Sunday, October 21, 2018

NHLErrata launch

Ever since I started collecting the NHL data it was very important to me to validate the collected information. So I created a set of checks that finally formed a whole library to test both the NHL boxscore feeds and the HTML reports.

I managed to fish out quite a few errors and inconsistencies. Some where systemic, and could be fixed in software, some required a manual intervention, sometimes by editing the source file, sometimes by providing the correct value overriding what the parser would read. These interventions, classified into a variety of types, formed another library.

But I also wanted to share the errors that I found and the fixes I figured out with the analytics community. So I decided to create a website dedicated to it. It took a while, but finally a couple of days ago I was able to open There you can find:

  • An overview of data sources.
  • Information on missing players and events
  • Information of broken reports, players and events
  • Systemic problems encountered with the reports

Both mentioned libraries, and are part of my scrape-to-database package on CPAN.

Monday, October 15, 2018

New features for the 2018/19 season.

After a period of deep dormancy we welcome the 2018/19 on our website with three major changes:

1. MoreHockeyStats is now a network of sites:

  • - the unusual statistics from the NHL for the league, the teams, the players, the coaches and the draft.
  • - the Elo ratings for teams, coaches and players, in general and for particular stats and situations.
  • - the errors we discovered while crawling and testing the NHL data.
2. All our data is now available through a sort of API. When a table is displayed, there is a set of links on how you can access the data:

  • JSON - receive the data displayed in the table as a JSON via direct link. 
  • CSV - receive the data in a CSV table.
  • Table - display the data as a pure HTML table with a direct link.
  • Link - direct link to the results displayed.
These links have a systemic structure and thus can be crawled by a bot.

3. Most of our tables now feature links to personal cards for teams, players, coaches and even the rinks. These cards display the visual changes to the stat displayed in the table for the particular team, player etc. We welcome ideas for the cards in the pages that do not have these cards yet. Here's a sample playercard:

Welcome, and have a great 2018/19 hockey season.