--- title: Using Python Webcrawling and Numpy to Build a Bad Bracket description: BrandVerity engineer John Rush tried building a bracket using a couple Python libraries. The results—while interesting—weren't exactly stellar. image: https://www.brandverity.com/hubfs/bracket_standings.png --- [![](https://www.brandverity.com/hubfs/bracket_standings.png) ](https://www.brandverity.com/blog/using-python-webcrawling-and-numpy-to-build-a-bad-bracket) # Using Python Webcrawling and Numpy to Build a Bad Bracket [John Rush](https://www.brandverity.com/blog/author/john-rush) Mar 29, 2016 **This is my first time doing a March Madness Bracket. I don't follow basketball - more of a Seahawks & Broncos fan.  I had about an hour before the deadline to get my bracket together, and being a developer I decided to play to my strengths. **   ### Theory: My bracket is based on 3 axioms: - Any team is skilled enough to defeat any other. (Upsets are possible) - You want the team with the most physical players. (Basketball is a physical game) - A player's weight is mostly muscle. (I've never seen a fat basketball player)   ### Scoring: Each player was given a physical score based on their weight divided by their height. All the players scores were averaged to get a team score. **Team with the best score wins the matchup. ‘Best’ meaning the highest average weight to height ratio.** ### Building the Bracket: ****We’re mostly a python shop so using requests and BeautifulSoup I built a simple crawler to scrape height and weight for all players and all teams in the NCAA. BeautifulSoup is *the* python tool to extract content, data, and pretty much anything on a webpage.****   | ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 ``` | ``` def table_parser(table): headers = [header.string for header in table.findAll('th')] results = [] for row in table.findAll('tr'): row_dict = {headers[i]: cell.string for i, cell in enumerate(row.findAll('td'))} urls = row.find_all('a', limit=1) for url in urls: row_dict['url'] = url['href'] results.append(row_dict) return results def fetch_player_data(team_list): for team_dict in team_list: if 'url' in team_dict: page_url = DOMAIN + team_dict['url'] + 'players' player_page = get_url(page_url) soup = BeautifulSoup(player_page) table = soup.find('table', {'class': 'tablesaw'}) player_data = table_parser(table) team_dict['players'] = player_data return team_list ``` | | --- | --- |   ****Then I used numpy to give each player a score and average all the players to determine team scores. Numpy is a very powerful scientific computing library. It's totally overkill for this project, but why not? There’s an office trophy at stake!**** Now I had a score for each team based on the full team name, but I then ran into a problem.   ### Data Filling: The ESPN site has only abbreviations for teams and the team logos. I didn’t know enough to go from the abbreviations to the full team names. (Did I mention I don’t follow basketball?) Luckily for me, ESPN put the logo next to each team. I scraped the bracket page for each logo/abbreviation combo, used google reverse image search and the ‘best guess’ to find the full team names. This let me build the bracket based on the scores. ![Team_Logo.png](https://www.brandverity.com/hs-fs/hubfs/Team_Logo.png?width=204&name=Team_Logo.png "Team_Logo.png")![best_guess.png](https://www.brandverity.com/hubfs/best_guess.png "best_guess.png")   ### Results: **Bottom of the BrandVerity Gr****oup - and in ESPN’s 6.6 percentile. Very Bad.** **![bracket_standings.png](https://www.brandverity.com/hubfs/bracket_standings.png "bracket_standings.png")** ### Analysis: The scoring method I used has a side effect - players who are shorter and weigh more score higher than taller, skinnier players. ****Turns out shorter, thicker teams do poorly in basketball. Who knew? ****   ### [We’re hiring!](https://www.brandverity.com/careers/) If webcrawling, big data and machine learning sounds appealing, get in touch! We are looking for passionate people with diverse backgrounds.   Topics: [Culture](https://www.brandverity.com/blog/topic/culture) - [Tweet](https://twitter.com/share) ###### Don't Miss Out Get the latest insights on brand protection, compliance, and paid search delivered right to your inbox. ###### Follow Us: [![](https://www.brandverity.com/hubfs/3.%20BRANDVERITY/Images/Icons/Updated%20Icons/Social/fb.svg)](https://www.facebook.com/brandverity) [![](https://www.brandverity.com/hubfs/3.%20BRANDVERITY/Images/Icons/Updated%20Icons/Social/twitter.svg)](https://www.twitter.com/brandverity) [![](https://www.brandverity.com/hubfs/3.%20BRANDVERITY/Images/Icons/Updated%20Icons/Social/linkedin.svg)](https://www.linkedin.com/company/brandverity) ###### Popular Posts - [What Is Trademark Bidding and Why Does It Matter](https://www.brandverity.com/blog/what-is-trademark-bidding-and-why-does-it-matter) - [Adwords URL Cloaking](https://www.brandverity.com/blog/23/adwords-url-cloaking/) - [New Guides to Make Sense of Trademark Bidding](https://www.brandverity.com/blog/new-guides-to-make-sense-of-trademark-bidding) - [Users Exploit Referral Programs with Ad Hijacking](https://www.brandverity.com/blog/1933/users-exploit-referral-programs-with-ad-hijacking/) - [How Do I Know If Trademark Bidding Affects Me?](https://www.brandverity.com/blog/how-do-i-know-if-trademark-bidding-affects-me) ###### Topics - [paid search](https://www.brandverity.com/blog/topic/paid-search) - [affiliate marketing](https://www.brandverity.com/blog/topic/affiliate-marketing) - [Events](https://www.brandverity.com/blog/topic/events) - [Search Engine Updates](https://www.brandverity.com/blog/topic/search-engine-updates) - [Trademark and Law](https://www.brandverity.com/blog/topic/trademark-and-law) - [Product](https://www.brandverity.com/blog/topic/product) - [Trademark Bidding](https://www.brandverity.com/blog/topic/trademark-bidding) - [Culture](https://www.brandverity.com/blog/topic/culture) - [News](https://www.brandverity.com/blog/topic/news) - [PPC](https://www.brandverity.com/blog/topic/ppc) - [Marketing Compliance](https://www.brandverity.com/blog/topic/marketing-compliance) ###### What you don't know will hurt you. Start monitoring and protecting your brand. [###### Request Demo ](https://www.brandverity.com/demo/) ![](https://dc.ads.linkedin.com/collect/?pid=28715&fmt=gif)