From The Washington Post Engineering team.

How The Washington Post Estimates Outstanding Votes for the 2020 Presidential Election

By Lenny Bronner, Jeremy Bowers and John Cherian

The Washington Post uses expected vote models to estimate how many votes remain for each candidate on election night. We built distinct models for the 2019 Virginia general election and the 2020 Democratic primaries. Today, we would like to introduce our third model, which we will use to estimate remaining votes during the 2020 general election.

How The Washington Post Generates Vote Flow Diagrams

By Lenny Bronner and Jameson Quinn

Several times during this election cycle, The Washington Post has produced estimated vote-share change diagrams. In this post we describe the pitfalls that came with our original diagrams and explore a method that mitigates these issues.

How The Washington Post Estimates Outstanding Votes During Primaries

By Lenny Bronner

For the primaries in 2020, The Washington Post has developed a model that estimates the candidates vote counts based on early results. These estimates update throughout the night as new results come in. We do this by comparing the current results to the last comparable election and see what the voters may have done this time. We also compute uncertainty estimates for our model. Read more about how our model works and where its pitfalls lie.

How The Washington Post Estimates Turnout on Election Nights

By Lenny Bronner

For the Virginia general election in 2019, The Washington Post has developed a model that estimates turnout for each state House and state Senate district before the election begins and then updates these estimates throughout the night as precincts report. We collected data from previous elections that most closely resembled this election night and used these to form our base estimates for turnout in each state House and state Senate district. A novel feature of our model is that it should respond to precincts that have already reported and update its estimates based on this new information.

Characterization and Early Detection of Evergreen Articles

By Shuguang Wang

The life spans of most news articles are very short, but a very small number of them maintain a timeless quality and are consistent of interest to the public. At the Post, we analyze the characteristics of these articles and has developed a model to automatically identify these so-called, evergreen articles.

Our New Engineering Blog, PostCode

By Leo Ji

Welcome to PostCode, the new engineering blog at The Washington Post. We’ve given our blog system a bit of refresh to take advantage of new systems in Arc Publishing, our publishing platform.

The Ad Performance and Safety Protocol: Cleaner, better, safer ad experiences on iOS

By Aram Zucker-Scharff

The Washington Post is taking a proactive approach to digital display advertising issues with a new project that is set to fix bad code, while saving brands from code errors and cross-platform bugs in ad code. The Ad Performance and Safety Protocol (APSP) focuses on accidentally harmful ad behaviors to ensure a better user experience for all readers on The Post’s Classic iOS App.

The Washington Post Among Initial Customers to Use Amazon DocumentDB

By Austin Graff

This week, Amazon Web Services announced Amazon DocumentDB and The Washington Post is among the first to use it. Below is the official announcement, including a quote from Washington Post Director of Data Science and AI Patrick Cullen. Today, Amazon Web Services (AWS), an company (NASDAQ: AMZN), announced Amazon DocumentDB (with MongoDB compatibility), a fast, scalable, highly available, and fully managed document database service that supports MongoDB workloads.

Recap: 2018 Internship and Hackathon

By Austin Graff

Technology is rapidly evolving, forcing industries to foster a posture of learning. One way we do this at The Washington Post is invite a group of interns to join us each summer. As they learn from us, we often walk away having learned more from them, the next generation of innovative minds.