium.com/nba-data-science">The NBA: A data-driven approach</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>